fmtDateTime's time-only branch appended '00' for seconds no matter what it was
given. The booking import stores what KItinerary hands it, which is '10:00:00',
so the value came out 17 characters long. dtLine only attaches the TZID to a
well-formed 15-character value, so those events lost their zone and went out as
floating times, which is the #1453 regression by another route.
It pads the way the ISO branch above it already does, so every caller is covered
rather than only the rental markers.
Rentals now emit two additional timed markers, Pickup and Drop-off, the
same additive shape hotels already get for check-in/check-out (#1721,
last unported piece). The rental's existing single-block event is
untouched.
Sides are derived in priority order: the endpoint role (from = pickup,
to = drop-off, reusing the already-loaded endpoints map); if no
endpoint on the reservation carries either role, positional first/last
by sequence (only once there's more than one endpoint, so a lone
endpoint is never both sides); otherwise reservation_time /
reservation_end_time, since imports frequently carry only that window.
Per side, an event only fires once a usable date and clock resolve;
zone resolution prefers the endpoint's stored timezone, then its
coordinates, then the reservation's linked place.
Follows the hotel `marker` closure's pattern closely (same esc/uid/
dtLine/isDate/isTime/usedZones usage, same Pickup:/Drop-off: summary
convention, same per-kind UID shape) without touching the hotel path
itself.
The variables only take effect when Node runs with NODE_USE_ENV_PROXY=1, and
that is set in the image. This page also covers Helm, Unraid and source
installs, where nothing sets it, so following it there changed nothing.
Also recommend NO_PROXY. Without it the proxy sees every request TREK makes to
itself, including the container health check.
Review of the previous commit surfaced that the departure-time lookup keyed a
trip-wide coordinate index, so a POI revisited on another day (a supported
pattern) leaked the other day's time into the prefill. Resolve the departure
time from the leg's own day instead; the name index stays trip-wide (a name is
coordinate-stable).
Also from review: wire the per-leg option on the mobile sidebar too (it was
desktop-only), drop the now-dead route-leg fallback in the outgoing menu opener,
and accept single-digit hours in the time normalisation. Adds a regression test
for the revisited-stop case and pins the hotel-bookend fallback time.
The per-segment travel-mode menu now carries a "Public transit" entry on every
editable connector: stop-to-stop legs and the boundary hotel/booking bookends
alike. Choosing it opens the existing transit search pre-filled with the leg's
origin, destination and the stop's departure time, instead of drawing an OSRM
road route.
A connector only renders when both endpoints are located, so the option is
inherently limited to legs that can actually be searched. A leg's coordinates
are resolved back to named, timed points from the trip's places, accommodations
and booking endpoints. TransitSearchPanel gains an initialTime prop so the
departure time can be seeded.
Dependabot cannot be pointed at dev. \ arget-branch\ in dependabot.yml applies to
version updates only, and security updates always open against the default branch.
Security updates are all this repository gets, since there is no dependabot.yml at
all, so every one of them arrives on main.
The branch check then labelled them wrong-base-branch and close-stale-wrong-branch
shut them 24 hours later. Thirteen of the seventeen Dependabot pull requests died
that way; the four that survived were retargeted by hand. The comment the check
leaves asks the author to click Edit, which a bot cannot do.
So retarget them instead. If that fails, because the branch was cut from main and
dev has moved on, it falls through to the existing path so a human sees it rather
than the check quietly passing.
Note this only takes effect once it is on main: pull_request_target loads the
workflow from the pull request's base branch, and these arrive on main.
* fix(mcp): accept per-leg times on create_transport and update_transport
metadata was pinned to Record<string, string>, so a legs array could not get
through and a multi-leg flight built over MCP lost every intermediate time. The
booking ended up with correct endpoints and no legs, which every reader then
treats as a single segment.
legs is now a real, typed parameter with its contract in @trek/shared, shaped
exactly the way the planner form and the importers write metadata.legs. The
folding checks what the readers depend on: one leg fewer than endpoints (the
AirTrail dedupe builds its signatures on that), day ids that belong to the trip,
airport codes that agree with their endpoint, and the times, where a stopover
endpoint carries the departure and the last one the arrival.
metadata keeps its type and its replace semantics. Segments smuggled through
metadata.legs as a JSON string used to be stored and then ignored by everything
that read them; they are refused now and point at the legs parameter instead.
Closes#1914
* fix(atlas): refuse a bucket list entry that is already on the list
Nothing gave a bucket entry an identity, so the same country could be added as
often as it was clicked. A wish is now identified by name, country, target date
and, where it has them, coordinates. A different target date is a different wish
and still lands, which is what the report asked for; two real places sharing a
name stay separate because their coordinates differ.
Empty string and NULL are normalised on both sides of the comparison, otherwise
the dateless case, which is the one that was reported, would slip through.
Deliberately no unique index and no migration. better-sqlite3 is synchronous and
the server is a single process, so the race an index would guard against cannot
happen here, and an index would have needed a destructive dedupe of existing
rows first.
create_bucket_list_item had no target_date parameter at all, so over MCP the
allowed "same place, other date" case was unreachable. It takes one now. The
plugin RPC reports the duplicate as bad params instead of letting it surface as
an internal error.
Closes#1898
* fix(settings): keep a free-form LLM endpoint in admin hands
Any account could name the address the server calls, and the guard in front of
it deliberately allows loopback and LAN so a self-hosted Ollama works. That is a
reasonable trade for whoever runs the instance, because the network is theirs. It
is not one for an arbitrary account, which could aim a request from inside that
network at a target only it chose and read how the target answered.
A free base URL, and the local provider that is one by another name, now come
only from an admin controlled source: the addon config, the instance-wide
defaults, or an admin's own settings. OpenAI and Anthropic stay available to
everyone, because they go to a fixed address with the user's own key.
The resolver is where this is enforced, since it is the single point both the
booking import and the plugin host pass through, and the only one that also
covers values already sitting in the database. The write path refuses as well,
so nothing is stored that would never be read.
The hint text promised something users no longer have, so it changed with it.
Closes#1772
* fix(calendar): span a date-only booking to its end date
A DATE event without DTEND is exactly one day per RFC 5545, so a multi-day
booking that carries an end date still collapsed onto its first day. It gets an
exclusive DTEND now, built the same way the stay branch already builds one. A
booking without an end date serialises byte for byte as before.
While in here: the stays query joins reservations on the accommodation, and two
bookings pointed at the same stay fanned it out into two events sharing one UID.
Duplicate UIDs in a feed are undefined behaviour and calendar clients pick one at
random.
The same "at most one booking per stay" assumption sat in the accommodations
service, where it was worse than cosmetic: accommodation_id carries no foreign
key, so deleting a stay took only the first booking and left the second pointing
at something that no longer existed. Both the sync and the delete walk all of
them now. The callers keep reading linkedReservationId; the list is additional.
Refs #1869
* fix(vacay): say that the region is required, because it is
The placeholder called the region optional while the add button stayed disabled
until one was picked. The wording predates multiple calendars per plan, which is
what introduced the gate. Only the value of the existing key changed, in all
locales, so the key set stays identical.
Two things the same screen got wrong were in the way of a clean answer:
The button was clickable while the region list was still loading, which put the
user in exactly the state the wording describes, and it is the state the report
was filed from. It waits for the list now.
Changing the country on an existing calendar kept the previous country's regions
on screen, so a German region could be picked for the Netherlands. The list is
cleared with the country, which only works in combination with the loading gate:
without it, the empty list would happen to enable the button instead.
Closes#1813
* fix(collections): let a saved place's address be corrected
The edit form offered name, category and labels, and the address only on the way
in. The cause was not the form: the update contract never declared address, so
the validation pipe stripped it and the UPDATE had no column for it. Coordinates
were pulled through for #1435 and the address was missed.
It travels the whole chain now, contract to service to both forms. No migration,
the column has always been there.
The mobile add sheet had no address input at all, showing whatever the search
result carried as read-only, so editing would have worked while creating still
would not. It has one.
The MCP tool description had drifted further than the address, never having
mentioned the coordinates either. Both are in it now.
Closes#1870
* fix(collections): keep the place search usable after the first save
The add dialog stays open on purpose so several places can be added in a row, but
after a save it gave no way back into the search. A search with no results was
silent, with no empty state and no toast, so typing another name and pressing
enter looked like nothing happened. The reporter described both halves and it is
the silent search that matches "nothing happens" most closely.
There is an empty state now, using a key that already existed in every locale and
was unused. The search row is sticky inside the dialog body, so it stays reachable
after the save instead of sitting above the scroll position.
Also fixed while in here: refreshing the collection after each add reset the
select mode, the selection and the label filter. In the very flow this dialog is
built for, the user quietly lost all three.
Closes#1921
* fix(vacay): make the phone year grid and the dock button do what they look like
Two separate faults met on this screen.
Tapping a day in the year grid did nothing unless the edit mode was already on,
and the only way into it was a floating button that looks exactly like the one
the dock draws in the same spot. A tap now opens that month in the editor, where
the cells are big enough to hit. At 390px the grid cells are around 21px, which
is too small to be mutating data with in the first place.
The dock's plus was not wired for this route, so it fell through to creating a
trip on the dashboard. Two different actions were stacked in one place, and the
wrong one was on top whenever the screen had not painted yet, which includes the
time its chunk is still loading. The dock leaves the slot to the screen now, and
holds the geometry so the tab groups stay symmetric.
Closes#1811
* fix(budget): settle in whole cents so balances stop drifting
The report is a balance that moves on its own, with nothing added, edited or
removed, on a trip where everything is in euro. Two mechanisms were behind it and
both are real.
The display currency is a fourth value, a per-user setting, so "everything is in
euro" never covered it. Each balance was converted and rounded on its own, and
the set no longer summed to zero: at 0.855 and 0.9312 it came out one cent over,
at 0.94 it came out even. That cent appears and disappears with the six hour rate
cache alone, which is exactly the symptom. The flows were rounded separately from
the balances, so the two never had to agree.
And a receipt line nobody was ticked for counted toward the total, and therefore
toward the payer's credit, while landing in no share. That difference is
permanent, and the client sent it that way.
The settlement converts once, at the edge, and does integer arithmetic after
that. Equal splits distribute the credited cents by largest remainder rather than
the entered total, so credits and debits of one expense cancel exactly even in a
foreign currency. Display conversion goes through the same largest-remainder
allocation, so the sum survives it.
With that, the 0.01 epsilon can go. It could never create a cent, but it stranded
them: a one cent balance was filtered out of the flows while staying visible in
the list, which is the "settle up offers nothing" half of the report.
Closes#1382
* fix(journey): reach the book export and the suggestion toggle on a phone
Below 1024px the hero card is dropped, and it held the only way to either action.
The floating bar that replaces it was never given them. It has them now, and the
hero is unmounted rather than hidden, so there is no second invisible copy of the
same controls in the document. The bar's own buttons moved onto the appearance
tokens while they were being touched.
The phone screen never had either action at all and now carries both.
The book itself was hardwired to English, cover and dates included, while the
keys for it had been sitting in every locale unused. It takes the page's
translator and locale, sets the document language, and prints mood and weather,
which were on screen but never in the book.
Closes#1848
* fix(journey): offer the connected photo providers in the phone entry editor
The desktop editor grew an external photo source in #1632; the phone editor that
came out of the mobile rewrite never did, so editing an entry on a phone had no
way to Immich or Synology. It mirrors the desktop mechanics, including the part
that is easy to miss: a save that creates the entry and then fails on the photos
keeps the sheet open, and without remembering what was just created the retry
creates a second entry.
The sheet no longer kicks off its own addon load. The list is loaded once on
boot and this only reacts to it; a second load races the other consumers and
overwrites the store when it lands.
Closes#1808
* fix(mobile): let the document scroll so Safari can collapse its URL bar
iOS Safari only shrinks its address bar when the document scrolls, and TREK's
never did: the root was locked and the mobile shell put a scroll container inside
it. On the bottom tab layout that costs about a sixth of the viewport,
permanently.
Below 768px the document is the scroller again and the full-bleed screens measure
their own height instead of leaning on the shell.
That makes the body scroll locks real for the first time on a phone, and they
were not ready for it. Three of them reset the overflow unconditionally, one of
those on every run of its effect rather than on unmount, so a system notice could
hand the page back while a sheet was still open. They all go through one
reference counted lock now, which is the part that had to land first.
Worth knowing for the issue: this reaches the flow screens. The trip planner, the
journal, the atlas and the vacay screen keep their own full-height scrollers and
still will not collapse the bar.
Closes#1809
* fix(mobile): keep the screen gradient visible behind the translucent cards
The gradient layer introduced with the scroll change sat at z-index -10. A
negative layer paints above the canvas but below the backgrounds of ordinary
blocks, and body carries an opaque one, so the gradient was hidden behind it.
The mobile cards are translucent and take their depth from what is behind them,
so every surface ended up compositing against a single flat colour and the whole
phone UI read as one tone.
The layer is positioned at z-index 0 now and the content above it at 10, which
puts both over body's background. Same visual as before the scroll change, and
the shell still owns no scroll container.
* test(reservations): teach the accommodation delete stub about the id lists
The RPC walks every linked booking since the delete stopped stopping at the
first one. The stub still answered with the two singular fields, so the loop had
nothing to iterate and the cascade events never fired.
* fix(settings): move the free-form LLM endpoint out of user settings entirely
Hiding it from non-admins closed the hole, but it left the option sitting in the
admin's own personal settings, which is not what was asked for and not where it
belongs: an instance has exactly one address it calls out to, and there was a
second, invisible place to set it.
The provider list is now the two hosted ones for everybody, the endpoint field is
gone, and the resolver takes the address from the admin-controlled instance
defaults for every caller including an admin. It no longer reads the users table
at all. The write path refuses the keys for everyone, so nothing can be stored
that would never be read.
The endpoint is configured on the addon in the admin settings, which already
carries provider, model, base URL and key, plus the Ollama model list. Nothing
is lost, it just has one home now.
Behaviour change worth a release note: an admin who set a personal Ollama
endpoint has to move it to the addon config. A non-admin was already in that
position after the previous commit.
Refs #1772
* fix(journey): make the book preview usable on a phone
The preview document carried no viewport, so a phone laid it out at its default
980px while the page is A4 landscape and about 1123px wide. It spilled out of
the frame, which reads as being zoomed far in. Declaring the real width lets the
browser scale the page to fit. Printing is unaffected, it goes by @page.
The preview chrome had the same problem in miniature: a title and two labels on
one row, sized inline, which does not fit a phone and fits translated labels
even less. It moved to a stylesheet, and below 640px the card goes full-bleed,
the decorative title steps aside and the two actions share the width.
Refs #1848
* fix(mobile): stop the shell trapping the screens' own overlays
Lifting the background layer into the positioned range meant the content wrapper
had to be positioned too, and a positioned wrapper with a z-index is a stacking
context. Every fixed overlay a screen renders inside it got clamped to that
level: the vacay view/edit FAB sits at z-50 precisely so it beats the dock at
z-40, and it disappeared behind it.
The layer goes back to a negative z-index and the wrapper back to no positioning
at all. What it needed in the first place was for body to stop painting over it,
which is one line in the phone media block rather than a new stacking context.
* fix(mobile): let the phone background override actually win
A media query adds no specificity, so the themed body rule further down the file
kept winning on source order and painted over the shell's background layer. The
override is on html body now.
The gradient was back to being invisible, which is the flat single-tone phone UI
again.
* fix(vacay): draw today's ring around the day instead of replacing it
On the phone, today short-circuited the day-cell matrix and returned a ring on an
empty cell. A vacation day or a company holiday logged for today was stored and
counted in the totals, but the cell showed nothing, so it read as if the tap had
not taken.
Today is an annotation now, layered like the school band and the shared-calendar
rings, which is how the desktop month card has always treated it. The strong ink
stays only where the cell has no fill of its own, since --m-ink is light in dark
mode and would disappear on a pastel.
* fix(map): stop serving a stale GL style document from the service worker
The Workbox rule for Mapbox matched every URL under api.mapbox.com, and the
style document lives there too. Under StaleWhileRevalidate the map was built
from the previous revision on every load, with the refresh landing one load too
late, so a style republished in Studio with different label languages kept
rendering the old ones. The network tab showed a 200 throughout, because the
service worker's own answer is a real 200, which is why this looked like the app
rewriting text-field at runtime.
The style now has its own rule ahead of the tile rules, on NetworkFirst so it is
current whenever the network is, and still cached so the map opens offline. Its
own small cache too: sharing an LRU with three thousand tiles made the staleness
come and go at random. OpenFreeMap styles had the same problem and are covered
by the same rule.
Closes#1924
* feat(places): link an expense to a place, the way bookings already do
Recording what a museum ticket or a guided tour cost meant leaving the place,
opening the Costs tab and typing it in again — while a booking has had a Costs
block in its own form all along.
The place form now carries that block. Same component, same flow: with nothing
linked it offers Create expense, which saves the place and then opens the Costs
editor for a new expense pointing at it; once linked it shows the expense with
edit and remove. budget_items gains place_id beside the reservation_id it
already had, with the same nullable FK and the same ON DELETE SET NULL. Desktop
and phone both, and the mobile editor carries the link through its own save path
rather than the desktop one.
The expense belongs to the place, not to a day assignment, so a place sitting on
five days is not five tickets. Hanging it off the assignment would have meant
adding a place to three days quietly creating three expenses and removing a day
quietly deleting one.
Deleting a place takes its expense with it in the same transaction and announces
it, on every path that deletes a place: REST, both MCP tools and the plugin RPC.
Closes#1298
* test(places): teach the RPC stubs about linkedExpenseIds
The delete paths read the linked expense before the place is gone, so the two
hand-built PlacesService stubs threw on the new method instead of reaching the
hook, the remove and the broadcast — four failures out of one missing line.
* fix(places): match imports on the provider id so renamed places are not duplicated
Re-importing a Google Maps list duplicated every place the user had renamed.
buildDedupSet only ever collected names and, for unnamed places, coordinates,
and isPlaceDuplicate returns on the name before it reaches the coordinate
fallback, so a renamed place was findable by neither. The ids were stored all
along, and even backfilled onto matched places, they were simply never read.
Match on google_place_id / google_ftid / osm_id first, keeping name and
coordinates in their old roles below it. Widening the coordinate check instead
would have merged the restaurant and the bar in the same building.
The collection copy carried a hand-mirrored port of the same predicate; it now
uses the shared helpers and gets the fix with them.
Closes#1550
* feat(calendar): put accommodations into the subscribed feed as all-day stays
A hotel keeps its dates on the linked stay, not on the reservation: the booking
form writes reservation_time = NULL for type 'hotel', and day_accommodations
carries start day, end day and the check-in/out clock. The feed only ever read
the reservation, so a stay showed up once on the arrival day at best, and not at
all when it came from the booking form.
Join the stay in and emit the reservation as an all-day range over every night,
which also keeps it correct when days are reordered later — the dates now come
from the day rows rather than from the copy stamped at creation time. Check-in
and check-out become their own timed events when the stay records a clock,
because an all-day event cannot carry the time you actually need reminding of.
The check-in window, when set, becomes that event's DTEND.
Closes#1586
* fix(oauth): treat a concurrent refresh as a race, not as a stolen token
Refresh rotation is a race by construction. Several MCP clients share one
refresh token, both post it within the same second, and the loser was read as a
replay: chain revoked, MCP sessions torn down, browser window asking to log in
again. That is the daily re-authentication people were seeing, not a short
lifetime — access tokens last an hour and refresh tokens roll for thirty days.
A revoked refresh token now only counts as a replay once it is outside a
30-second grace window, and only when the rotation that revoked it left a
successor still alive. Inside the window the caller gets its own pair branched
off the same parent, so both clients walk away with a working token. Explicit
revocation leaves no live successor and a chain killed by a real replay has none
either, so neither can slip through the window.
Closes#1007
* feat(collections): mark a place visited from the trip it belongs to
A place saved in several lists had to be opened in each of them to say you had
been there, which is the wrong way round: you find out you have been somewhere
while looking at the trip, not while browsing the library.
The save picker now shows each list's own status and cycles it in place, with one
action for "visited, wherever this is saved". The planner's selection bar and the
mobile bulk toolbar carry the same thing for a whole selection — select all, one
tap, done. Matching happens on the server with the signals findMembership already
uses plus the source link a place saved from this trip carries, so a copy renamed
inside a list is still found. Lists you may only read are skipped rather than
refused; a shared list you cannot edit should not stop you marking your own.
Closes#1469
* feat(journey): jump to the top or back to the last entry in a long journal
Everything that adds to a journal sits at the top — the entry button, the
gallery upload — and where you left off sits at the bottom, so adding one photo
to a long journey meant scrolling the whole way twice.
Two buttons float over the middle of the feed, clear of the Add Entry column on
one side and the reorder arrows on the other, each appearing only when there is
more than a screenful to travel in that direction. They live in a zero-height
sticky box so they take no layout space and never push the timeline around.
While in the header, the status pill also picks up the height of the round
buttons beside it.
Closes#1088
* fix(packing): give the item name the row back on phones
An item name had roughly a third of the row to itself: a spelled-out sharing
sentence, an empty weight column and two round action buttons took the rest, so
anything longer than a word was cut off.
Sharing is now the icon it always was next to a label rather than a sentence in
the layout, an unset weight prints nothing instead of a dash, and edit/delete
sit behind one overflow button that opens the same inline menu the bag picker
already uses. The to-do list needed none of this — it wraps its chips onto a
second line already.
Closes#1525
* docs(wiki): cover the six community changes
- Calendar feeds: accommodations as an all-day range plus check-in/check-out
- Collections: setting the visited status from a trip, and how places are matched
- Journey: the jump buttons on a long journal
- Packing: what the phone row keeps and what moved behind the overflow button
- Places: re-importing a list no longer duplicates renamed places
- MCP setup: the rotation race no longer ends an OAuth session
* test(collections): assert the stored status instead of a private method
tests/ sits outside the build tsconfig, so reaching into getPlaceById only
showed up in the dedicated typecheck step. Reading the row is the better
assertion anyway — it pins what was persisted, not what a getter returns.
The bucket-list place search opened its result list upward with
position:absolute inside the desktop Atlas panel, which is overflow-hidden
for its rounded corners. The panel leaves ~125px above the input, so the
top of the 160px list was painted outside the clip box — the first result
was invisible and unreachable, since that is the container's own top and
not a scroll offset.
Portal the list to the body and position it fixed off the input row,
flipping up or down by available room and capping its height to that room.
This matches CustomSelect, whose month/year dropdowns already escape the
same panel this way.
Fixes#1899
Clicking a logged day that a weekend-days change had turned into a weekend
day was swallowed by the client, while the entry kept counting against the
entitlement. The server already allows the removal (only conversions and new
inserts are refused on a blocked day), so mirror that: a blocked weekend cell
holding the acting person's entry stays clickable and clears it with the
entry's own fraction/kind. Logging a new weekend day stays blocked.
Fixes#1897
* feat(journey): serve the GPX tracks behind a journey
A journey has no trip of its own, but every entry records the trip it came from,
and a trip's routed geometries live on its places. So the tracks are derivable
without any new storage: GET /journeys/:id/tracks walks that link and returns the
geometries, dropping elevation nothing on the map reads and skipping anything
that is not a usable line.
* feat(journey): draw those tracks on the journey map
Both renderers, Leaflet and GL, plus the mobile map-timeline that shares them.
Solid and in the track's own colour with a white casing, so it reads as a
recorded route rather than as the dashed line that merely connects entries in
time order, and stays legible on satellite tiles.
* feat(journey): load the tracks alongside the journey
Fetched separately from the journey itself: most journeys have no tracks, and a
failed lookup should cost the map its lines rather than the whole page.
* docs(journey): document the GPX tracks on the map
* fix(journey): label tracks with the house map tooltip
bindTooltip's default box ignores the appearance tokens, so it sat there in
Leaflet grey next to markers already using .map-tooltip. Same class now, and an
unnamed track gets no tooltip at all instead of an empty one.
* refactor(atlas): drop the last-trip tile from the stats bar
The row beside the continent counts is statistics; a single trip sitting in it
was neither, and the trip it named is already one click away in the country
detail below. Its two locale keys go with it, along with atlas.lastVisitLabel,
which nothing had rendered for a while.
* feat(atlas): resolve a coordinate to its country and region
Groundwork for searching the atlas by place: /locate answers with the admin1
region a point falls in, resolved against the same bundled polygons the visited
colouring uses, so the answer is always something the map can highlight. A
country without admin1 coverage resolves to a country and a null region, which
is a real answer rather than a failure.
* feat(atlas): search the map by place, not only by country
Typing a city geocodes it through the same /maps/search everything else uses and
offers the hits below the instant local country matches. Picking one flies there
and opens the dialog for the region it sits in, so Milan marks Lombardy without
the user knowing that is where Milan is.
The country list stays offline and instant; only the place lookup waits on the
network, debounced and sequence-guarded so a slow answer cannot overwrite a
newer query.
* fix(atlas): count cities instead of shop names
The visited-cities figure scanned the address left to right and took the first
part that was not a country and matched a latin-only pattern. The first part of
a geocoded address is the place itself, so a trip to Tokyo counted Shibuya Sky,
Shibuya Parco and Family Mart as three cities, while an address written in
Japanese, Cyrillic, Greek or Korean counted none at all.
Read from the administrative end instead, skipping the country, anything with
digits, and the region already resolved for that place in place_regions, which
is what separates the city from the region above it. On one real trip the count
went from 20 entries, nearly all of them streets and shops, to Paris, Shibuya,
Minato and three Berlin boroughs.
* i18n(atlas): translate the place-search heading
* docs(atlas): document place search and how cities are counted
* fix(atlas): stop the search panel closing itself while results arrive
The dropdown closed on mouseleave, which was survivable while it held only local
country matches and appeared complete in one go. Geocoded places land a moment
later and grow the list under a cursor already resting there, so the panel shut
itself seconds after opening. Closes on a click outside now, Escape unchanged.
Two suites under client/tests/ still looked for "Export Calendar" by accessible
name. They live outside src/, which is why the targeted runs while building this
missed them and only CI caught it.
* feat(places): hand a trip back as GPX
TREK reads GPX but never wrote it, so a planned trip could not reach the offline
map or the handheld it was planned for. The importer already fixes the mapping:
a wpt is a place, a rte or trk is a place with route_geometry, so writing is that
reversed. Days have no import counterpart and become routes of their stops in
order, which is the part that puts a plan on a device.
XMLBuilder comes from the same fast-xml-parser the importer parses with, so
nothing new is pulled in.
* feat(dayplan): a GPX button beside PDF and ICS
Scopes are worded by what someone wants on their device rather than by the GPX
element: the whole trip, just the places for an offline map, just the days to
follow a plan. An empty trip says so instead of reporting a failed download.
* feat(mobile): GPX in the export sheet, which is no longer calendar-only
The sheet held the PDF export next to the calendar entries and was still called
Export Calendar; with GPX beside them the name was simply wrong.
* i18n(dayplan): translate the GPX export into all locales
* docs(map): document the GPX export
* fix(dayplan): drop the GPX tooltip and centre both export menus
The tooltip sat under the cursor exactly where the menu opens, so hovering the
button produced both at once. The label lives on as an aria-label. Both menus
now open centred under their button instead of hanging off its right edge.
* feat(collections): preview which of a trip's places a list already holds
The bulk import endpoint has been there since the select-mode copy, but nothing
calls it and it only reports duplicates after the fact. A dialog needs the verdict
up front, so /:id/importable/:tripId returns the trip's places with the duplicate
check the import itself applies, plus whether any day holds the place.
* feat(collections): import a whole trip's places into a list
Two steps: pick a trip, then tick its places. Places already on the list are
greyed out and unpickable, and the ones no day holds start selected — those are
the plans a trip left behind, which is the reason to import at all. Sorted so a
mostly-saved trip reads as the two that are left rather than a wall of grey.
* feat(collections): reach the trip import from the filter row and an empty list
* i18n(collections): translate the trip import into all locales
* fix(collections): stop the label dialog's buttons wrapping mid-word
Three actions in a max-w-sm footer wrapped into two lines each. The assign
button also painted white text on the accent, which disappears on a light one.
* docs(collections): document the trip import
* feat(atlas): show wishlist countries as hatched, stable colors, remove action
* i18n(atlas): translate the wishlist removal keys into all locales
* feat(atlas): offer the wishlist removal in the mobile country sheet
* docs(atlas): document the wishlist hatch and the removal action
* test(atlas): assert the stable country colour instead of a palette literal
---------
Co-authored-by: Mauro Remondi <mauro.remondi@sourcesense.com>
Co-authored-by: Maurice <mauriceboe@icloud.com>
The per-segment transport request now carries a direction, but the server
schema defaults it to 'outgoing'. Only send direction for the incoming
boundary-leg case so the outgoing payload stays identical to the pre-#1281
shape and existing callers keep the same wire format.
With block_weekends on, toggling a weekend day via MCP or the plugin RPC
still added the entry and burned allowance — the guard lived only in the
web UI (mcp-smoke-test I-02). toggleEntry now rejects inserts and
in-place conversions on blocked days with { error: 'weekend_blocked' }
(the votePoll pattern); removing an existing entry stays allowed so
stray weekend entries remain cleanable. The controller maps the code to
a 400, the MCP tool to an isError result.
Copying a trip nulled budget->reservation and reservation->accommodation
links (TEXT accommodation_id vs number-keyed map) and dropped
budget_item_members/payers, assignment_participants, budget currency/
exchange_rate/expense_date/ticket_json and reservation url. All links now
remap to the copied rows; external_* sync identity is intentionally not
copied; the to-do reset is now documented (smoke-test I-01).
* Open a running trip on today
Asked for in discussion #1567: during a trip, opening the planner starts
at day 1 and you scroll to find out what is planned for today.
The mobile shell has done this since the plan tab was built - it seeds
the active day with today when the trip is running. The desktop day plan
never did, so the same trip behaved differently depending on the screen
you opened it on. It now selects today's day, expands it and scrolls it
into view, once per mount and only while nothing is selected yet, so a
deep link or a day picked before switching tabs still wins.
The date comparison moved into one helper both sides use. It is
deliberately local rather than toISOString(): 08:00 in Tokyo is still
yesterday in UTC, and a trip day is a calendar day where the traveller
is standing.
Also fixes a note losing its colour when dragged to another day. A move
is a delete plus a create, and the create was still sending the three
fields notes had before colours existed.
* Open a collection place in a map app
Asked for in discussion #1455: a saved place is somewhere you intend to
go, but the only way to get directions to one was to add it to a trip
first, where the button has been for a while.
The collection detail gets the same picker off the same helper: Google
Maps, Waze, Apple Maps and OpenStreetMap, filtered by what the place can
actually be located with. A place with coordinates offers all of them; a
place typed in by hand with just a name and an address still offers the
two that take a search string, and a place with neither shows no button
at all.
It sits on the meta line beside the address, quieter than the link chips
below it, because it is an action on the place rather than a piece of it.
* Drag a place off the map onto a day
Asked for in discussion #891: assigning a place means finding it again in
the sidebar, even though it is right there under the cursor on the map.
The drop half already existed and did not need touching. Every day, every
row and every gap between rows in the day plan reads a placeId and calls
onAssignToDay with the position it was dropped at, complete with the
toast and the undo entry. What was missing was a second place to pick a
place UP from - the only one was the row in the places sidebar.
Markers become that second source. Both renderers build their markers as
real DOM nodes, so one helper covers Leaflet and the GL variants alike,
and it sets the same two channels the sidebar row does: dataTransfer for
the drop, and window.__dragData because browsers hide dataTransfer while
the pointer is merely hovering. The day plan cannot tell the two apart.
Not offered on touch: HTML5 drag is a pointer feature, and on a phone
the day plan is not on screen next to the map anyway.
* Show the landing time in the PDF
Asked for in discussion #1310: the export prints a flight's departure and
nothing else, so a reader cannot tell how much of the day is left for
anything after it.
The day plan has drawn both times as "14:30 - 16:45" for a while; the
export never got the rule. Now it does, for any transport that begins and
ends on the same day - a train and a ferry read the same way a flight
does.
Across a span it stays as it was: the arrival belongs to the arrival day,
which already prints it as its own time, and repeating it on the
departure day would put tomorrow's clock next to today's departure.
* Rich formatting and colours for notes
Asked for in discussion #1629: notes should carry formatting, links and
colour, so a warning reads differently from a recommendation.
Notes have rendered Markdown for a long time. The problem was that the
only way to write it was a bare textarea, so unless you already knew the
syntax there was nothing to tell you it existed. Both note fields get a
formatting bar - bold, italic, strikethrough, code, link, lists, quote -
which edits Markdown rather than replacing it with a rich-text model, so
the stored value stays plain text that the PDF, the plugin API and every
existing note keep reading unchanged.
Links now open in their own tab. A bare anchor navigates the planner
away, which loses the open day and the map, and without rel=noopener the
opened page can navigate this one - notes are content other trip members
wrote.
Day notes get a colour from a short shared palette. It tints the card,
the border and the icon circle, and it reaches the PDF as the spine and
rule rather than a fill. The dialog is two columns now, with a live
preview of the card beside the picker, because a row of swatches says
nothing about what a tinted note looks like next to a place.
On the phone the thirty-two icon grid was most of the screen, so it is
folded behind the chosen icon, which doubles as the colour preview. The
detail field there got the same formatting bar and the same 2000
character ceiling as the desktop one - they write the same column.
* Fix the tests that pinned the old note caps, and rework the note actions
The e2e suite builds its own day_notes table by hand rather than going
through schema.ts, so it never got the new column and every note insert
came back 500. Three more tests wrote down the 250 character body cap
and the old create() arity.
While in there: the edit and delete buttons on a note card were a 10px
icon in a 14px box, side by side, and they took width away from the text
on every note whether you were hovering or not. They are stacked in a
capsule now, 30px targets with their own hover states, floating over the
right edge of the card only while the pointer is on it. The text keeps
the full width the rest of the time.
* Centre the day badge when there is no weather to stack under it
The badge is pinned to the top of the row because with weather it is a
tall stack that has to start there. Without a location there is no
weather block, so all that is left is a 26px circle floating above the
day title.
* Render the place subtitle in the places list, and drop the form hint
The row in the places sidebar printed the description raw while the same
line in the day plan renders it, so a place written with the formatting
bar read as _underscores_ in one list and italics in the other. Still
clamped to one line - it is a list entry, not the inspector.
The "Markdown supported" line under the notes field goes: the bar above
it already says so.
Requested in discussion #1658: an expense carries an amount and a
category, and everything else about it ends up crammed into the title.
The column was already there. It just was not available, because the
costs UI stores an itemized receipt in it as TICKETJSON:{...}, so an
expense split by receipt could never carry a note - and saving one any
other way wrote note=null, which quietly wiped whatever had been typed
into the budget table. The receipt gets its own column and note goes
back to being text. Old clients that still send the blob as a note are
routed to the right column instead of clobbering the note.
On desktop the row now has fixed columns, so the amount does not drift
left when one row has a note and the next does not. The category moved
to the card edge as a tab like the mobile card does, which is what frees
the line the note sits on. Collapsed it is one line, and the same
element grows in place when opened.
The dialog is two columns now: what the expense was on the left, how it
is shared on the right, grouped into panels the way the trip dialog
does it.
On the phone the eleven category pills became a dropdown, which is where
the room for the note field came from, and a note on a card unfolds when
you tap it.
A place that has shut down for good is never the answer to "where should
we go", so it should not be in the results at all.
Google knows this and says so in businessStatus, we just never asked for
the field. It goes into the field mask and CLOSED_PERMANENTLY drops out.
CLOSED_TEMPORARILY stays - a restaurant on holiday next month is still
worth planning around - and results without the field are parks and
viewpoints rather than businesses, so they stay too.
OSM states it a few different ways. Re-tagged objects (disused:amenity,
abandoned:shop) never matched the selectors to begin with, but plenty
keep their original tag and gain a marker instead, and those did come
back. Same for opening_hours=closed/off.
Roadmap census 14 -> 8 (5 seam-pinned, 3 verified-permanent), resources.ts DoD line closed, README in-container list and counts corrected. Includes the SystemNoticesService unit test the src/nest coverage ratchet demanded after the addonEnabled threading.
* Tell a plugin where it is mounted
A plugin sits in one of six host surfaces and cannot see which: a trip tab and
the plugin page fill their container and scroll themselves, a dashboard widget
and the settings frame report their height, a detail slot is a narrow column in
someone else's panel. The contracts differ, and on a filling surface the height
report is discarded, so a plugin cannot even discover the difference by trying.
trek:context now carries viewport: the surface by name, phone or desktop, the
fill flag, and the insets the host already keeps clear. The insets describe
space that is held free, not padding to add — the mobile trip tab is the only
surface with chrome floating over it, and the numbers come from the same
variables the native tabs use so the two cannot drift.
* Seat the plugin tab like a native one
The mobile plugin tab kept clearance at the bottom and none at the top, so the
plugin's first row sat under the status bar and behind the floating back button.
The bottom value was a copied 84 that ignored the safe area on the devices that
have one, and the padding sat on the container rather than in the scroller.
Both clearances now come from the variables TabScroller uses, so the tab cannot
drift away from the native ones. The frame also states its surface, and pins
color-scheme for the reason both settings mounts already pin it: Chromium paints
a white canvas behind a transparent frame otherwise.
The tab also had no name anywhere on screen. It is not in the dock, so nothing
lit up there either, and the header had a branch for every other tab. It gets
the same treatment, built from the tab entry the planner already assembles.
* Keep the widget card intact on the phone
The card is built from the dashboard's glass vocabulary, which is scoped to
.trek-dash. The phone dashboard is a different route and mounts these widgets
outside that subtree, so background, border, shadow and blur all resolved empty
and the widget stood on the gradient as a bare title over an iframe. Each value
now falls back to the mobile token the surrounding cards are made of.
* Give the design kit a mobile layer
The kit was written for the desktop and stayed there: a filled card with a drop
shadow where the mobile design is translucent and border-led, inputs at 13px
which make iOS Safari zoom on focus and never zoom back, feedback on :hover
which never fires on a touchscreen, and one media query in the whole stylesheet.
It reads the new viewport field now and sets data-form-factor, data-surface and
data-fill on the document, plus the insets as variables. Everything mobile hangs
off data-form-factor="phone", so nothing changes on the desktop: the card loses
its shadow and takes the mobile surface tokens, targets grow to 44px, inputs to
16px, and pressing something scales it instead of waiting for a hover that will
not come. A filling surface anchors the body and scrolls an inner element,
because its height report is discarded by the host.
Plus the pieces a v4 screen is actually made of, so an author does not have to
reverse-engineer them: eyebrow, count pill, section header, status dot, round
icon button, field, segmented control. Each falls back to a global token where
the mobile family is absent, so they work on both.
* Document what a plugin needs on a phone
The guide had nothing on the subject: no mention of the second token family, of
who scrolls, of what sizes to expect, or that the kit brings almost nothing
responsive. It also promised more than it delivered, claiming the palette was
complete and that applying it matches the host "exactly" — fonts are named, not
shipped, so that was never true.
The new section covers the six surfaces and which of them fill versus report a
height, what insets mean (space already held free, not padding to add), the
mobile palette with the branch-on-absence pattern, and the four rules that
decide whether it feels native. It also says plainly that the dev preview cannot
show a phone width yet.
* Offer Waze and Apple Maps beside Google Maps
A place could only be opened in Google Maps, and the request in #919 was for the
apps people actually drive with. Google keeps the precise link it always had:
ftid, then place id, then the details URL, and coordinates only as a last resort,
so it still lands on the entry inside a mall rather than on the roof. Waze and
Apple Maps take coordinates and nothing else, which is why they are absent for a
place with no position, and why they open with navigation armed instead of on a
place view.
No setting. Every app that can resolve the place is offered, except Apple Maps,
which is dead weight on Android and Windows and therefore only appears on Apple
platforms. iPadOS reports itself as a Mac; both sides of that ambiguity have the
app, so the check does not need to tell them apart.
The picker is a new component rather than the context menu, which has no notion
of a submenu. It is portalled because both call sites sit in panels that clip
their overflow, and it is measured before it is painted: near the bottom of the
window it flips above the trigger instead of running off the screen, and it is
pulled back inside when it would overflow to the right. With a single app left
to offer there is no menu at all, the button opens it directly, exactly as it
behaved before.
The three surfaces take it the way each has room for. The inspector button
collects them behind one entry, the day-plan context menu lists them flat
because a menu can grow, and the phone keeps its single circle and opens the
picker as a sheet.
* Move OpenStreetMap into the navigation menu, and name the destination
OSM was still a button of its own beside the one that now collects the map apps,
which is the arrangement the menu exists to end. It belongs with the others.
The bigger correction is what Waze and Apple Maps are handed. Both document a
query parameter next to the position, and the pair is what makes a destination
legible: coordinates say which place is meant, the name is what the driver reads
instead of a number. Neither gets a name without a position, because there are a
lot of places called "Bahnhof". Google is unchanged; it can already do better
than a name through its place id.
* Stop drawing the section count as a pill inside a pill
SectionHeader wraps its count in a CountPill, and all three callers passed one
in already, so every section header carried two stacked pills: a lighter
rounded background peeking out from behind the real one. Pass the number and
let the header do the wrapping.
* Give the traveler filter a label and a way back
The avatars floated between the toolbar and the first section with nothing to
say what they were, and once a filter was on the only way out was remembering
which face had been tapped. It is a labelled bar now, closed off by the same
divider the rows below use, with an explicit reset that appears while a filter
is active. Both texts already exist in all 23 languages.
* Line the traveler filter up with the section titles
It carried the sheet's usual 18px of horizontal padding while SectionHeader
right below it starts at 2, so the two labels sat on different left edges.
* Filter the place list by stars instead of sorting by them
Sorting put the best-rated place first and left every other one on the list,
which does not help when the question is "what does the group actually want to
do". It is a floor now, the same one the collections bar uses: All, or 5+ down
to 1+, with unrated places dropping out once a floor is set. The button carries
the active floor so the list is never quietly shortened.
The row it sits in used to render only when the trip had categories, which hid
the star control on a fresh trip. That is exactly where someone collects places
to rate, so the row is always there now and only the category dropdown depends
on there being categories.
places.sortByRating is replaced by places.filterByRating in all 23 languages,
since the old label described a behaviour that no longer exists.
* Keep the collections map when a filter matches nothing
Clicking a label with no places behind it made the map disappear and the page
reflow to full width. The layout hung off the FILTERED set: no visible places
meant nothing mappable meant no split. A filter should empty the map, not tear
it out, so the layout now asks whether the list has anything mappable at all,
while the markers keep following the filter.
* Give the label filters a tray of their own
Two changes to the filter row. Select and manage-labels drop their captions and
keep icon, title and aria-label, which puts them on the same footing as the
add-place button that was already icon-only.
The labels themselves were pill-shaped buttons with their own borders sitting
loose among square-cornered dropdowns, so a list with a handful of them read as
a pile rather than as one control. They share the filter buttons' radius and
type now, and sit in a recessed strip led by a tag glyph: the chips carry no
border of their own, because a frame inside a frame was half the noise. Manage
stays quiet at the end of the strip until hovered.
* Put the collections controls on one line
The filter row sat at about 31px while the buttons floating over the map are
38, and the icon-only add button was the odd one out even within its own row.
Scrolled to the top, where the row and the map's top bar are side by side, that
reads as three different heights.
One --col-ctl-h for both, fixed rather than derived from padding so a button
with an icon and one with a word cannot drift apart, and the row starts at the
same offset the map's own bar does. Phones raise the variable instead of
patching individual min-heights, which keeps the square buttons square.
* Offer Apple Maps on desktops that are not Apple
Apple Maps has had a web version since 2024, so the link resolves to a usable
map on Windows and Linux rather than to nothing. Hiding it there was a guess
about a limitation that no longer exists.
Android keeps the exemption: the web version works there too, but nobody
navigating from an Android phone reaches for Apple Maps, and the row of choices
is short on purpose.
* Move the label filter into the map's top bar
The filter row was running out of room while the map's own bar had space to
spare, so the labels ride up there next to the collapse button, in the same
glass as the controls beside them and scrolling sideways rather than wrapping.
They fall back into the filter row whenever no map is on screen — a narrow
window in list view, or a list with nothing mappable in it — so they can never
end up somewhere the user cannot reach. Both places render the same component
now instead of two copies drifting apart.
* Stop the active select button drawing a dark icon on a dark fill
Two rules with the same specificity claimed the same class. On a dropdown
`.open` means the panel is down, so the label stays dark; on the select button
it meant the mode is on, which fills it with the accent and needs the light
text. The dropdown rule sits later in the file, so it won, and the icon
disappeared into the fill the moment the button was clicked.
The state is `.on` now, which is what the other toggles on this page
(col-map-btn, col-iconbtn) already call it.
* Open the place in Google Maps, not the coordinates
Reported in #1278 and still reproducible: without a Google identifier the link
carried nothing but latitude and longitude, so Maps dropped an unlabelled pin
instead of opening the entry. The name and the address were both on hand and
both thrown away.
Google documents query=PLACE_NAME,ADDRESS for exactly this case, so that is
what the fallback sends now. A name with no address still falls through to
coordinates, because "Central Station" on its own would confidently open the
wrong city, and a pin in the right spot beats a link to the wrong place.
The precise paths are untouched: an ftid or a place id still wins, and they are
what a place searched through Google carries. This only changes what happens
for the ones that carry neither, which is every place added from OpenStreetMap,
by hand, or before those fields existed.
* Put the atlas sparkline behind the text it was covering
The graphic is absolutely positioned and comes after the label, value and delta
in the DOM, so with no stacking order of its own the browser painted it on top
of them. On the dashboard's atlas row a status card is about 190px wide at
1440px, and the 80px sparkline covered more than half the delta line (#1787).
Give it z-index 0 and lift the three text rows to 1. It is decoration, so it
stops taking pointer events while we are here.
* Let a sentence-length sheet subtitle wrap
TileHeader truncates its subtitle to one line. That is right for a filename or a
date, and wrong for the reorder hint: between the 40px icon tile and the 34px
close button there is not enough width for a sentence, so it was cut off
mid-word (#1814), and every translation is longer than the English one. Opt in
with subWrap where the subtitle is prose; the file sheets stay single-line.
* Translate the phone entry view
MobileEntryView carried its own English label tables while the desktop view
resolved the same values through i18n, so a German reader got "Amazing" and
"Sunny" on a phone and the translated chips on a tablet (#1846). Which view
appears is decided by two different breakpoints, which is why it looked
arbitrary. The labels are keys now, the edit action, the photo counter and the
pros/cons headings go through t(), and the date formats against the chosen
locale instead of the browser default. No new keys: all of them already exist
in all 23 files.
The colour classes stay local. The desktop table carries raw hex values and no
dark-mode variants, so sharing one table would change how this looks.
* Name Google account linking in the Cloudflare bot section
The section claimed the block was specific to ChatGPT, so someone hitting it
with Gemini had no reason to read on (#1667). Google exchanges the code from its
own servers, which Bot Fight Mode blocks the same way, and the symptom differs
enough to be worth spelling out: the browser half succeeds and linking fails
afterwards, with a 200 on /api/oauth/authorize and no /oauth/token at all. Also
note that Bot Fight Mode ignores IP allowlists, which is the first thing people
try.
Both advisories are the same mistake: a route that hands out a bearer token
treated trip membership as authorisation. Membership lets someone read the trip
while they are signed in. It does not let them mint a URL that reads it without
an account, keeps working after they are removed, and can be rotated out from
under the owner's calendar subscriptions.
The calendar feed token now sits behind TripAccessGuard plus
@RequirePermission('share_manage'), the same shape the rest of the trip-scoped
controllers use, which also replaces a hand-written access check that GET never
called at all. GET is gated with the writes because it returns the credential
itself, not its status. The share link's GET does what its own POST and DELETE
already did, and what the invite link next door already does for exactly this
reason.
Two things had to come with it or the gate would have been decoration. The MCP
tool get_share_link read the same token on membership alone, so it would simply
have become the new way in. And feed_token rides along in every trip payload
because TRIP_SELECT reads t.*: any member could have read the value out of
GET /api/trips/:id, the summary tool or the plugin RPC. It is blanked once in
the select and stripped in the two places that select the row directly.
The three service writes take the acting user and scope their UPDATE to a trip
that user can reach. The guard decides who may manage the credential; this only
makes sure a caller arriving another way cannot write to an id it guessed.
Client side, "Subscribe to calendar" is now behind the same permission on
desktop and mobile. Without that a member would have kept an entry whose enable
button silently does nothing, because IcsSubscribeModal swallows every non-200.
The one-off ICS download stays where it is: that is a file they may already read.
Instances that want members to manage these links can lower share_manage to
trip_member in the admin panel, which the tests cover.
Landing on /login while already signed in, by typing the URL or going back, showed the login form again (#1810). PublicRoute takes a redirectAuthed flag on those two routes and sends the visitor to the start destination instead, capturing the auth state at mount so a fresh login is left alone during the takeoff animation. Only from a bare URL: ?redirect=, ?invite=, ?oidc_code= and ?oidc_error= all have to reach useLogin intact.
A transport also carries a top-level reservation_time, and that branch derives its zone from the linked place, which a flight does not have, so both ends landed as floating times in the subscribed feed (#1453). Consult a timed departure endpoint first instead and keep the IANA zone per side. When the arrival side has no clock of its own, the DTEND still comes from reservation_end_time, so a rental imported with only a geocoded pickup keeps its multi-day return.
The admin panel's test button reported a key with an HTTP referrer restriction as invalid, because the validation call was the one Google request that did not send a Referer. It sends the same header googleFetch already sets for every real request.
typecheck:tests landed with the contract ratchet; the scheduler migration
merged after it, and its idempotency-cleanup test stubs register() with no
parameters. Reading calls[0][2] back off a no-parameter mock types every
recorded call as [], which is a tuple index error, so dev has been red since.
Give the stub the real register() signature and drop the cast it needed.
Both were left open by the audit and are closed here.
days, packing, photos and reservations each built a service outside the
container at module load for nobody: only their own delegation tests imported
them, and three domain docblocks were still telling readers to go through them.
photos.bridge's own comment pointed at src/services/, a directory deleted two
PRs ago. The cases that exercised them kept their assertions and call the
service the bridge delegated to, so nothing that was proved stopped being
proved. Twelve bridges remain, each with a live out-of-container caller.
Coverage measured src/** but gated only src/nest/**, so the boot path, the
scheduler, the MCP handler and the migrations sat under no floor at all. That
is the same 9,602 lines where this audit found its worst defect, and a
swallowed EADDRINUSE is exactly the kind of thing an untested boot path hides.
Seven globs now pin what is measured today. They are floors to raise, not
targets that have been met: src/*.ts is at 39% statements and src/db at 38%
branches, and writing those numbers down is the point.
src/demo/** is deliberately left out. It measures 0%, and a floor of zero
asserts nothing; it needs tests before it needs a threshold.
The docblock said the ws handshake consumes tokens from outside the container.
That was true when it was written and stopped being true one PR later, when the
transport became a gateway that injects this service.
The store still has to be module state, for a reason the comment did not give:
four bridges build EphemeralTokenService by hand, and per-instance state would
give each of them an empty Map, so a download token minted through a bridge
would never validate. An audit lens flagged the stale reasoning; the
requirement survives it.
An adversarial audit over the three merged migration PRs. Sixteen findings
survived triple refutation; these are the ones with a defect behind them.
The serious one is mine, from the gateway move. ws registers
`error: this.emit.bind(this, 'error')` on the http server it attaches to, and
the adapter now runs inside buildApp, before listen(). So from that moment the
http server HAS an error listener and Node stops throwing on one. The adapter's
handler was `() => {}`, which meant EADDRINUSE was forwarded there and dropped:
the process survived, the listen callback never ran, so the banner never
printed and the scheduler never started. A container that looks healthy and
serves nothing. Reproduced, then fixed at both ends: index.ts owns the fatal
decision on the http server, and the adapter logs instead of swallowing.
Four DTOs were tightening contracts rather than describing them, which is the
thing the pipe-before-handler ordering makes easy to do by accident:
- OauthClientCreateDto replaced 'Name is required' with a Zod field report, and
pre-empted the handler's 403 on an instance with the MCP addon off.
- BookingImportConfirmDto demanded per-item fields the service never reads, so
a caller assembling items itself got a 400 about `source`.
- TagCreateDto/TagUpdateDto typed name and color as strings; the legacy route
bound whatever arrived, so a numeric name used to create a tag.
- PluginUserSettingsUpdateDto typed config as a record, where the handler
answers 200 for anything it cannot use.
Two of my own tests were false-green, which the audit demonstrated by mutation
rather than by reading. WSGW-008 asserted on frames that never carry the user
object at all, so it passed with the password_version strip deleted; it reads
the retained identity now. GEO-009 made the same assertion before and after the
write; it distinguishes a cached negative from a miss now, through a hasCached
seam that says what getCached cannot. Both were re-run against the mutation
they exist to catch, and both fail on it.
booking-import geocodes every venue and endpoint in one request loop, up to
thirty sequential calls. Folding maps onto the shared Nominatim cursor put that
loop on the interactive lane, so an import starved every concurrent keystroke
for half a minute. It takes the background lane now, which is what the two-lane
split was built for.
Also: the README's in-container bridge list had outlived one entry and was
missing another, and the journey e2e still mocked a module this branch deleted.
The plugin host's job scheduler now uses the cron package (the one
@nestjs/schedule wraps), so node-cron and @types/node-cron leave the
dependency tree entirely. Verified against both real libraries that the
accepted cron grammar is identical — no plugin's declared schedule tightens —
and pinned that with an unmocked compat test.
The transport was a hand-rolled setupWebSocket(server) that index.ts kicked
off with a dynamic import after listen(). It is a @WebSocketGateway now, with
DatabaseService and EphemeralTokenService injected instead of imported, and
the heartbeat owned by a lifecycle hook rather than an interval nobody stops.
The wire protocol does not move. Nest's WsAdapter dispatches on
{ event, data } and every deployed client sends { type, tripId }, so
TrekWsAdapter translates rather than asking clients to change. It also builds
the ws server itself instead of delegating: the base's port-0 branch uses
noServer and performs the upgrade from its own registry, which never consults
verifyClient, so delegating would have dropped the origin check and answered
101 to anyone. The per-socket error listener stays attached first, because an
unhandled one kills the process (#1576), and the flood guard counts every
inbound frame before parsing, because malformed frames are exactly the ones
worth limiting.
Four things had to stay where they were, and each is a silent failure if it
does not:
The socket registry stays module state. Seventeen bridges and
notifications.instance build `new RealtimeService()` outside the container;
rooms on a provider instance would leave their broadcasts going to an empty
map, with no error and no log.
The socket id stays a monotonic integer. The client echoes it as X-Socket-Id
and broadcast excludes the originator with Number(excludeSid). A uuid is NaN,
NaN === NaN is false, and every client would receive its own writes back,
which shows up as drag-and-drop that jumps rather than as an error.
buildApp creates the http server and binds the adapter before app.init(), and
callers take it from getHttpServer(). Nest binds gateways during init, so a
later adapter binds nothing; an adapter given the app binds to Nest's own
internal server, which this process never listens on. Both boot green with no
client able to connect.
RealtimeGatewayModule is imported by AppModule alone. RealtimeModule is
@Global, so a gateway there follows into every hand-assembled TestingModule,
and Nest's SocketModule then loads an adapter none of them set: it reaches for
@nestjs/platform-socket.io and calls process.exit(1). Eighteen e2e harnesses
died that way before the split.
src/websocket.ts survives as a re-export of the three fan-out primitives. 115
test files vi.mock that path to assert on broadcasts.
The two transport suites pass unchanged, and 30 new cases pin the pieces
directly: the pv and MFA gates, the numeric socket id, the plugin announcement
firing before the room check, both sides of the origin allowlist, and the
flood guard counting frames no handler wants.
The trip/todo reminder enable gates (notify_trip_reminder, notify_todo_due)
are now read inside each 9 AM tick instead of once at boot, so toggling them
takes effect at the next run without a restart. Saving notification settings
no longer re-arms anything: the auth-side startTripReminders() call and the
shouldRestartScheduler result field are gone (no controller consumed it, so
the PUT /api/auth/app-settings response shape is unchanged). The boot
'enabled/disabled' log lines are informational banners now.
BODY_CONTRACT_ALLOW_LIST is empty. The last seventeen handlers type their
body with a createZodDto class, so the fail-closed boot gate has nothing left
to forgive; it still refuses a boot for a new handler that reads @Body()
without one.
Several of those contracts describe rather than constrain, and that is the
whole care in this change. The pipe runs before the handler, so a schema
stricter than the handler moves a user-visible answer:
- TagsController.create answers { error: 'Tag name is required' }, pinned
byte-for-byte in two tests, and the pipe formats as `field: message`. The
DTO leaves name optional and the handler keeps the check that owns the
message. Same shape for the booking-import confirm body, which additionally
gains per-item validation the handler never did.
- PluginsController.activate and install answer 503 when plugins are disabled,
and check that BEFORE looking at the body. A schema that rejected first
would turn that into a 400 for an operator whose server has plugins off.
- setEgressHosts reads a non-array `hosts` as "clear the list", which is the
documented reset. Typing it as string[] would have removed a working admin
action rather than tightened anything.
- PluginRoutesController.route answers { route: null } at 200 for everything
it dislikes, so the client falls back to straight lines. A 400 there would
turn a graceful degradation into a visible error.
files.bridge is gone. The multer options are built by
MulterModule.registerAsync from an AllowedFileTypesService leaf, so the
fileFilter reads the operator's extension list from the container instead of
from a second FilesService built outside it. Nest injects
MULTER_MODULE_OPTIONS with @Optional(), so a token that fails to resolve is
not a boot error: it silently falls back to defaults, with no filter and no
size cap. FILE-020 already proved that for trip files by uploading an .svg
over HTTP and expecting 400; journey uploads had no such case, and now do.
VIDEO_UPLOAD is untouched, including the filename callback that forces .jpg
regardless of what the client sends, which guards the public share proxy.
PluginFrame gets an onError. A frame whose document fails to load left a
blank element at full height, which reads as a hung app rather than as a
plugin that is gone.
typecheck:tests runs in CI now, with the test tree at zero errors. tz-lookup
is left as it is: the build puts it in a lazy helpers chunk, not the entry
chunk, and the remaining split would trade a synchronous render for an async
one to move 172 kB that is already off the critical path.
Client CI was red on this branch while dev was green, and no client file had
changed. The 536 test files all passed; the job failed on three unhandled
rejections, all `window is not defined` from MDashWidgets, attributed to
whichever case happened to be running when they landed.
The widget fetches api.frankfurter.dev directly and sets state from both
handlers with no unmount guard. MSW runs with onUnhandledRequest: 'warn', so
that request was never intercepted: it left the runner, came back after the
environment was gone, and React tried to schedule an update against a torn
down tree. Load dependent by construction, which is why dev happened to pass.
Two halves, because either alone leaves the other broken. The widget now
aborts in flight work on unmount and on refetch, and treats an abort as "no
one is waiting" rather than as a failure to report. That matters outside the
runner too: this is a third party call on a mobile screen a user can leave
immediately.
The handler set gains the third party endpoints the app calls directly rather
than through /api, so no suite reaches the real host. An empty rate list is
the shape the widget already handles.
Also clears twenty more test files, each at its root cause. Two more were
passing for the wrong reason: oidc.service.test.ts built AuthService with an
AtlasService in the membership slot, undefined for userCleanup and mailer,
and the mailer stub in the tokens slot, the same shift admin.service.test.ts
had. And CAL-010 claims to serialise a flight byte for byte, but
createReservation ignores reservation_time, so the row is written with a null
time and CalendarService drops it: the recorded snapshot is truthful about
the data the case builds and silent about the flight it advertises.
ephemeral-tokens.ts was free functions, and AuthService owned the cleanup
timer only because it happened to be the domain that mints tokens. Issuing
them and sweeping them are different jobs, and FilesService and TokenService
mint or consume them too.
EphemeralTokenService owns the lifecycle and hands the three operations out.
The store and the interval stay module state next door, and that is
load-bearing rather than a leftover: src/websocket.ts consumes ws tokens on
the upgrade handshake, outside the container, before any injector is in
reach. Per-instance state would hand it an empty Map and close every
connection with 4001, which no test would catch because the ws suites drive
the transport directly. WS-027 through WS-029 pin the session gate that runs
through exactly that path.
The MAX_STORE_SIZE back-pressure is the second reason: createEphemeralToken
returns null at 10,000 live tokens and TokenService turns that into a 503. A
second store would raise the real ceiling with nobody noticing.
Its own leaf module rather than a provider in AuthModule: TokensModule mints
download and ws tokens as well, AuthModule already imports TokensModule, and
TokensModule is documented as a deliberate leaf that three graphs reach for.
Same shape as MailerModule, which broke AuthModule <-> NotificationsModule.
Twelve hand-wired construction sites in tests needed the new collaborator.
The typecheck named all twelve; before this branch they would have been
undefined at runtime.
Sixteen more test files are cleared here too, each at its root cause. Four of
them were passing for the wrong reason:
files.controller.test.ts drove two handlers with shifted arguments, a string
tripId where the trip row belongs and undefined where tripId and id belong.
It stayed green because FilesService is fully stubbed and the stubs ignore
what they are given, so DELETE /:id/permanent and DELETE /trash/empty never
covered that the trip row reaches the permission check.
conflictUpdate.test.ts was missing real collaborators on PlacesService and
PackingService. None of the exercised paths reach them, so it passed by luck.
assignments.service.test.ts builds its service by hand, and the one consumer
of the collaborator it was missing swallows every throw in a bare catch. A
broken journey reconciliation would still have shown 28 green tests.
scheduler.test.ts mocks two modules with paths that resolve one directory
above the server root, so they mock nothing at all. Harmless only because
src/scheduler.ts imports neither.
conditions.test.ts lost noTrips when the condition kind landed; oauth.test.ts
lost the RFC 8707 resource indicator when it became required. Both are
fixtures that stopped matching a contract that grew, and both are filled in
with the value that keeps every existing case reading as it did.
budget.service.calc.test.ts typed its prepare mock from the default
implementation rather than from prepare's own signature, so all ten later
mockImplementation calls had to reproduce that exact three-member literal.
One stmt() helper names the widening instead.
admin.service.ts reached getOnlineUserIds with require('../../websocket')
inside a try/catch — the last lazy require of the ws singleton in the tree,
and exactly what RealtimeService exists to prevent. The facade gains the
third pass-through next to broadcast and broadcastToUser, delegating at call
time like its siblings so per-file vi.mock stubs keep flowing through.
The catch stays at the admin call site rather than moving into the facade: an
admin list that shows everyone offline is a better answer than a 500, but
that judgement belongs to this caller, not to every future one.
Typechecking the tests turned up two more things behind the arity errors.
admin.service.test.ts built AuthService with an AtlasService in the
membership slot and no mailer at all, so its last four collaborators were
shifted by one. Nothing failed, because no case there reaches a path that
uses them.
Its getOidcSettings/updateOidcSettings wrappers referenced methods AdminService
no longer has. Nothing called them: OIDC settings moved to OidcService and the
two cases moved with it (ADMIN-SVC-047/048 now live in oidc.service.test.ts).
Only the wrappers stayed behind.
server/tsconfig.json has include: ["src"], so tests/ was never typechecked.
A service that gained a constructor parameter did not break its hand-wired
call sites at compile time — the argument arrived as undefined and threw at
runtime, if anything read it at all. The notifications sweep hit 46 such
sites and only a stash-diff against a throwaway tsconfig found them.
tsconfig.tests.json makes that permanent. It is not wired into CI yet: the
tree still reports 170 errors, and turning it on now would block every
unrelated PR.
Three of the first fixes were not type noise.
DAY-SVC-034 asserted that createAccommodation rolls back a failed
reservation insert. That method moved to AccommodationsService in the
days/accommodations split, so the case was calling undefined — which throws,
which satisfied .toThrow(). The rollback has been unchecked ever since,
while the case reported green. Repointed at the owner; the invariant holds.
host-rpc.test.ts typed its response as RpcResponse & Partial<RpcError>.
Both carry ok, one as true and one as false, so the intersection is never
and all 97 property reads in the file were errors nobody could see.
auth.controller.test.ts passes {} as the body in cases that stub the service
and assert on the mapping around it. That is correct as written — the point
of several of them is that the error path never reads the body — so anyBody()
names the intent instead of inventing fields the assertions do not read.
Three changePassword calls were genuinely missing the @Res() argument.
Nine fire-and-forget sends reached notifications.bridge through
import('...').then(({ send }) => ...). The laziness was working around a
module cycle that no longer exists, and it cost twice: it hid the edge from
the module graph, and it handed each send a NotificationsService built
outside the container rather than the container's own.
The sweep needed a method rather than a grep. tests/ sits outside
tsconfig's include, so a service that gains a constructor parameter does not
break its hand-wired call sites at compile time — the argument arrives as
undefined and throws inside the send. Adding tests to a throwaway tsconfig
and diffing the errors before and after named all 46 sites exactly. The same
config reports 288 pre-existing errors in the suites, almost all deliberate
partial constructions in the harnesses, which is why tests/ is not in the
build's include; that cleanup is its own piece of work and is written down
in the nest README.
notifications.instance.ts holds the out-of-container instance, apart from
notifications.bridge.ts on purpose: several suites replace the bridge
wholesale with vi.mock(..., () => ({ send })), and a sibling bridge taking
its instance from there would break under those mocks.
The suites that assert on a send now hand their spy to the injected service.
The reservations e2e overrides the provider, because its app resolves the
real one out of its own container and never saw the module mock.
streamImmichAsset answered { error, status } when the effective owner had no
Immich connection stored, and left the response untouched. All four callers
ignored that return, so the client got no status line and waited out its own
timeout. Its Synology counterpart has always written through
handleServiceResult; this is the same shape.
Every other exit already went through pipeAsset, which writes on all paths
including its catch, so the method is void all the way down now — and the
PhotoProvider interface drops the error return it was carrying purely to
describe this defect.
Three cases pin it, including the one that matters for shared albums: the
connection that gets read is the OWNER's, so a viewer who happens to have
Immich configured must not paper over an owner who does not.
Dispatch to a photo backend was `switch (photo.provider)` in
photo-resolver.service.ts twice and, in journey-public.controller.ts, an
if/else that handed anything not named immich to Synology — where it 404'd
from inside that provider's id parser. Adding a third backend meant finding
all three.
PhotoProviderRegistry.get(id) replaces them, over adapters registered under
the PHOTO_PROVIDERS token in memories.module.ts. Adapters rather than
`implements` on the services themselves: ImmichService and SynologyService
are also the album and browse surfaces, and folding the interface into them
would drag their argument orders and return shapes along.
PhotoAssetRef is the part worth having. The two services took the same three
values in different orders — Immich as (userId, assetId, ownerId), Synology
as (userId, targetUserId, photoId) — so every dispatch site had to remember
which was which, and a transposition would have compiled and served another
user's photo. One place builds the ref, one place decrypts the Synology
share passphrase.
`local` stays an explicit branch at both dispatch sites: the bytes are on
this disk, there is no row for it in photo_providers, and its asset info
comes from the stored row rather than from a backend.
The rename memories/ -> media/ is deliberately not here. It is import churn
across 275 files, and coverage-thresholds.mjs buckets on one path segment,
so merging photos/ in would collapse two ratchet entries into one and keep
the looser of the two.
Nine *.mcp.ts files reached through addons.bridge for one boolean each, and
platform.routes plus the legacy registrars still do. That was never a
preference: a decorator's options object is built while the class is being
defined, so the `when:` closure had no `this` to reach an injected service
through, and the only thing in scope was a module-level singleton that
builds its own AddonsService outside the container. Nine domains hid an edge
to the addons domain from the module graph to answer "is this addon on".
@trek/nest-mcp hands the gate its declaring instance now, resolved at
attach rather than captured at module load, and addons/addon-gate.ts turns
that into addonGate(ADDON_IDS.X) over the injected service. The option is
declared with method syntax so a predicate written against its own
controller class stays assignable.
Two more edges went the same way, both checked for cycles first:
reservations.mcp injects AssignmentsService (AssignmentsModule never
reaches back), atlas.mcp and journey.mcp inject AuthService.
The tests that forced a toggle by mocking the bridge module now write the
addon row the admin panel writes, so they exercise the real read.
Four in-container uses stay, each documented in the nest README: places.mcp
(a real Days -> Places -> Assignments cycle), budget/packing/costs against
trips.bridge, user-cleanup against budget.bridge, and the two multer
interceptor options, whose fix is MulterModule.registerAsync.
Atlas throttled to the 1.1s OSM asks for and cached its answers in a 50k
map; maps did neither, so five interactive routes fired straight at the
service while the two atlas backfill loops politely waited their turn.
From Nominatim's side that is one instance ignoring the rate limit.
Both go through geo/nominatim.client.ts now. Three things the naive merge
would have got wrong:
- The shared client sends UA, the instance-identifying User-Agent, not the
generic string atlas used. A shared generic UA is throttled harder.
- One cursor across five interactive routes and two background loops would
have put a keystroke behind however many places the backfill enqueued, so
there are two lanes over the one cursor and background yields.
- The cache key carries the query shape. Atlas asks at zoom 3 and 8, maps at
18; a coordinate-only key answers a street lookup with a country.
The throttle cursor and the cache stay module state, the way the permissions
and FX caches do, because the bridges build their collaborators outside the
container and a second copy of the cursor is the bug being removed here.
GeocodingService owns only the cleanup timer's lifecycle, which atlas used to
start as an import-time side effect.
Also folds the three haversine implementations into common/geo.ts, keeping
the clamped form: asin of a value a hair over 1 is NaN, and a NaN distance
fails every comparison it feeds instead of throwing.
tests/setup.ts zeroes the interval. Nineteen maps cases drive these paths
back to back with real timers and were sleeping nineteen seconds for it.
Two controllers declared the identical @Controller string from two modules and
each rebuilt the trip access check its own way: booking-import ran it as a
guard chain, integrations ran it by hand inside the handler, and
BookingImportService carried a third copy that nothing called at all.
They never actually collided - the four file-upload sub-paths and the one
AirTrail sub-path are disjoint - so this is not a bug fix. It is one access
path instead of three implementations of the same one, and the dead copy is
gone.
The addon requirement sits on the AirTrail handler rather than the class.
Class-level would have 404d the four upload routes whenever AirTrail happened
to be off. AddonGuard still runs first in the chain, because a disabled addon
has to answer 404 to an anonymous caller before anything answers 401, and
ADDON-GUARD-006 pins both halves of that.
One deliberate behaviour change: the AirTrail route now gets its trip 404 from
the guard, ahead of the body 400, because a guard runs before the pipe. That
is the ordering every other trip-scoped controller already has and nothing
pinned the other way for this route.
The class rename would have stopped the server booting: the body-contract gate
keys on ControllerClass.methodName and throws on stale entries. Three entries
renamed, and the list re-sorted - GATE-006 caught that the rename had broken
the ordering it asserts.
booking-import lost the controller from its coverage denominator, so the
branch floor was about to fall. It goes up instead: the off-request parse had
no case for its catch arm, so a throw would have left the job on 'running' and
the widget spinning, and the binary probe behind /api/health/features had no
test at all - that one boolean decides whether the client offers booking
import.
The directory is empty after the airtrail fold, so it goes. The ESLint rule is
the part that matters: the directory disappearing is not what keeps it gone,
because the way it grew was one file at a time, each one reasonable on its own.
Any import of a services/ path is an error now, with the message pointing at
the recipe in the module guide.
Proved rather than assumed: a probe file importing the old airtrail path fails
the rule, and the whole package still lints with zero errors.
Not changed: the 91 provenance comments in shared/, client/ and server/ that
name a services/*.ts file. They are statements about where a schema or a
service came from, and they stay true as history. Rewriting them all would put
91 files of noise over the change worth reviewing, and the paths they cite are
already gone from the tree, which is the honest signal that they are history.
src/services/ held one thing: airtrail/, 1305 lines of plain functions over the
better-sqlite3 singleton. They are four providers and two pure helper modules
now, and the directory is empty.
Five hidden edges go with them. The db singleton in three files, the raw
broadcast from src/websocket in two, and the reservations, addons and audit
bridges - all module-level imports that the module graph could not see.
The scheduler stops reaching for it at call time. scheduler.ts ran
require('./services/airtrail/airtrailSync') inside the cron tick, so a wiring
mistake surfaced as "AirTrail sync tick failed" every five minutes rather than
at boot. It takes the dependency through SchedulerDeps now, the same structural
contract backups and placePhotos already used.
One edge stays a seam, deliberately. AirtrailSyncService needs
ReservationsService, because the inbound pull applies remote changes through
the real reservation update path - endpoint restamping and all - and
reimplementing that against raw SQL would be worse than the coupling. The
write-back trigger runs the other way, from reservations.controller, and
injecting there would close
ReservationsModule -> AirtrailModule -> ReservationsModule. airtrail.bridge.ts
holds that one function with the reason in its docblock, rather than a
forwardRef hiding it.
The write-gate suite was module mocks over free functions and is stubs through
a constructor now; it gained four cases on the way (auth failure, remote
delete, disconnected owner). The domain ratchet goes from 19/0/10/19 - set when
integrations was two thin controllers - to 69/52/72/71, which is what the
folded code measures. It is a 1:1 move, so it arrives with the coverage its
tests always gave it.
place-enrichment landed on dev without an entry in the per-domain coverage
table, so it sat on the 80/80/80/80 catch-all while measuring 99.5 statements
and 99.0 branches. That is the exact hole the per-domain ratchet was
introduced to close: one well-covered domain quietly carrying nineteen points
of headroom nobody would notice losing.
The rest of that work needs nothing from this branch. It injects
DatabaseService rather than reaching for the db singleton, reads no
process.env, carries a Zod contract in @trek/shared with a colocated DTO, adds
no allow-list entry, and its controller declares @UseGuards(JwtAuthGuard) -
which is what keeps it authenticated under the default-deny guard this branch
introduces, since a declared chain makes the global guard stand down rather
than answer first.
Every split on this branch dropped the giving domain's ratchet without a line
being tested any worse: coverage follows the source file, so when
well-covered code leaves, the weaker branches that stay behind become the
whole denominator. days fell from 82 to 77.5 percent branches that way, feeds
to 76.7, trip-read-model to 62.5.
The rule in vitest.config.ts is that the only way past a number is to raise
it, so these are tests, not edits to the thresholds. days is back at 92.1,
calendar at 96.2, trip-read-model at 100.
feeds and trip-members and trip-read-model had no unit test at all - feeds'
whole threshold rested on one e2e, which is exactly the file whose contract
this branch rewrote. They have one now, and the read-model test pins the
thing in that file that would leak data if it regressed: getTripSummary and
bundle both thread the viewer id into packing.listItems so another member's
private items stay hidden, and nothing asserted that before.
The five thresholds this branch introduced move up to what they now measure,
with a few points of slack. The pre-existing entries are left alone on
purpose: they were generated from a Linux run, CI measures there, and a
handful of branches are platform-specific - raising them from a Windows
measurement is how you get a failure that reproduces nowhere local.
The guard section still said BE-Phase 2 was in progress and listed
TripAccessGuard, default-deny and the MFA policy as open. All three landed on
this branch, and two of them changed how a contributor is supposed to write a
controller, so the file that tells them how needs to say it: what
GlobalAuthGuard does with a declared @UseGuards chain, why places splits along
"does this handler take a DTO-typed body" rather than along the domain, and
why TripOwnerGuard must not go near PermissionsService.
It also still pointed at middleware/mfaPolicy.ts in two places, which this
branch deleted.
The Express shell set a 100kb limit and stopped applying it the moment the
Nest instance took over parsing: bootstrap passed bodyParser:false and
nothing replaced it. The number happened to stay the same, because it is also
the framework default - but it lived nowhere anybody would look for it, and
one NestFactory default away from silently becoming something else.
It goes through Nest's own parser rather than a second express.json(),
because that parser carries the verify hook that keeps req.rawBody, and the
plugin webhook routes need the exact payload bytes to check a provider's
HMAC. Adding a parser in front of it would have taken rawBody away without
failing anything.
MISC-009 and MISC-010 pin it from both sides.
The bodyParser flag on applyGlobalMiddleware goes with it: one caller, which
passed false.
Not moved: cookieParser stays on the raw Express layer rather than becoming a
configure(consumer) middleware. The platform routes - the OAuth consent flow
among them - are registered on that instance before Nest routing exists, so a
Nest-slot cookie parser would simply not run for them. helmet, CORS,
compression, the forced-HTTPS 301 and the res.on('finish') logging stay for
the same reason: they have to see requests Nest never routes.
The last live non-shell middleware verified the JWT and read the users row a
second time on every /api request, then threw the loaded user away - two
jwt.verify calls and two SELECTs per authenticated request, because it ran
before Nest and could not see what the auth guard had already resolved. The
guard reads req.user.
It also corrects an asymmetry that had grown quietly. The middleware decided
on two literal path lists, and every public /api endpoint added after those
lists were written - /api/config, /api/help/*, the public journey and share
routes, /api/health/features - answered a logged-in user without MFA with a
403 while answering the same request from a stranger perfectly well. Nothing
intended that; the lists simply stopped being maintained. Keying on @Public()
means the exemption cannot drift from the route again. This is a deliberate
behaviour change and the only one here.
@MfaExempt carries the reason for the other kind of exemption: the routes
that let somebody complete setup. Without them the policy locks out every new
account, which is why the middleware had its second list.
One ordering trap on the way: global guards run before route guards, so the
auth guard standing down on a controller with its own @UseGuards chain left
req.user unset when the MFA guard ran behind it, and it waved everything
through. AUTH-020 caught it. It resolves the user now without refusing - the
declared chain still owns the 401.
Nothing in src/nest/** was an APP_GUARD. Every protected route said so
itself, 62 times, which means a forgotten @UseGuards was a silent bypass
rather than an error. GlobalAuthGuard inverts that: a route is authenticated
unless it carries @Public() or @OptionalAuth(), or declares its own guard
chain.
Reading the declared chain is what makes this behaviour-neutral. Nest runs
global guards before route guards, so without that check the global guard
would 401 an anonymous request to an addon-gated controller before its
AddonGuard could answer 404 - inverting an order four e2e suites pin, and
leaking which addons are installed. A controller that declared a chain keeps
deciding for itself.
It is a new class rather than JwtAuthGuard gaining a Reflector: that guard is
constructed positionally in its own test with a context stub that has no
getHandler, and the two want opposite behaviour anyway.
@Public() takes a required reason, and the reasons are the diff worth
reading: the invite token IS the credential, help assets are loaded by <img>
which cannot send one, logout has to work with an expired session or nobody
can sign out.
The real deliverable is the inventory. validateRouteGuards runs in buildApp
next to the body-contract gate and refuses to start when a route answers
anonymously without being on a reviewed list of 25 - or when a listed one no
longer does. An allow list nobody prunes stops being a list.
Default-deny found two controllers on its own: SharedController and
OidcController both sit in a file next to a guarded controller, so a
file-level search for unguarded ones missed them. 37 tests went red and named
them. That is the mechanism working before the change shipped, not after.
Sixteen of the thirty-four entries in the body-contract allow list were the
JourneyController, and a strict schema already existed for six of them. That
is the trap: eight of these handlers validate their own body and answer with
their own text - 'Title is required', 'trip_id required', 'provider and
asset_id required', 'entry_date is required', 'user_id required' - and those
bodies are pinned by e2e and integration cases. Adopting the strict schemas
would have the pipe answer first, with a different body. That is a
client-visible change wearing the costume of a refactor.
So the schemas follow the memories doctrine instead: loose objects, and every
field the handler type-checks itself is unknown here. They describe the shape
that already works, which is what the boot gate needs, and every hand-rolled
400 keeps answering. The spec inverts accordingly - it now asserts that each
schema ACCEPTS what the handler still rejects, which is the property that
makes the claim checkable.
Two of them are optional at the top level, and that is not decoration: a
multipart upload with files and no text fields arrives with no body at all,
and JOURNEY-INT-045 pins 'No files uploaded' for it. Without the optional the
pipe answered 'body: Invalid input: expected object' instead. The suite
caught it.
Backup joins them - its schema was already a passthrough, because the service
normalises the body itself.
Allow list: 34 entries down to 17. Tags and the two OAuth entries stay for
now; tags reads @Body('name') per property, which is a signature change
rather than a contract one.
src/nest/** had no owner guard at all. Handing a trip over and creating or
deleting guests each asserted access.user_id !== user.id inline, and the
string they answer with differs per route, so the message travels in the
metadata rather than living in the guard.
TripOwnerGuard deliberately does not inject PermissionsService.
checkPermission returns true for every admin and the trip actions are
admin-lowerable, so routing ownership through it would quietly let any admin
transfer somebody else's trip. OWNER-004 pins that: an admin who is not the
owner gets 403.
places gets protection on the five routes where a guard is provably safe:
list, get, unrate, image and delete. Not on create, update, the imports or
the uploads. A guard runs before the pipe, so guarding a route with a
DTO-typed body would answer the trip 404 where the suite pins a 400 - that
includes PUT :id/rating, which is why the rating pair is split. The three
multipart routes are excluded for the other reason: answering while the
client is still streaming destroys the socket, and the caller sees
ECONNRESET instead of a status.
DELETE :id carries @RequirePermission('place_edit'). The first attempt at
this lost that check and answered 200 where 403 belonged, so the e2e asserts
the argument list rather than the status: a check that returns the right code
for the wrong reason fails now.
TripsService injected fourteen other services. Almost all of them were there
for two read aggregates - the MCP trip summary and the offline bundle - and
for the member roster, not for anything the write path does. Reading is what
fans out; writing is not.
Membership goes to trip-members/: listMembers, addMember, removeMember,
transferOwnership and the four guest methods, plus the six routes, which keep
their paths. It is a NEW module, not the trip-membership/ leaf next door:
deleting a guest erases their plugin data and re-splits their expenses, so it
needs auth and budget, and AuthModule already imports that leaf. Giving it
those dependencies there would have closed a hard cycle. This one is a sink -
nothing imports it back except trips.
The read aggregates go to trip-read-model/, which is where the nine-service
fan-out now lives.
What is left is the write core plus day generation, on eight parameters
instead of fourteen: database, reservations, days, permissions, budget,
vacay, realtime, unsplash. The last four are all reachable from updateTrip's
date-shift transaction, which is why the count does not fall to three.
Trips still reaches budget through ReservationsModule, so trips.bridge.ts
survives this. Saying otherwise would have been the easy claim to make and it
would not have been true.
The routes lived in reservations/ while the SQL lived on DaysService, so the
controller delegated every call across a domain boundary to reach its own
data, and neither module owned the thing outright. The seven statements move
into AccommodationsService unchanged and the domain gets its own directory,
module, DTO, RPC and MCP surface.
What stays in days/ is the reorder side: assertNoInvertedAccommodation and
resyncAccommodationDays are about re-dating day rows, and the stay is the
thing being carried, not the thing doing the carrying.
Two edges fall out of this. AccommodationsService drops its DaysService
dependency, and DaysMcp drops DatabaseService and PlacesService, because
create_place_accommodation was their only user - which lets DaysModule stop
importing PlacesModule and closes one leg of the latent
Days -> Places -> Assignments -> Days cycle.
DAY-SVC-016..026, 029 and 035 move to accommodations.service.test.ts with
their case ids intact. The reservations e2e container assembles both modules
now: it builds a single domain module on purpose, which is exactly what
catches a move like this, and it caught it.
days/ carried two fachlichkeiten with a full file set each: days.controller,
days.service, days.mcp, days.rpc, days.dto next to the five day-notes ones.
The note files move out unchanged; because the new directory sits at the same
depth, not one import path inside them changed.
The two modules deliberately do not reference each other in either direction:
dayExists is raw SQL on the days table in the note service, and DaysService
reads day_notes the same way, so a module edge would buy nothing and cost a
cycle.
DayNotesModule has to be registered in PluginsRuntimeModule as well as
AppModule - DayNotesRpc owns four of the 113 wire methods and the registry
boots with requireTotalCoverage, so leaving it out would stop every test app
that assembles the plugin runtime.
The days e2e container assembles both modules now: the note routes still nest
under the days prefix, and a container that builds only DaysModule would 404
on them while production served them fine.
The second flake recorded as 'costs' is not a timing flake. It is a live
HTTPS call: budget.service.ts settlement() reaches getRates()
unconditionally, unlike the two guarded call sites above it, and
ExchangeRatesService fetches api.frankfurter.dev with a 10 s abort.
tests/integration/budget.test.ts stubbed fetch nowhere at all, so the suite
talked to the internet - fast when the API answers, ten seconds per call
when it does not. The stub fails closed, which changes no assertion: every
settlement case here is single-currency, so rates never enter the
arithmetic. tools-budget-advanced had the stub inside a single case; hoisted
so a second settlement case cannot quietly reintroduce the call.
The five ReDoS budgets in maps.service.test.ts measured wall-clock time,
which on a loaded machine reports how long the test waited for a core.
MAPS-024 has gone red at 523 ms against an unchanged file for exactly that
reason. They now measure CPU time, which is what catastrophic backtracking
actually burns. This does not soften them: CPU time is always <= wall-clock
time, so every budget got stricter, and pool:'forks' keeps the measurement
to this file.
The all-trips feed merges many trips into one calendar. It did that by
generating each trip's finished ICS and then scanning the text back apart:
one pass for the VEVENTs, one for the VTIMEZONEs, plus a string replace to
splice the refresh hints into the header. The scans had to be structural
rather than regexes, because a user-supplied SUMMARY can legitimately
contain the literal 'END:VEVENT' and a lazy match would truncate the event
there.
CalendarService now returns the parts - calendar name, filename, VTIMEZONE
blocks keyed by TZID, and the VEVENT blocks - and exportICS serialises them.
Both scanners are gone.
CAL-010 pins the whole serialised document as a snapshot, with DTSTAMP and
the UID counter normalised so it pins the calendar rather than the minute it
ran in. It is what proves this changed no bytes: it was taken before the
model went in and still matches.
Two tests still pointed at the health controller deleted in d37af204:
health.di.test.ts loses its only subject, and the AppModule wiring test now
resolves FeaturesController instead - it is the one test that builds the
whole container, so it should resolve something that still exists.
exportICS and its four module helpers were 330 lines of RFC 5545 inside
TripsService, and FeedsService had to import the whole trip aggregate to get
an ICS string out of it. They move to a calendar domain unchanged - same
statements, same escaping, same folding, same VTIMEZONE fallback - so the
emitted bytes are identical on both the download and the feed.
The 21 unit cases move with the code and keep every assertion, which is what
makes that claim checkable rather than asserted. FeedsModule now imports
CalendarModule instead of TripsModule, and its e2e container drops from the
trip aggregate to two small modules.
NotFoundError/ValidationError move to common/ so a second domain can raise
them without importing the first one for its error types; trips re-exports
them, so the nine files that import them from there are untouched.
Fixes a harness that tsc cannot see on the way past: mcp-test-controllers
built TripsService with 12 arguments for 14 parameters, leaving unsplash and
userCleanup undefined.
GET /api/health/features lived in booking-import, which meant the client's
capability probe was reachable only because that domain happened to be
registered. Move it to health/ and give health an actual module instead of
two loose entries in app.module.
The extractor probe follows into a leaf module rather than health importing
BookingImportModule: features() needs isAvailable(), not llm-parse,
reservations, budget, maps and places.
The /api/_nest/{health,me,echo} smoke endpoints go. They proved boot, DI,
the cookie guard and the Zod pipe back when nothing else did; every one of
those now has real domain coverage, and nothing outside their own test ever
called them. The container probe is GET /api/health on the raw express layer
and is untouched.
Three things you could see on dev1.
Pictures were "roughly right" rather than right, because a coordinate search
only knows what was photographed within 300 m. Wikidata's P18 is one picture
somebody chose to represent that exact object - "Zurletzteninstanz.jpg" for the
restaurant, not a statue down the road - so it now leads, then the Commons
category, and geosearch is the last resort it should always have been. Small
venues carry a wikidata tag far more often than a Commons category, so this is
also what reaches the places the other sources miss.
Descriptions stayed on Wikipedia after the switch to Wikivoyage because the
week-long result cache predates it. It carries a format version now; a release
that changes the shape or the sources invalidates its own entries instead of
serving last week's for another six days.
Ratings and opening hours were being fetched from Google and then dropped on the
floor - the fact list only looked at OpenStreetMap tags. Both are shown now,
whichever provider supplied them.
Also: the column follows a place opened from the map, not only one picked from
search. A POI tap arrives as prefillCoords and never touched the selection, so
the column kept showing the previously added place.
Pictures are a tight three-column grid and only the picture in play is credited
in full, on one line. Crediting all of them at two lines each drowned the
column, and the licence attaches to the one that gets used - whose credit also
stays visible in the inspector after saving.
The request carries the place record the dialog already fetched, so the server
skips a details lookup it was making in parallel with the client's own. That
lookup was measured at 12.8 s for a large OSM relation.
A restaurant got nothing: no wiki article, and a coordinate search around it
returns whatever was photographed nearby rather than the place itself. Three
changes, all on free sources:
OpenStreetMap facts. A tagged restaurant carries its cuisine, its opening hours,
a link to its menu, outdoor seating, step-free access and diet tags. Those
arrive with the details lookup the dialog already makes, so showing them as
chips costs nothing and is the single biggest gain for exactly the places that
were empty before. Only positive facts are listed - "no outdoor seating" is
noise.
Commons categories. Where a place has a wikimedia_commons tag, its category is
pictures OF it; geosearch falls back in when the category is empty.
Wikivoyage before Wikipedia. Same API, same licence, but it describes a place
for someone about to visit rather than opening with area in square kilometres.
Two sentences instead of three, because this sits beside a form.
The column also follows the section-heading idiom Vacay and the dashboard share
(uppercase overline, wide tracking), its header matches the Saved places panel
beside it, and a lead picture with a row underneath replaces the two-column
grid that left a hole on every odd count. Credits wrap over two lines instead
of truncating - a cut-off name credits nobody.
Saved places stops after ten with a Show more button; search and filters still
run over everything, not just the visible page.
Form spacing is one rhythm (12px between groups, 8px inside, 4px label to
field) and Notes matches Description in height. The OSM notice fits on one line
and no longer claims there are no photos, which stopped being true when the
column started fetching them from Wikimedia. The clipboard hint is gone; paste
still works.
The nightly photo sweep deletes every candidate nobody picked, which is most of
them. The cached result still pointed at those files, and the read filtered them
out - so the morning after, a place with perfectly available pictures showed an
empty column, and kept showing one for the rest of the week the entry lives.
A missing candidate now invalidates the whole entry instead of shrinking it.
The column showed "details could not be loaded" for every place. The endpoint
was answering correctly, just after ten seconds, and the shared axios instance
gives up at eight.
The downloads sat in a loop: five Commons hits meant five round trips and five
decodes in sequence, even though the slot limiter already allows five at once.
Both provider listings now run together and the bytes are fetched concurrently,
which brings a warm request back to about a second.
Candidate keys move with it. They now number the provider's list rather than the
surviving strip, so a picture that fails to download leaves a gap instead of
renumbering the ones behind it - a key has to point at the same file on the next
request or the cached result goes stale against the photo cache.
The client asks for enrichment with a longer timeout, and the Wikimedia calls
carry one of their own: a cold TLS handshake from a fresh container was measured
at eight seconds on its own, and no pictures is a better answer than a request
that never comes back.
NOTICE knew about geoBoundaries, OSM and OurAirports but not about Wikimedia,
Wikipedia or Google Places, all of which TREK has been calling for a while.
Also fixes two dead paths in the place contract - it still pointed at
server/src/routes/places.ts and services/placeService.ts, both long gone.
Accommodation cards read the place from page-local state, so a new thumbnail
only showed up there after a reload. place:updated now asks the planner to
reload them, but only when the image actually changed - renaming a place must
not trigger a refetch.
The inspector shows the stored credit under the avatar. Crediting a CC BY-SA
picture only in the moment it is chosen is not enough; the obligation travels
with the picture.
Also lets a GIF hero through to the PDF. Uploads accept gif, the PDF extension
check did not, so one silently fell back to the category icon.
Picking a search result fills a column on the left of the add-place dialog -
pictures and a description, no extra click. Someone who already knows the place
ignores it and saves exactly as before.
The column stays mounted with an empty state instead of appearing on the first
pick, otherwise the dialog jumps sideways mid-typing. Requests are cached in
memory and in sessionStorage and aborted when the selection changes, so clicking
back and forth between two results does not pay twice.
Every thumbnail carries its author, its licence and a link to the file page.
That is not decoration - most of these are CC BY-SA, and reusing one without
naming the author does not satisfy the licence. The description is shown but
never written into the place on its own; a button copies it across, and it is
disabled while the user already has a description of their own.
Desktop only. The mobile sheet is untouched.
Reads fail-open, unlike the three older places_* switches. Those read
=== 'true', which is why migration 185 had to backfill a row on every existing
install; there is nothing here that warrants repeating that.
The three hand-rolled toggles next to it move to ToggleSwitch while I am in
there - they were bare buttons with an inline transform and no aria-pressed or
label, so a screen reader could not tell their state.
Fans the new provider primitives out into one answer for a place that is being
looked at but not saved yet: a strip of pictures and a description, each with
its source.
Free sources are the normal case - most instances have no Google key at all, so
Commons and the OSM tags have to carry this on their own and Google is what an
admin adds on top. Pictures are copied into the existing photo cache under
"<placeId>~pN" keys rather than hotlinked: Google's photo URLs expire (migration
107 had to repair exactly that once), and a provider URL in the strip would send
every viewer's address to Google or Wikimedia. Whatever nobody picks is
unreferenced and the nightly sweep clears it out.
The result is cached for a week in place_details_cache under a third expanded
value, so no migration. Rate limited per user, and a provider being down answers
empty rather than failing the dialog.
Commons geosearch already returned five files and we threw four away, so a
whole strip costs nothing extra. It now comes back with the licence metadata
too: without LicenseShortName and the file description page there is no way to
credit a CC BY-SA image, and a picker full of them needs that.
Wikipedia descriptions resolve through the OSM wikipedia tag rather than
guessing the article from the place name, which lands on the wrong article for
anything ambiguous. Google photo references and photo bytes are separate calls
now so a caller can list without downloading, and the editorial summary has its
own lean field mask - asking for it through the expanded details would drag
reviews along and move the call into the pricier tier.
Asserting the module registration meant importing AtlasModule, which pulls
AtlasMcp and the whole MCP SDK behind it into a file that otherwise
constructs a controller with a stub. It does not resolve when the file runs
on its own.
PROFILE-014 already covers the registration, and covers it better: it calls
GET /api/auth/travel-stats through the full app.
atlas.service.test.ts carries the admin1 GeoJSON bundle warm-up from #48, and
its ATLAS-UNIT-021..024 cases sit close to the 15s limit whenever the machine
is busy. Eighteen more cases in the same file is not what they need.
Same DB-per-file pattern, no shared state, case IDs unchanged.
The last of AdminService's borrowed SQL. oauth_tokens belongs to oauth/,
which already had a user-facing view of the same table; AdminService carried
a second one only because the panel route is under /api/admin.
Both survive, under names that say which is which. They are not duplicates:
the user-facing listOAuthSessions is scoped to one user, the admin one spans
all of them and joins the owner username, and the admin revoke kills the
client session for whoever owned the token.
OauthModule gains its first export. Its header said nothing inside the
container injects the domain, which stops being true here.
invite_tokens gates signup. Its consumer is the registration path in
AuthService, not anything in admin/, which held the four management methods
only because the routes sit under /api/admin.
Worth spelling out because the names invite it: invite_tokens and
trip_invite_tokens are different tables for different jobs, and this one has
nothing to do with adding a member to a trip. An earlier reading of the plan
had these going to trip-invite/, which would have been the wrong home.
The routes keep their paths and their admin guard; AdminController injects
the service instead of carrying another domain's SQL.
Also recorded, in the file, why the app-settings pair did NOT move to
settings/ with the other app_settings readers: updateAppSettings runs the
lockout guard that stops an admin disabling every login method, and that
guard is auth's own resolveAuthToggles. Moving it would put an AuthModule
import inside a leaf module half the container pulls in.
The second cut into AuthService. Display settings, avatar, the third-party
API keys a user pastes in and the member directory are not identity: no
password is checked, no session is issued, no MFA secret is touched.
UserProfileService takes those eight methods verbatim, with DatabaseService
as its only injected dependency. AuthService keeps register, login, JWT,
MFA, password reset and account deletion, and is now roughly 900 lines
lighter than it started this branch.
Not exported from AuthModule: only its own controller needs it, and an
export nobody imports is noise.
The controller test harness builds AuthController positionally, so the two
new parameters landed in optional slots and the stubs for the moved methods
moved with them.
AuthService carried identity, profile, settings and tokens at once. Tokens
are the cleanest of those to cut loose: they touch one table plus the
ephemeral-token store, and nothing in them needs to know how a password is
hashed or how a session is established.
TokenService takes the six user-facing methods plus verifyMcpToken, and the
two admin-side ones that only lived on AdminService because the admin route
did. The pairs are genuinely different queries rather than duplicates, so
both survive under names that say which is which: the user-facing ones scope
every statement by user_id and treat the session sweep as best-effort, the
admin ones do neither.
Left behind on purpose: verifyJwtToken, which is login identity and belongs
next to the cookie logic, and isDemoUser, which is a demo gate that happens
to read the users table.
TokensModule is a leaf with nothing but DatabaseService, because auth, admin
and the MCP transport all import it and anything added here lands in three
graphs at once.
Both bridges and both controller test harnesses construct these classes
positionally and had to be pulled along; the new argument goes in an
optional slot so untouched call sites stay untouched.
getTravelStats sat on AuthService for historical reasons only. It reads
places, trips, visited_countries, place_regions and reservation_endpoints,
and subtracts the countries hidden in Atlas, so every one of its sources is
an Atlas concern.
Keeping it there had a real cost: AuthModule had to import AtlasModule, and
that edge is the reason atlas.mcp.ts reaches AuthService through auth.bridge
instead of injecting it. The direction is now free to reverse.
GET /api/auth/travel-stats keeps its path. The client calls it there, so the
route moves to a small controller in atlas/ that holds the /api/auth prefix
rather than changing the URL. PROFILE-014 runs against the full app and
proves the endpoint is untouched.
KNOWN_COUNTRIES moves along, since getTravelStats was its only reader. Both
bridges construct AuthService by hand and needed the argument dropped.
default-user-settings, notification-preferences and oidc leave AdminController.
The first two were pure pass-throughs on AdminService that did nothing but
forward and cast; the third had its SQL there for a domain that already had a
module of its own. AdminService loses SettingsService and
NotificationPreferencesService entirely.
Two things came out of the move rather than going into it.
The OIDC lockout guard now has tests. It refuses to clear the SSO config while
password login is disabled — otherwise the instance has no SSO to log in through
and no password form either — and nothing covered that branch before, because it
sat among admin tests that only exercised the happy path.
The admin e2e app had to learn about the split. It assembles AdminModule alone,
so the moved routes 404'd there while working in production; it now imports the
three domains too. That is the same failure mode the module-provider guards
catch for RPC classes, and it is worth knowing that an e2e app which builds a
subset will find it first.
Also fixes the audit route note: audit-log stays on AdminController on purpose.
AuthModule imports AuditModule, so an admin-gated controller in AuditModule
closes a cycle — the same trap the addons toggles hit.
Twelve admin packing-template routes sat on AdminController and reached
PackingService through eleven pass-through methods on AdminService that did
nothing but forward and cast. PackingService already owns all three template
tables, so the routes move to AdminPackingTemplatesController next to it and the
pass-through block is gone. AdminService no longer injects PackingService at all.
The three places flags were raw app_settings SQL on AdminService, sitting beside
bag-tracking and collab-features which it already delegated to AddonsService.
They move to AddonsService, which owns every other flag, and AdminController now
calls it directly rather than through delegating methods.
The toggle ROUTES stay on AdminController, which is a departure from the plan and
worth stating: an admin-gated controller in AddonsModule makes that module import
AuthModule, and PluginGuardsModule imports AddonsModule precisely because it is a
leaf — Nest refused to build the graph. Same for the addon CRUD, whose PUT
cascades into PluginRuntimeService. Both notes are in the code.
Paths, response shapes, the {error,status} envelope, the create-201 split and
every audit action are unchanged.
assignments, budget, collab, accommodations, reservations and booking-import
drop their copied requireTrip/requireEdit helpers for the guard. Status codes and
bodies are unchanged.
places is deliberately NOT converted, and that is the finding of this PR rather
than an omission. Its e2e suite documents that the body pipe answers 400 ahead of
the trip 404 — a parity shift somebody made on purpose — and a Nest guard always
runs before the pipe, so converting it silently reverses that ordering. The first
attempt also dropped the permission check on DELETE, which the integration suite
caught as a 200 where a 403 belonged. places keeps its in-handler check, and
README now says why so the next person does not rediscover it the same way.
collab applies the guard per handler for the same reason files does: its note-file
upload route would otherwise be refused before multer had read the body, and a
response sent mid-upload destroys the socket.
Two coverage notes. The per-domain thresholds are regenerated because deleting a
well-covered helper lowers a file's ratio without anything becoming less tested —
the drops are 0.4 to 1.2 points across six domains and match the helpers removed.
And the atlas integration suite gets the same admin1-bundle warm-up the unit suite
got: ATLAS-013 was paying 3.1s for the one-off decompression inside its own 5s
budget, which is why it only ever failed under load. It now runs in 116ms.
Every trip-scoped controller carried its own three-line requireTrip, and a
requireEdit next to it: eighteen copies of a security check, each of which had to
remember that no access is a 404 'Trip not found' and never a 403 — the
distinction that stops a stranger learning which trip ids exist. One of them
getting it wrong would have looked like nothing.
TripAccessGuard resolves :tripId once, refuses what the user cannot reach and
hands the row to the handler through a @Trip() param decorator (a param
decorator rather than a req.trip cast, because the request type that used to be
widened for that is gone). @RequirePermission carries the same action string the
domain services pass to checkPermission, so the HTTP and MCP paths cannot drift
apart on which right they demand. Status codes and bodies are unchanged.
Piloted on days, day-notes, todo, packing and files. Two things worth knowing:
Files applies the guard PER HANDLER and leaves the upload route on its own
check. Guards run before interceptors, so a class-level guard would answer 404
before multer had read the multipart body, and a response sent while the client
is still streaming gets the socket destroyed — the caller sees ECONNRESET
instead of the 404. Refusing after the body has been read costs the bandwidth
and keeps the answer readable.
The services keep verifyTripAccess and canEdit: the large majority of their
callers are *.mcp.ts tools, which never pass through an HTTP guard. canEdit lost
its incidental controller-test coverage in the move, so it now has direct tests
asserting the action and the shared flag — the same pair the guard reproduces.
Also fixes the backup zip-slip test that failed only on CI: it asserted a
drive-letter path, which is absolute on win32 and an ordinary relative segment on
Linux, so the branch it was guarding never fired there.
src/services/backupService.ts moves to src/nest/backup/backup.impl.ts, imported
only by its own domain now that the controller reaches it through BackupService.
src/services/ is down to airtrail/.
It stays a module of free functions rather than becoming methods on the service.
The restore path closes and reinitializes the core DB handle, which is the single
most dangerous sequence in the server, and rewriting its shape in the same step
as the move would make a regression there impossible to bisect. The db-lifecycle
question is the last step of this fold, not part of it.
The move also pulled 512 lines out of ungated src/services/ into a gated domain,
which is how the per-domain ratchet caught that the per-entry extraction loop had
no tests at all: the zip-slip refusal, the running decompressed-byte cap (the
declared size in the central directory is attacker-controlled, so only the
running total catches a lying archive), a corrupt stream not being dressed up as
a size refusal, and the reopen failure that has to report restart-required rather
than success. Six cases now cover it, and backup goes from 84% to 96% statements.
PluginsModule was 21 controllers, 24 domain imports and 10 providers in one
class, which made it the hub most of the Nest graph ran through: AdminModule
imported it for a single cascade-disable call and inherited all of it. It is now
PluginsRuntimeModule (supervisor, capability router, hook contracts),
PluginContributionsModule (the 14 read-only hook controllers, moved to
contributions/), PluginOAuthModule (the credential leaf, moved to oauth/) and a
composition root that keeps the install and delivery controllers. AdminModule
imports the runtime alone.
The three plugin files that reached for the db singleton alongside the injected
DatabaseService no longer do. plugin-host-state and signature-status take the
connection from the caller, and the three per-user settings functions that were
bare exports 'so the RPC host wiring can call them directly' are now
PluginUserSettingsService - the host wiring is providers, so it injects them.
Coverage moves from one repo-wide 80% to a per-domain ratchet. The single number
let a well-covered domain subsidise a thin one: booking-import sits at 50% and
integrations at 20%, and the aggregate still cleared the bar, so either could
have lost another ten points unnoticed. Each domain now carries its own measured
floor, regenerated by scripts/coverage-thresholds.mjs.
Two test-side findings on the way:
expect(undefined).toContain(SomeClass) PASSES in this vitest version - it only
throws for a string needle. Every '<X> is listed in its module providers' guard
was written that way, so all twelve of them silently stopped guarding the moment
their class moved. They go through expectRegisteredProvider now, which asserts
the array shape first.
The atlas region test paid for the one-off decompression of the multi-MB admin1
bundle inside its own 5s budget, which is why it failed under load and passed
alone. The bundle is warmed in beforeAll instead.
With requireTotalCoverage on, a test app that assembles a subset of AppModule
now fails to boot: the admin e2e suite pulls PluginsModule in through
AdminModule, but WeatherModule was only ever reachable from AppModule, so
weather.get had no handler and validate() refused the start.
The invariant the gate needs is that importing PluginsModule pulls in every
domain that owns part of the wire surface. WeatherModule was the one domain
missing from that list.
The last 20 wire methods move onto the decorators: db.* and the three
unconditional methods (DbRpc), meta.* (MetaRpc) and the host-mediated surface
of users/broadcasts/notify/ai/oauth/scheduler (HostSurfaceRpc). That closes the
ledger at 113/113, so the legacy has() blocks in rpc-host.ts are gone and the
router is down to what was never domain-specific: build the granted subset of
the dispatch map, dispatch, audit, map errors.
PluginHostDepsFactory was the 26-argument wiring sheet behind those blocks. With
the handlers in the domains it has two dependencies left, so it is now
PluginRpcHostFactory: the plugin's own sqlite handle, the inter-plugin router
bound with this plugin as the caller, and the audit sink.
The 18 host-to-plugin hook calls get the same treatment. They used to spell out
the fn name and the timeout at each controller, which meant nothing checked that
a hook:* grant on the consent screen had a consumer behind it, that the fn
matched what the SDK documents, or that two call sites agreed on the budget.
PluginHooks declares all 18 with @PluginHook and performs them, and the
controllers call it by name.
With both halves complete, requireTotalCoverage goes on: a KNOWN_METHOD without
a handler, or a hook without a consumer, now fails app boot instead of surfacing
as a PERMISSION_DENIED nobody can explain.
Aiming the new files at 95% rather than at the 80% gate turned up something worth
fixing rather than testing around: `error.issues[0]?.message ?? 'bad input'` had
been copied into twenty handlers. That is twenty two-branch fallbacks that no test
can reach, because a real Zod failure always carries at least one issue. They are
now one schemaMessage() in rpc-params.ts, with one test that does reach both sides.
The rest of this is the branches the handlers inherited from the router: the
`!= null` link ternaries, the "service said no" paths, the ValidationError and
NotFoundError translations, and a reject case per schema-validated method.
Two limits found while measuring, both real rather than laziness. istanbul counts
each @PluginController line as a branch and only ever takes one side, so eight
branches across the rpc files can never be covered and a two-method file caps out
around 50%. And the three update schemas stay deliberately permissive, because the
migration's rule is parity over tightening, so their reject branch is defence in
depth no payload reaches. Everything else went from 50 uncovered branches to 29.
collab, journal, atlas, vacay and collections: 32 more methods, so the registry
owns 93 of 113. rpc-host.ts is at 565 lines, down from 1332, and the deps factory
at 384, down from 944.
These five are the risky group, because their addon gates lived inside the deps
factory's closure rather than in the router. A handler that calls its service
directly still works, just without the check, and neither the compiler nor the
router tests would notice. So addon-gates.rpc.test.ts walks all 37 gated methods as
one table and asserts two things for each: with the addon off it is refused with
that addon's message, and the service is never reached. The table length is pinned,
so a later method added without a row is a failure rather than a silent gap.
One ordering detail worth knowing: on the atlas writes the router evaluated its
params before the addon gate inside the closure, so a malformed country code still
reports BAD_PARAMS rather than the addon. The handlers keep that order.
The collections service reports refusals as status-tagged errors rather than typed
ones, so mapCollectionError came across with the handlers: 403 and 404 become
RESOURCE_FORBIDDEN, 400 and 409 become BAD_PARAMS, and anything untagged stays a
HOST_ERROR instead of being dressed up as a permission problem.
The branch threshold over src/nest/** dipped to 79.89% after the move: the new
*.rpc.ts files carry the defensive branches the router closures had (?? '',
|| current, optional cascades), and none of them were exercised.
Adds the cases that reach them: a booking whose cascade does not fire, an update
that keeps the stored title, a bag colour of the wrong type, a non-array userIds,
a missing bag across all three bag writes, and the block delete without a partner
reservation. Also corrects two assumptions I had written into the booking tests:
title and type are required by the shared schema, so the handler's own fallbacks
sit behind it as defence in depth rather than as the first line.
Eleven more methods, so the registry owns 61 of 113. rpc-host.ts is at 771 lines,
down from 1332.
costs is the first addon-gated domain, and the addon check sits in a different
place in each of the two reads: inside the membership callback for getByTrip, so a
caller without trip access never learns whether the addon is on, and before
everything for listMine, which has no trip to check. Both orderings are asserted.
The writes keep their own refusal wording ("no permission to edit costs on trip N")
instead of going through requireTripEdit, which words it differently.
The booking paths are cascade-heavy and stay that way: creating a reservation can
create an accommodation and a budget item, deleting one removes both again, and
deleting a lodging block takes its partner reservation and budget item with it.
Each cascade is a broadcast the web app relies on, so they are asserted by event
name and order.
CostsRpc reaches the trip list through trips.bridge rather than by injection.
BudgetModule cannot import TripsModule, because TripsModule already imports
BudgetModule; budget.mcp.ts uses the same escape for the same reason.
Twelve more methods, so the registry owns 50 of 113. rpc-host.ts is down from 1332
lines to 830 and the deps factory from 944 to 500.
trips.update keeps its two field-level gates: archiving and changing the cover each
need their own admin-configurable permission on top of trip_edit, so a plugin
cannot do either on a trip it may otherwise edit. Member management keeps its own
grant for the same kind of reason, since adding a member grants trip access, and
the owner still cannot be removed through it because that would orphan the trip.
canEditTrip stays in HostDeps for now: the db:meta gate still resolves an entity's
edit permission through it and has not moved yet.
Eight more methods, so the registry owns 38 of 113 and rpc-host.ts is under 1000
lines for the first time.
itinerary.* lives in assignments/, not in days/, because AssignmentsModule already
imports DaysModule and putting the handler the other way round would have closed a
cycle. It keeps riding on day_edit, which is what the app gates it on.
The delete path in places carries an ordering constraint that is silent when
broken, so it has two tests of its own: the journey hook has to run BEFORE the row
goes, because journey_entries.source_place_id is ON DELETE SET NULL and afterwards
there is nothing left to detach, and the place id has to be scoped to the trip
before the hook runs at all, because the hook keys on the place alone and would
otherwise detach another trip's entries even though the delete itself is refused.
Left over from writing the demo-mode case: a cast statement with no effect, which
eslint flags as no-unused-expressions. It is the one error in an otherwise
warnings-only run, so `eslint .` exits 1 and the Server Tests job fails on lint
before it runs a single test.
Six more methods, so the registry owns 30 of 113, and the deps factory is down to
660 lines from 944.
Files are the awkward corner of the plugin surface and stay that way on purpose.
Reading a file's bytes is a separate grant from reading its metadata, because a
passport scan is more sensitive than its filename, and the three writes use three
distinct rights instead of one domain action. FILES-RPC-006 pins that by watching
which permission each operation actually asks for.
This is also the only plugin path that touches disk, so the tests lean on it: the
extension blocklist, the size cap checked before the read rather than after, a
path that resolves outside the uploads root, and the demo-account upload refusal
that mirrors the REST guard.
Nine more methods, so the registry owns 24 of 113.
The #858 privacy logic existed three times: PackingService had the primitives, the
controller layered emitToViewers and broadcastUpdate on top privately, and the
plugin deps factory carried a standalone copy of all four, with a comment asking
for them to be kept in lockstep by hand. Migrating packing meant either copying
them a fourth time or fixing that, so the two private controller methods moved
into PackingService and the factory copy is gone. Controller and PackingRpc now
call the same code, and the controller suite proves it by passing with all 45
assertions unchanged.
The four public/private transitions get one test each, because each one is a way
to leak a private item to the whole trip room, plus a test that the privacy is
read before the write rather than after: reading it afterwards would report
"already private", skip the room delete, and leave the item on every member's
screen.
categories, weather, rates, todos and daynotes: 11 more methods, taking the
registry to 15 of 113. Handler bodies and messages are verbatim; the ledger test
tracks the split.
Two things the move surfaced rather than introduced. PluginGuards could not stay
in PluginsModule: that module imports 22 domain modules, so a domain importing it
back for the guards would close a hard cycle and the only way out would be
forwardRef. It now lives in a PluginGuardsModule leaf that imports nothing which
leads back, the same shape MailerModule got for AuthModule.
And the todo and daynote writes broadcast. Those events are not decoration, they
are what keeps an open session from showing stale data after a plugin write, so
they came across from the deps factory with the handlers, and there is a test
that a failed write broadcasts nothing.
Worth knowing for the next domains: todos are edited under 'packing_edit', not
under a todo-specific action, and TODO-RPC-003 pins that. Each new class also
asserts it is actually listed in its module providers, since a decorated class
nobody registers is invisible to discovery and answers PERMISSION_DENIED with no
other symptom.
Wires the registry into the host and takes the first domain across, so the kit
lands with a worked example rather than as scaffolding nobody uses yet.
The registry is an additive, optional fourth constructor parameter on
PluginRpcHost, so all 107 existing construction sites keep compiling untouched;
that is what makes the rest of the rollout a per-domain move instead of a
big-bang one. bindInto runs LAST in the constructor, so a method still owned by
one of the has() blocks collides loudly instead of being silently overridden.
tags is the right pilot: four methods, a service that is already DI-native, and
it exercises the read/write permission split, the userless refusal and a
BadParams write. Handler bodies and messages are verbatim; only deps.foo(...)
became this.tags... and uid became ctx.actingUserId.
dispatch now strips the supervisor's _inv marker. The supervisor resolves the
acting user from it before dispatch and then passes the request on untouched, so
it has been visible to every handler and is in effect a reserved param name.
Nothing reads it: auditResource takes named fields only, and the one handler
that touches the params object as a whole (packing.setBagMembers) reads a named
field off it. Dropping it keeps a future strict parse from rejecting calls every
shipped plugin makes.
Two tests carry the rollout from here. rpc-coverage.test.ts derives both owner
sets from the code and proves they partition the method vocabulary exactly, so a
PR that moves a method and forgets to delete the legacy registration fails in
CI. And because a decorated class that no module registers is invisible to
discovery, with every call to it coming back PERMISSION_DENIED and no other
symptom, tags.rpc.test.ts asserts TagsRpc is actually listed in providers.
A decorated handler cannot reach the router's private tripRead / requireActor /
requireTripEdit, nor the canEditTripAs and requireAddon that live inside the
deps factory's closure. PluginGuards holds all of them so a *.rpc.ts class can
inject what it needs.
Every message is copied character for character, including the two asymmetries
that look like typos and are not: a trip read says "trip reads", while
requireActor appends the word "writes" to whatever noun it is given, and some
domains throw their own inline variant rather than calling it at all. The tests
pin the exact strings, not just the refusal, because shipped plugins read them.
The method-to-permission association exists twice today: declaratively in
METHOD_PERMISSION, and imperatively as 44 if (has(...)) blocks in rpc-host.ts,
with nothing asserting the two agree. protocol-paths.test.ts checks that each
method has some permission, never that it is the same one the router gates on,
and plugin-audit.ts inherits any drift silently.
@PluginMethod binds the two together. Its permission argument is typed as
MethodPermission<M>, so it resolves to the single literal the table assigns to
that method and a mismatch does not compile. validate() re-checks the same
argument against the same table at boot, because types evaporate under an as
cast. @PluginHook does the same against HOOK_PERMISSION.
UNCONDITIONAL_METHODS was missing from envelope.ts, so the three methods the
router registers outside every grant check had no vocabulary of their own.
Adding it here, with a compile-time proof that it stays disjoint from
KNOWN_METHODS: giving those three a METHOD_PERMISSION row would change which
methods isAuditable considers, and settings.get is deliberately the one
unconditional method that is not audited.
The kit is written to stay extractable into its own package, so an ESLint rule
pins what it may import. Note the rule is a deny list on purpose: these patterns
are gitignore-style, and a leading ** would exclude the parent directory of
every allowed path, which makes the ! re-includes silently ineffective.
The hook->permission map existed verbatim twice (the host supervisor and the
plugin SDK) and the permission list four times (envelope, SDK manifest, and
both admin panels in the client). All of them were in sync when measured -
63 entries each, 15 hook entries each - so this changes no behaviour. It
removes a quiet outage: a hook whose permission is added on only one side
fails silently in production, with the plugin installing, activating and then
simply never being called.
protocol/envelope.ts is the single source now. METHOD_PERMISSION was typed
Record<KnownMethod, string>, so its value side was unconstrained -
'db:read:trps' compiled and made the method unreachable forever. It is
`as const satisfies Record<KnownMethod, KnownPermission>` now, which catches a
typo'd value, a missing key and an extra key. The derived KnownPermission and
MethodPermission<M> are what the @PluginMethod decorator will check its
argument against.
Deliberately no explicit annotation on METHOD_PERMISSION: with one, typeof is
that Record, MethodPermission<M> widens to the whole union, and the check
silently evaporates.
server/scripts/gen-plugin-facts.ts imports envelope.ts and writes the SDK and
shared copies. It does not scrape. The guard it replaces pulled the block out
of plugin-supervisor.ts with a regex that goes null on a reformat, and it ran
only from prepublishOnly - drift would have surfaced as a broken npm release.
The generated files are checked in because publish-plugin-sdk builds from a
standalone checkout with no server present.
The SDK's re-exported types stay wide. plugin-sdk compiles under strict, and
grantGaps indexes HOOK_PERMISSION with a plain string, so an `as const` there
is TS7053 and would break the publish. Strictness belongs on the host side.
--check also fails on a permission with no admin.plugins.perm.* translation
and on one the CLI has no hint for. Both are silent today.
plugin-sdk/** was in no workflow path filter, so nothing in CI ever ran its
tests. It is in the filter now, alongside a job that runs the drift check.
journeyService (1324 LOC, 36 exports) and journeyShareService (203) become
JourneyDomainService and JourneyShareService. Every SQL statement, error string
and broadcast is the one that shipped.
The two could never be split: the share service gates on isOwner, which lives in
the domain service, so it injects it.
JourneyDomainModule is a leaf carrying just the two services. Places,
assignments and the plugin host each need a handful of journey functions;
importing JourneyModule for that would drag MemoriesModule and both photo
providers into their graphs. Same shape as TrekPhotosModule.
The 23 tools in src/mcp/tools/journey.ts become JourneyMcp. src/mcp/tools/ is
down to _shared.ts and registerTools() no longer fans out to anything.
journey.bridge.ts keeps four exports, for the journey resources in
src/mcp/resources.ts that register at import time.
McpAccessMode is host-augmentable now. journey:share is neither read nor write,
so the three share tools would have had to be opaque predicates - and predicates
bypass the access policy entirely, invisible to the boot-time scope gate. A
scope typo in one would ship as a silently ungated tool. McpAccessModeRegistry
mirrors McpAccessGroupRegistry; unaugmented, mode stays 'read' | 'write'.
Fixes a latent bug in tests/helpers/mcp-test-controllers.ts: it constructed
PlacesService with five of eight arguments. tsconfig only includes src, so
nothing typechecked it, and the missing dependencies were harmless until the
journey skeleton hooks landed on the place write paths.
journey.mcp.ts arrived from src/mcp/ at 14.5% branch coverage - the legacy
registrar had no test file of its own. 40 tool tests take it to 93.4%, and 25
cases for the skeleton lifecycle, the photo gallery and the preference row take
the domain service from 73.9% to 82.8%.
Pressing "log out" clears isAuthenticated while the page you were on is
still mounted, so ProtectedRoute re-renders for it and sends you to
/login?redirect=<that page>. Log out from the dashboard and the next login
goes to the dashboard - beating the startup destination every time after
the first, which is exactly how it looked: worked once, never again.
A session that ended on its own still gets the return ticket, because
coming back to where you were is the whole point of that parameter. A
deliberate sign-out is a fresh start and now gets none, so the startup
destination decides. The Navbar already hinted at this with its unused
noRedirect state; the flag lives in the store instead, since all seven
call sites of logout() need it, not just that one.
RootRedirect waited for whoever else might call loadSettings, which on a
cold start meant waiting for a request queued behind everything else a
launch fires off. Traced it in the browser: the settings came back as the
388th request, long after the redirect had given up and gone to the
dashboard. That is what "it still just opens the dashboard" was.
It now asks for the settings itself - the store de-dupes, so this costs
nothing when a load is already in flight - and login spends its 2.6s
takeoff animation fetching them instead of leaving that time idle, so by
the time it hands over they are already there.
Also folds the five copies of setShowTakeoff + navigate-after-2600ms into
one takeOff(), since all of them needed the same fix.
The mirror answered "dashboard" for a browser that had simply never been
told, and RootRedirect took that as the user's choice - so the setting was
ignored on every device except the one that set it, and on that one too
after a logout cleared the mirror. The setting itself was stored correctly
the whole time, which is what made it look like nothing happened.
readStartDestination now says "unknown" rather than guessing, and a device
in that state waits for the real settings instead. That costs one wait,
once per device; afterwards the mirror answers and the launch is as fast
as before. Bounded by a timeout, because loadSettings deliberately leaves
isLoaded false on a failed request so it can retry - without the floor, a
launch with no backend would sit on the spinner forever.
Includes the tab ids the ?tab= parameter takes - they are the historic
German names, which is not something anyone would guess. That matters for
the people who asked: a shortcut or a browser wrapper can point at a tab
without the setting at all.
Desktop and mobile both get it, in the style of the file they live in -
the On/Off button pairs on desktop, segments and a picker sheet on mobile.
The tab picker only appears once "Active trip" is chosen, since there is
nothing for it to mean otherwise.
Labels come from the same keys the planner's tab bar uses, so the picker
can never name a tab differently from the tab itself.
Settings -> General -> Startup: start on the dashboard (unchanged default)
or straight in the active trip, on a tab of your choice. Three clicks and
three loads less for someone entering expenses while actually travelling,
which is where the request came from.
The redirect sits on '/' because that is what a shortcut, a bookmark and
the installed PWA all launch (manifest start_url). /dashboard keeps
working as itself - redirecting there too would leave no way back to it.
The preference is mirrored into localStorage, the way the language and
the theme snapshot already are, so the first paint can decide without
waiting on GET /api/settings. That keeps the default launch exactly as
fast as before: only someone who asked for the trip pays for the lookup,
and if it fails or they have no trip they get the dashboard. Right after
login the store already holds the real settings, so those win there.
Login stops hardcoding /dashboard as its fallback and hands the decision
to the same place, so there is one answer to "where does TREK open".
/trips/:id?tab=finanzplan opens the planner straight on that tab. Read in
the state initializer, not an effect, so the planner never paints the plan
view and swaps a frame later. The param is consumed on arrival; the
session memory keeps the choice across a reload.
Two things the tab machinery got away with while nothing ever started on a
tab other than plan:
- the guard evicted anything not in TRIP_TABS, but enabledAddons is an
optimistic guess until the addon feed answers, and collab is guessed off
- so a requested collab tab was thrown away a render before we knew the
addon was on. It now waits for the feed, like it already did for plugins.
- the lazy loads hang off handleTabChange, which the tab we *start* on
never goes through. Opening on Files showed an empty list, and that was
already true for a tab restored from a previous visit.
Tab ids move to constants/tripTabs so the settings picker can name them
the way the tab bar does instead of keeping a second list.
GET /api/trips/active answers with the trip a user means by "my trip"
right now: the one running today, else the next one starting, else the
most recently started. Archived trips never qualify.
That order is the dashboard hero's (sortTrips), on purpose - the startup
redirect this is for has to land on the same trip the dashboard features,
or the two would disagree about which trip is yours.
Kept apart from list() because it runs on a first paint: four columns of
one row instead of every trip with its per-trip day and place counts.
Completes the pass over the providers that moved into the gated tree. The
session handling is the part that most needed it: Synology invalidates a SID on
timeout (106), on a duplicate login (107) and when it does not recognise it
(119), and the service is supposed to clear it, re-login once and repeat the
call. That retry had no case at all — a regression there would have looked like
"Synology randomly stops working" and nothing would have failed.
Also pinned: a SID that no longer decrypts after a key rotation triggers a fresh
login rather than a dead session; a stored device id rides along so a trusted
device skips OTP; an app-level error that is not a session code is NOT retried;
the album listing merges personal, shared and shared-with-me, survives a partial
source failure, and collapses an id that appears in two sources; and an SSRF
block stays a 400 with the guard's own message while any other transport failure
is a 500.
The seven files in services/memories/ become injectable services. Every SQL
statement, error string, status code and header moves verbatim; what changes is
who owns what.
photoResolverService was two files in one, so it stayed two: the trek_photos
rows are a repository under photos/, the provider dispatch is
PhotoResolverService under memories/. That is the photos/ vs memories/ boundary —
storage there, dispatch here — and it is also why the module graph is a straight
line: TrekPhotosModule is a leaf both halves import, so neither needs a
forwardRef.
pipeAsset and the ServiceResult envelope stay free functions over the Express
response. pipeAsset is deliberately not a StreamableFile: it forwards the
upstream 206 and Content-Range so a <video> can seek (#823), swaps Cache-Control
to no-store on an error status, and answers differently once headers are sent.
The trek-photo cache keeps its in-flight map at module scope. That map is the
stampede guard, and the scheduler builds its own instance through a lazy require
outside the container — a per-instance map would give them separate guards and
ten tiles would fetch the same asset ten times, silently.
journeyService's dynamic import() of synologyService is gone: it existed only to
duck the import cycle, which no longer exists.
MemoriesModule imports RealtimeModule explicitly even though it is @Global, so a
single-domain e2e TestingModule can still resolve the broadcast.
The providers arrived at 60% branch coverage or less, because services/memories/
sits outside the gate and the integration suites drive the domain over HTTP
without reaching the failure paths. New suites cover what had no case at all:
the access checks that decide who may fetch someone else's photo, the stampede
guard, the thumbnail mtime check, the poster-less-video 404, and Immich end to
end with safeFetch stubbed — including the hidden-asset filter older servers
depend on (#1474) and the http -> https canonical detection that must NOT
rewrite a changed host.
PR #1844 went red with every suite green: the diff moved MCP tools from
src/mcp/, which the gate ignores, into src/nest/**, which it measures. That did
not change any behaviour - it revealed that airports.mcp.ts had never had a test
in either world, and dragged the aggregate to 79.86%.
Worth stating in the gotcha list, because journey.ts and _shared.ts still have
to make the same move.
Two preparations the provider conversion needs, and neither changes behaviour.
The ServiceResult envelope, the provider-agnostic asset shapes and pipeAsset
move to nest/memories/memories.helpers.ts, which touches no database. The
controllers and the provider services can name them without importing each
other - the same reason notifications/notification-events.ts exists. What is
left in services/memories/helpersService.ts is the DB-backed half (the access
checks, the album-link lookups) and it folds with the providers.
pipeAsset stays a free function over the Express response on purpose. It is not
a StreamableFile candidate: it forwards the upstream's 206 and Content-Range so
a <video> can seek (#823), swaps Cache-Control to no-store on an error status,
and has to answer differently once headers are already sent. All three
controllers take @Res() anyway.
The three copies of `path.join(__dirname, '../../../uploads')` become one
UPLOADS_ROOT. __dirname moves with the file and none of these would have thrown
on a wrong depth - mkdirSync and multer create the wrong tree without a word,
and every stored photo lands somewhere nobody serves from. UPLOADS-ROOT-001
pins it. The depth happens to be identical for nest/memories, so nothing moves
today; the point is that the next move only has one place to get wrong.
photoResolverService was two files in one. One half registers and drops rows in
trek_photos and touches no photo provider at all; the other dispatches a stream
or an info lookup to immich or synology. Keeping them together meant every
consumer that only wanted to store a row - the unified trip-photo service, the
journey service - dragged both providers in behind it.
The storage half becomes TrekPhotosRepository in nest/photos, over
DatabaseService. That is also where the photos/ vs memories/ line belongs:
storage here, provider dispatch with memories. photos.bridge.ts serves the two
modules still in services/ that register rows, and dies with them.
Every statement moved verbatim, including the two rules worth stating out loud:
getOrCreate re-encrypts a passphrase onto an existing row rather than inserting
a second one, and deleteIfOrphan never deletes a local photo, because its bytes
are ours and the file would outlive the row.
The coverage gate reads src/nest/** only, so the tools that lived in src/mcp/
were never measured. Moving them exposed what was never covered:
- airports.mcp.ts sat at 0% branches - search_airports and get_airport had no
test at all, in either world
- the two weather catch blocks, which turn a provider failure into an isError
result instead of a transport-level error
- the transport guards: trip access, day ids belonging to another trip, the
not-a-transport-type check on update, and delete_transport in full
- the price path, where a transport with a price links a budget item and
broadcasts budget:created
None of this is new behaviour; it is behaviour that shipped untested because it
sat outside the measured tree.
One case documents a subtlety worth keeping: the weather fallback message only
fires for a rejection with no .message at all. `new Error('')` still carries an
empty one and passes straight through, so the case rejects with a non-Error.
The prompts registrar is gone. Each prompt moves to the domain that owns its
data instead of a shared file that had to import three bridges to serve them.
packing-list goes to PackingMcp and budget-overview to BudgetMcp, both reusing
the `when` addon gate their tools already carry — so a disabled addon keeps the
prompt out of the listing, exactly as the `if (isAddonEnabled(...))` around the
registration call did. Both keep taking the trip title through trips.bridge:
TripsModule imports PackingModule and BudgetModule, so injecting TripsService
would close a cycle. Same documented in-container exception as atlas.mcp.ts, and
the registrar reached for it the same way.
token_auth_notice goes to a new AuthMcp — it is a statement about how the
session authenticated, not about trip data, and it injects nothing because the
message is static. Its `if (isStaticToken)` guard becomes `when: (ctx) =>
ctx.isStaticToken`; the registry hands `when` the session context, so an
OAuth 2.1 session never sees the prompt listed at all, which is what the
registrar achieved by skipping registration.
src/mcp/tools/ is down to _shared.ts and journey.ts. journey stays until
services/journeyService folds, which waits on memories.
Three of the four remaining hand-written registrars are gone. The tools keep
their names, descriptions, input schemas, annotations and result shapes; the
registration-time scope gates (`if (canWeather)`, `if (canGeo)`,
`if (!canWrite(scopes, 'reservations')) return`) become declarative access
markers, resolved by trekMcpAccessPolicy like every other decorated domain.
get_weather / get_detailed_weather move into a new WeatherMcp over the injected
WeatherService, including the catch that answers with the upstream message and
isError rather than throwing - a client sees "Weather service not available."
instead of a transport failure when the provider is down.
search_airports / get_airport move into a new AirportsMcp with no injected
service at all: the airport table is a static dataset. AirportsService owns the
flight-endpoint backfill against the database, which neither tool touches.
create/update/delete_transport move into ReservationsMcp rather than a domain of
their own, because a transport IS a reservation - same table, same service, same
reservation_edit permission - and that class already injects everything they
needed from the four bridges the registrar was importing.
That was budget.bridge's last outside-container consumer, so
linkBudgetItemToReservation is pruned with it; the file is down to one export and
dies as soon as BudgetMcp stops needing AuthService.
journey.ts stays: it still reaches into services/journeyService, which cannot
fold until memories does.
immich -> unified -> photoResolver -> immich was a real cycle, every edge a
value import; journeyService only got past it with a dynamic import(). Nothing
in that cluster can become a provider while it stands, because Nest would need a
forwardRef through three modules.
It had exactly one cause: syncAlbumAssets and syncSynologyAlbumLink called
addTripPhotos, which lives in unifiedService. Both functions are two things
stapled together - fetch the album from the provider, then write it to the trip
- and only the first half is provider-specific.
So the providers keep the half they own (collectAlbumSelection,
collectSynologyAlbumSelection) and unifiedService keeps the half it owns
(syncImmichAlbum, syncSynologyAlbum). Every error string, status code and return
shape is the one the combined function produced, including the per-provider
wording the split had to preserve ('Could not reach Immich' stays in immich).
The plan called for pulling addTripPhotos and getOrCreateTrekPhoto into one
neutral module instead. Together they do not work as a pair: such a module still
needs immich and synology to fetch, so the cycle only moves.
There is a second cut that works - photoResolverService is really two files, and
its trek_photos CRUD half touches no provider, so splitting that out also opens
both loops with less churn. It is not the one taken here because it leaves
immich -> unified standing, i.e. a provider depending on the trip aggregate.
That direction is what tied the knot in the first place. The CRUD split is still
worth doing, but as the photos/ vs memories/ module boundary, not as the cycle
fix.
server/src/services/memories/ is a DAG now: helpers <- immich/synology <-
photoResolver <- unified.
The last nine memories entries in the body-contract allow list are gone: three
each on the immich, synology and unified controllers. They read @Body() as
Record<string, unknown> and coerced by hand.
The schemas are deliberately permissive and loose, and that is the point rather
than a shortcut. Paging arrives as `2` or `'2'` because the controller does
`Number(page) || 1`; synology_skip_ssl arrives as a boolean or the string
'true'; auto_upload is accepted in any shape because only a real boolean is
acted on. A schema that described the ideal body instead of the real one would
turn working clients into 400s, and parity outranks tightening. Narrowing them
is a separate, client-visible change.
Loose objects for the same reason: a strict object strips unknown keys, and
`selections[]` goes straight to the services as `Selection[]` - stripping there
would be a silent data loss the day a provider adds a field.
This unblocks be-memories-fold, which could not move a controller whose body had
no contract to move with it.
Four files taken apart rather than moved, because they only worked as a knot.
The mailer becomes its own leaf module. AuthService sends the password-reset
mail while NotificationsModule imports AuthModule for the demo gate in
NotificationsMcp - a mailer inside the notifications domain would have closed a
hard cycle that only a forwardRef could open again. Webhook and ntfy stay with
notifications; auth has no business importing them.
The channel registry stays a module singleton and does NOT become a provider.
PluginRuntimeService pushes its channel getter in at onModuleInit, and
notifications.bridge.ts builds its own NotificationsService for the scheduler
crons and the memories senders. A provider would hand the bridge a second, empty
registry - every plugin channel would then go quiet with no error anywhere.
CHOVR-015 drives that seam end to end: a channel published through the runtime,
delivered through the bridge. Nothing pinned it before; the existing cases all
set the source themselves.
The built-ins no longer self-register by side-effect import. They used to exist
only because notificationPreferencesService happened to `import
'./notifications/builtins'`; removing that line would have silenced email,
webhook and ntfy without a failure. They are built from the injected transports
and registered from NotificationsService's constructor, so both instances have
them.
No cached nodemailer transport: MailerService.createTransport re-reads the SMTP
config and skipTlsVerify per send, exactly as the three inline call sites did,
so an admin changing them does not have to restart.
Preferences become a provider over DatabaseService. Eight places built
NotificationsService by hand, so a tests/helpers/notifications.ts factory keeps
the next constructor parameter a one-file diff instead of an eight-file one.
The event names, the channel ids and the ExternalChannel contract move to
nest/notifications/notification-events.ts, which imports nothing.
The cluster only looked acyclic: notifications.ts and channelRegistry.ts both
reached back into notificationPreferencesService for NotifEventType, and both
edges were `import type`, so they compiled away. Putting any of those names in a
value position - an enum, a const array, an injected class - would have made the
cycle real, with builtins.ts as a third ring. Nothing can fold while that is
true.
The drift guard against the shared NotificationEventKey moves with the union it
guards; it sat three files away from it.
* refactor(server): move the webauthn config and the account erasure into nest/auth
webauthnConfig resolved the RP ID through its own getSetting() on the db
singleton; it is a WebauthnConfigService over DatabaseService now, injected into
AuthService (passkey_configured) and PasskeyService (four ceremonies). readEnv()
is still read per call, so an operator flipping the RP ID at runtime takes effect
on the next ceremony instead of after a restart.
userCleanupService follows as UserCleanupService. It keeps taking
removeUserFromBudgetItems from budget.bridge rather than injecting BudgetService:
BudgetModule imports AuthModule for the demo guard in BudgetMcp, so AuthModule
cannot import BudgetModule back, and breaking that edge the other way drags
BudgetController, BudgetMcp and its trips.bridge chain into every graph that
imports AuthModule.
The legacy module had no tests of its own - the erasure only ever ran through
AdminService.deleteUser and TripsService.deleteGuest - so USER-CLEANUP-001..009
pin both halves, including the orphan data dir that no permissions row can
describe and the all-or-nothing deletion transaction.
* refactor(server): fold the oauth 2.1 server out of services/
694 lines move into OauthService over DatabaseService, AddonsService and
AuditService. The thin wrapper that only forwarded is gone, and so is the second
path to the MCP addon flag - validateAuthorizeRequest read it through
addons.bridge while the wrapper read it through the injected service, which would
have let the two drift.
The pending authorization codes stay a module singleton in oauth.pending-codes.ts
instead of becoming instance state. The SDK still serves /oauth/authorize,
/oauth/register and /oauth/consent from its own router, mounted on Express before
app.init(), so it reaches the domain through the new oauth.bridge.ts - two maps
would break the authorization-code flow silently, since the SDK only ever answers
"Authorization grant is invalid.". The import-time sweep keeps its .unref() and is
a named function now so a test can drive it.
scopes and sessionManager are imported deep, never through the src/mcp barrel:
the barrel re-exports revokeUserSessionsForClient and pulls the domain bridges
back in, which is the nest -> mcp -> nest cycle. Same comment as admin.service and
auth.service carry.
The 70 cases of the legacy suite moved over unchanged, bound to a hand-built
service. Fifteen more cover what free functions never let anyone reach: the code
store at capacity, an expired code, an expired access token, refresh against a
public and an unknown client, the RFC 8707 resource checks, and that a code
written through the container is read back through the bridge.
* refactor(server): move the wiki reader into the help module
wikiService had exactly one consumer, HelpController, and touches no database -
it stays a plain module rather than becoming a provider, the way
auth/ephemeral-tokens.ts does, so the text and asset caches keep being one per
process without any ceremony around it.
The WIKI_DIR anchor needed one more level: src/services sits three below the repo
root, src/nest/help sits four. The existing path-resolution case caught it, which
is what it is there for - a wrong anchor makes help degrade to the GitHub fallback
without anybody noticing.
Also brings the nest README back in line: services/ is down to five top-level
files, and the passkey and budget entries no longer point at paths that were
deleted since they were written.
Two small leftovers, moved to the domains that own them.
ephemeralTokens holds the short-lived tokens the WebSocket handshake and the
file-download links use. It already had startTokenCleanup/stopTokenCleanup with
an unref-ed handle — the right shape. What was missing is anyone calling stop:
index.ts kicked the sweep off with a bare require() after listen() and nothing
ever ended it. AuthService owns that lifecycle now, through
onModuleInit/onModuleDestroy, which is where it belongs — it is the service that
issues the tokens.
The store itself stays module state. websocket.ts consumes tokens from outside
the container, and a token store is only useful if the issuer and the consumer
share one. Same reasoning as the weather cache in #1835.
inAppNotificationActions is a 22-line handler registry with exactly one
consumer; it moves next to it as notifications/in-app-actions.ts.
services/ is down to 9 files, from 44.
Finishes be-weather-airports-fold. The README has listed weather as the
reference implementation since the migration started, while 513 lines of it sat
in the legacy layer.
The file is nest/weather/weather.impl.ts now, and WeatherService owns what it
could not own before: the cache sweep. That sweep was a bare setInterval at
module scope, so every process that so much as imported the file started a
five-minute timer and nothing ever stopped it — there was nothing to call. It is
startCacheCleanup/stopCacheCleanup now, driven by onModuleInit/onModuleDestroy,
and the handle is unref-ed so a cache sweep can never hold the process open.
The old test file says as much: its second line is "Prevent the module-level
setInterval from running during tests", with a vi.useFakeTimers() to do it. That
workaround is no longer what keeps the timer quiet.
The cache itself stays module state, deliberately. The MCP weather tools call
getWeather directly from outside the container, and a cache is only worth having
if every caller sees the same one. Hoisting it into the class would mean handing
the registrar the container instance to get exactly what a module singleton
already provides — the same trade the place-photo cache resolved the other way,
because there the scheduler needed the instance anyway.
Coverage is the reason this PR carries eleven new tests rather than none. Code
in services/ is ungated; the moment it lands in src/nest it counts against the
80% branch threshold, and weather.impl arrived at 75%. The gap was in paths
worth having: a failing archive lookup with and without an upstream reason, a
far-future date whose climate rows are all null, an expired entry being re-
fetched, and the sweep evicting one on its own. Plus the lifecycle itself —
start is idempotent, stop is idempotent, the handle is unref-ed. 79% now.
services/ is down to 11 files, from 44.
Two thirds of be-weather-airports-fold. weatherService (513 LOC, a process
cache and a setInterval at import) is left for its own PR.
airportService split along the line that was already there. The dataset lookup
touches no database — it reads one JSON file and keeps it in memory — and both
MCP registrars call it from outside the container, so findByIata/searchAirports
are free functions in nest/airports/airports.data.ts. backfillFlightEndpoints
needs the connection and became a method on AirportsService.
That backfill was the interesting part. db/database.ts ran it at module eval
through a bare require() back into services/, so the module graph closed a cycle
that only held because `db` happened to be initialised by the time the require
resolved. It runs in onApplicationBootstrap now: it repairs rows, nothing boots
on it, and the container is fully built by then.
The cycle had a visible cost. tests/setup.console-noise.ts carried a suppression
for `Cannot find module '../services/airportService'`, because vitest's
transform pipeline could not resolve that require at all — it fired in roughly
sixty test files. That suppression is now narrowed to the only case left: an e2e
TestingModule whose fixture DB has no reservations table.
distanceService split the same way: haversineKm is arithmetic and moved to
common/geo.ts; getFlightDistanceKm reads the database and became a private
method on AuthService, its only caller.
AirportsModule imports DatabaseModule explicitly. @Global only reaches modules
that are in the graph, and the airports e2e builds its container from
AirportsModule alone — exactly the trap the nest README now documents.
The search ranking gets its first direct tests. It sat in services/, outside the
coverage gate, and only ever ran incidentally through the route; moving it into
src/nest made the gap measurable (16% branches). The ten cases pin the scoring
ladder — exact IATA short-circuit, ICAO over prefix, IATA prefix over city
prefix, city prefix over substring — and the missing-dataset fallback, which no
test had reached. 90% branches now.
The last file in be-places-helpers-fold, and the one that had to be a provider
rather than a move: it carries three in-memory structures that only do their job
if the process has exactly one of them.
recentFailures stops a screenful of markers hammering an API that is down
inFlight stops a stampede when several requests miss the same placeId
knownOnDisk saves an existsSync per cache hit
A second instance gives the stampede guard its own inFlight map, so two
simultaneous requests for the same uncached place both call the provider. That
is why the scheduler takes this one from the container instead of constructing
it: SchedulerDeps gains a placePhotos entry alongside backups, filled from
app.get() in index.ts — the shape PR #1827 was built for.
The module-scope mkdirSync becomes onModuleInit, and GOOGLE_PHOTO_DIR is read
per call through RuntimeEnvService rather than frozen at import. That is not
cosmetic: TREK_PLACE_PHOTO_DIR is the seam the tests use to keep off the real
uploads tree, and its own test used to have to import the module dynamically to
set the variable in time. The __dirname fallback moved a level with the file and
is corrected accordingly.
Consumers: maps, places and share inject it; places.helpers.reclaimPhotoCache
takes it as a parameter, since it is a free function.
Six test seams follow. Four path mocks become injected stubs with the same mock
functions behind them; the collections case keeps the REAL implementation,
because what it asserts is that removeIfUnreferenced counts collection_places as
a reference (#1081) — a stub there would assert nothing. The share e2e overrides
the provider in its container.
services/ is down to 14 top-level files, from 44 when this started.
Two of the three files in be-places-helpers-fold. placePhotoCache is left for
its own PR — it holds three module-level caches and its sweep hangs off the
cron, so it needs both a provider with real state and a second entry in the
scheduler contract.
kmlImport.ts is 212 lines with no state, no database and no I/O — parsing,
sanitising and coordinate arithmetic. It is nest/places/kml-import.helpers.ts
now, unchanged, next to the places.helpers.ts that already consumes it.
placeImage.ts moves to nest/places/place-image.ts as free functions. Making
reclaimPlaceImage a provider would have split it from PLACE_IMAGES_DIR, which
cannot be injected: nest/common/place-image-upload.ts reads it at module scope
to configure multer's storage, before any container exists.
That constant is the one thing here that could have broken quietly:
path.resolve(__dirname, '../../uploads/places')
__dirname moves with the file. From src/services that resolved to
<server>/uploads/places; from src/nest/places the same string would have
resolved to <server>/src/uploads/places — a directory multer would have happily
created, leaving every uploaded place image in the wrong place and every
existing one unreachable. The path is now '../../../uploads/places', verified to
resolve to the identical absolute directory, and uploads-dirs.test.ts covers it.
Two of the three leftovers in be-misc-fold. The third (wikiService) needs a Zod
contract in @trek/shared first and is left for its own PR.
llmConfig.ts was pure movement: 70 lines of encryption-aware config shaping with
no state and no database, sitting in services/ while all four of its callers are
in src/nest. It is nest/llm-parse/llm-config.ts now, same functions, same file.
unsplashService.ts needed an actual provider. Its key resolution reads both the
env and the users table (env var, then the caller's stored key, then any
admin's), so the injected connection was unavoidable. Its own module because
trips and places both reach it, and because folding it into either would have
made the other import a trip/place service for an image search.
The four exports became methods on UnsplashService. The SSRF guard in
saveUnsplashCover is untouched — it still refuses any host but the Unsplash
image CDN and still goes through safeFetch — as is the unauthenticated-endpoint
fallback when no key resolves (#1449).
The env read stays live through RuntimeEnvService rather than a registerAs
token. readEnv() is what the function did before, and a boot-frozen token would
change when an operator's UNSPLASH_ACCESS_KEY takes effect.
Two test seams follow the code: the unsplash unit tests bind the four methods so
every existing assertion reads unchanged, and the trips controller's path mock of
the module becomes an injected stub.
services/ is down to 17 top-level files.
The README is the binding spec, and seven PRs have landed patterns it does not
mention. Someone reading it today would rebuild things that now exist, and would
walk into two traps that cost real time this week.
Added to the foundation section: common/ and why those helpers are free
functions, auth/jwt-verify.ts and why JWT_SECRET must not move to app-config,
query-helpers/, trip-membership/ and the cycle that put it in its own module,
AddonGuard + @RequireAddon with its ordering requirement, and demo-write.
New subsection on reaching the container from outside it. The choice is settled
now: hand the dependency in, declared structurally, the way index.ts fills the
scheduler's SchedulerDeps. Bridges stay correct for the MCP registrars, but the
README should say what they cost — a bridge rebuilds its services with `new`, so
one added constructor parameter meant editing five bridge files in a single PR.
Two gotchas that are cheap to read and expensive to rediscover:
Never gate a multipart upload with a guard. Guards run before the parser, so
throwing leaves the body unread and Node resets the connection — the client
gets ECONNRESET rather than the 403.
@Global only applies to modules that are in the graph. An e2e TestingModule
built around one domain module does not get AppConfigModule for free, and a
provider injecting RuntimeEnvService takes every suite importing that module
down with it.
Also corrects an assumption the migration plan makes and this README implied:
the trip-access and canEdit methods on the domain services are not dead weight
waiting for TripAccessGuard. Across the five domains piloted for it, 40 of the
46 callers are *.mcp.ts tools, which never pass through an HTTP guard. Only 6
sit in controllers. Counted, not estimated.
Two pieces of copy-paste in the guard layer.
Three addon guards — collections, airtrail, journey — did the same thing and
differed only in the addon id they read and the word leading the 404. They are
one AddonGuard plus a @RequireAddon(addonId, label) decorator now, read through
the Reflector. A handler-level decorator overrides the controller's, and a route
without the decorator is not gated at all.
The demo-mode upload block was copy-pasted into six handlers, each carrying its
own copy of the message. All six call isDemoWriteBlocked now, against the same
DEMO_WRITE_ERROR body.
That one is deliberately NOT a guard, though all six sites are handlers and a
guard is the obvious shape — and this PR did build one first. Guards run before
the multipart parser, so throwing from one leaves the request body unread and
Node resets the connection: the client gets ECONNRESET instead of the 403.
PROFILE-015 caught it. The check has to sit inside the handler, after the
interceptor has drained the upload, for the response to arrive at all. The
function carries that reasoning so the next person does not rediscover it.
Two smaller findings on the way there.
The DemoWriteGuard attempt broke eight e2e suites that never touch demo mode:
AuthModule listed it as a provider, and RuntimeEnvService was not resolvable
inside a TestingModule built around one domain module. AppConfigModule is
@Global, but @Global only applies to modules that are in the graph. The five
modules whose controllers now inject RuntimeEnvService import it explicitly, the
same way PermissionsModule is handled.
And both new guards get a test asserting the decorators are actually applied.
Once a check moves out of a handler, that handler's own tests stop noticing if
the decorator is dropped — the assertions on __guards__ and the @RequireAddon
metadata are what keeps the four upload endpoints and the four addon route
groups gated, and gated on the right addon.
middleware/auth.ts exported seven things. Two of them — extractToken and
verifyJwtAndLoadUser — are the canonical session check every auth surface goes
through. The other five (authenticate, requireCookieAuth, optionalAuth,
adminOnly, demoUploadBlock) are Express middlewares with no production caller
left; the guards in nest/auth replaced them and only comments still mention
them.
The two live functions move to nest/auth/jwt-verify.ts and the file goes, along
with middleware/validate.ts (nothing but its own test imported it) and
middleware/idempotency.ts (its only caller was `authenticate`, and
IdempotencyInterceptor has replaced it). AuthRequest and OptionalAuthRequest go
with them — src/typings/express.d.ts is what types req.user now.
Eight importers repoint: the three guards, auth/files/plugins-proxy services,
platform.routes and the global MFA policy.
Two deliberate departures from the plan for this package.
JWT_SECRET does not move to app-config. It is not an env value: it is generated
into data/.jwt_secret, and the admin panel rotates it at runtime through
updateJwtSecret. `export let` is precisely what makes a rotation take effect
in-process — a registerAs token would freeze the boot value and silently break
rotation.
The verify also keeps reading the db singleton rather than DatabaseService. That
change is not a one-line swap: it gives the three guards a constructor
dependency, which means registering them as providers everywhere they are used.
Measured: 38 directories use those guards and 21 of them do not import
AuthModule, on top of 34 e2e tests that each build a container around a single
domain module. That is its own PR, not a rider on one that deletes dead files.
tests/unit/middleware/auth.test.ts covered real behaviour, so it is not simply
deleted. Its extractToken cases and the five verify cases it drove through
`authenticate` come back as tests/unit/nest/auth/jwt-verify.test.ts, now hitting
verifyJwtAndLoadUser directly. Four cases are new: the password_version gate —
the whole reason this function exists — had no direct coverage at all. Its three
adminOnly cases are dropped because AdminGuard, their successor, is covered by
three e2e suites.
The scheduler runs outside the Nest container but needs four things inside it:
backupService, trekPhotoCache, placePhotoCache and airtrailSync. It reaches all
four with a require() or await import() at call time. This settles how that
should work before four separate folds each invent their own answer, and wires
one consumer as proof.
It takes the dependencies from the container rather than growing four more
bridges, for a reason that showed up while writing the previous PRs: adding one
constructor parameter to a service meant hand-editing five bridge files that
rebuild their services with `new`. app.get() has no such cost, and hands back
the container singleton instead of a second instance.
The scheduler declares what it needs structurally — a `backups` with a
createBackup — and never names BackupService. That is not fastidiousness: an
import of the class would pull in src/nest/backup → src/services/backupService →
src/scheduler, which is the exact cycle the lazy `await import()` in runBackup
existed to dodge. With the dependency handed in, the scheduler stops importing
backupService entirely and the cycle is gone; backupService still imports the
scheduler for VALID_INTERVALS, but now only in one direction.
index.ts fills the contract from the app buildApp() already initialised, right
before the first cron starts, the way bootstrap.ts hands the /mcp handler its
registry.
Two smaller things fall out. The Nest BackupService was swallowing
createBackup's prefix parameter, so it could only ever produce 'backup' names;
it forwards it now. And the scheduler can be started without a container at all
— its unit tests do exactly that — so runBackup says so in the log instead of
throwing, with a case covering it.
The other three consumers stay on their require() calls; they move with their
own folds.
Mechanical, and it ends the migration visibly.
45 files imported the whole react-dom default export to reach one function.
They now import createPortal by name, across 54 call sites; six files already
did it this way. main.tsx keeps its react-dom/client import, which is a
different module and already correct.
forwardRef is gone from every component. In React 19 ref is an ordinary prop,
so the wrapper only obscured the signature:
NumericInput — a plain DOM ref, no imperative handle involved
JourneyMap — useImperativeHandle stays, only the wrapping goes
JourneyMapGL — same
JourneyMapAuto — same
Each takes `ref?: Ref<Handle>` in its own Props interface now, so the ref is
part of the declared surface instead of arriving through a generic parameter.
The two journey bindings in glLazy shrink to one line each. They existed as
forwardRef wrappers purely to carry the ref past the lazy boundary; with ref an
ordinary prop it rides through the spread like everything else, which makes all
six bindings in that file identical again.
The five React.FC annotations go. Three were `Record<string, React.FC<{size?:
number}>>` for provider icons in files that never import React at all — those
only typechecked through the UMD global — and are now ComponentType. The two
pages become function declarations.
TranslationContext uses the 19 context shorthand, which leaves its React default
import unused, so that goes too.
Also drops two dead vi.mocks of react-dom/server. Their comment still said "used
in createMarkerIcon", which stopped being true when marker icons moved to the
in-house serializer — nothing under src/ imports react-dom/server any more
except the parity test that exists to compare against it.
The last React-18 leftover in the dependency tree. v4 declared
use-sync-external-store as a hard dependency and its main entry imported the
shim; v5 declares no dependencies at all and calls React.useSyncExternalStore
directly. Only zustand/traditional still reaches for the shim, and nothing here
imports it, so it no longer reaches the bundle.
The upgrade itself needed one line. What it needed more of was checking, because
the one thing v5 changes about behaviour does not show up as a test failure.
v4 subscribed through useSyncExternalStoreWithSelector, which memoised the
selector's result. v5 hands React a plain getSnapshot of
`() => selector(api.getState())` and compares with Object.is. A selector that
builds a fresh object or array therefore returns a new reference on every call,
and React reads that as "the store changed" — an infinite render loop, or at
best a re-render on every parent render.
All 294 store selectors return a property or a primitive. None builds an object
literal, array literal, filter/map/slice result, `?? []` default, Object.keys or
a new Set/Map, so none of them needed useShallow. The suite confirms it: 533
files, 11,165 tests, and zero "getSnapshot should be cached" or "Maximum update
depth" warnings anywhere in the output.
The one source change is the seedStore test helper, which was typed against a
hand-written `{ setState: (partial, replace?: boolean) => void }` shape. As of
v5 setState is an overload pair — partial with replace?: false, full state with
replace: true — and no single signature is assignable to that. Typing it against
zustand's own StoreApi fixes all 129 resulting errors and is what the helper
meant in the first place.
* refactor(server): route every trip-access check through DatabaseService
Twelve Nest services carry the same verifyTripAccess/canEdit pair. Six already
asked DatabaseService.canAccessTrip; the other six went through
services/tripAccess, a nine-line re-export of the identical function. Same
query, same result, two ways to reach it.
All twelve now take the injected route, so services/tripAccess is gone.
The bigger duplication was in the types. Thirteen files declared their own
`Trip` for the same return value, in three different shapes:
NonNullable<ReturnType<typeof verifyTripAccess>> 6x
{ user_id: number } 4x
TripAccess 3x
The narrow one is why assignments, days, places and accommodations all wrote
`as Trip | null | undefined` on the call — the cast existed to shrink
TripAccess down to a local type that was never meant to differ from it. All
thirteen now say TripAccess, and the four casts are gone with no other change.
middleware/tripAccess.ts goes too: it was already dead in production — nothing
under src/ imported requireTripAccess or requireTripOwner, only comments in the
services that had absorbed them. Its test went with it.
* refactor(server): fold the query helpers and trip joins into the container
Two more legacy files whose only reason for staying outside the container was
that they predated it.
services/queryHelpers.ts held five exports doing two different jobs. Three are
the batch loaders that keep the list endpoints off N+1 queries — they hit the
database and are now QueryHelpersService over the injected connection, imported
by assignments, days, places and share. The other two, formatAssignmentWithPlace
and ratingAggregate, touch no database at all; they sat next to the loaders only
because the loaders were free functions too. Those stay free functions, in
common/rowShape.ts.
The README note on the assignments service claimed the loaders had to stay in
services/ because the day, share and place services were unmigrated. All three
have since moved, so the reason is gone.
services/tripMembership.ts becomes TripMembershipService in its own module. It
could have folded into auth or trip-invite, but both call it — as do oidc and
the plugin RPC host — and putting it in either would have made AuthModule and
TripInviteModule import each other.
Neither module is @Global, matching PermissionsModule: e2e TestingModules
resolve them transitively through each consumer's explicit import.
Test seams changed shape accordingly. Two path mocks of services/tripMembership
became injected stubs; the trip-invite e2e overrides the provider in the
container instead. The oidc test keeps the real implementation, which is what
OIDC-SVC-045 was always asserting.
ratingAggregate moved out of the ungated services/ tree into the coverage-gated
one, so it gets the direct tests it never had — as does the tags/participants
pass-through, which two long-dead fixture variables in the moved test file had
been waiting for.
Also drops four stale comments that outlived their subject: the requireTripAccess
middleware they pointed at was deleted in the previous commit.
Eight files in services/ hold no state and touch no database: avatarUrl,
conflictResult, demo, passwordPolicy, timezoneService, cookie, and the two
crypto helpers. They are imported from 65 places, 62 of them already inside
src/nest, so having them live in the legacy layer was pure inertia.
They stay plain functions rather than becoming providers. Three importers sit
outside the container and would have no way to reach an injectable one:
db/migrations.ts runs before the app is built, and middleware/auth.ts and
middleware/mfaPolicy.ts run beside it. Wrapping these in DI would buy ceremony
and nothing else.
mfaCrypto and apiKeyCrypto move together into common/crypto — they read the
same ENCRYPTION_KEY, and the directory says so.
services/ drops from 44 files to 36, 9,870 lines to 8,493.
One trap worth recording: src/app-config and src/nest/app-config both exist.
cookie.ts imported '../app-config' from services/, which resolved to the former;
after the move the same string silently pointed at the latter, which has no
readEnv. It is now explicit about which one it wants.
Nothing else changes — no signatures, no behaviour, no tests rewritten beyond
their import paths.
Two more things every trip paid for whether or not anyone used them.
downloadTripPDF hung statically off the days sidebar, so opening any trip
pulled in the whole PDF builder. Both call sites are already async and already
wrap the call in try/catch, so a failed chunk surfaces as the export error the
handler had anyway.
TransportModal was already rendered conditionally, which means lazy bites
immediately. It is worth more than its own 32 kB: it is the only path to
TransitSearchPanel, and that chunk carried tz-lookup — about 200 kB of packed
zone geometry.
TripPlannerPage 339.78 -> 305.70 kB
entry chunk 258.24 -> 218.70 kB
TransitSearchPanel 198.67 -> 26.51 kB (and now on demand)
TripPDF 226.02 -> 27.78 kB (and now on demand)
eager path unchanged at 1,483 kB
Worth stating plainly: tz-lookup itself did not go away. It sits in
PlaceFormModal.helpers now, and PlaceFormModal is still a static import of the
planner, so a trip still loads it. Prying that loose means touching
placeOpenState, whose consumers assume a synchronous lookup — a separate change
with its own care.
* fix(map): drop the casing layer before its source on teardown
Tearing down the reservation overlay logged an error on every map unmount:
Source "trek-reservations" cannot be removed while layer
"trek-reservations-lines-transit-casing" is using it.
The casing layer arrived with the transit paths in 3.2.0 but was never added to
destroy(), which only removed the main line layer. The try/catch around it gave
no protection either — the engine reports this by firing an error event rather
than throwing — so the source was silently left behind on the style.
The layer id is a named constant now, so the place that creates it and the place
that removes it can no longer drift apart.
The lifecycle test grew the assertion that would have caught it: the fake map
now refuses to remove a source that a layer still references, the same way the
real one does. Verified against the unfixed code — two tests fail without the
change.
* fix(map): stop untrusted values reaching the marker HTML
Both map renderers build their marker HTML as a string and hand it to
L.divIcon / innerHTML. Two of the interpolations carry values a user controls,
and neither was escaped.
place.image_url is the worse of the two, because it is not admin-only: any trip
member with edit rights can set it. placeUpdateRequestSchema is a plain record
of unknown values, and the length guard in places.controller covers name,
description, address and notes but not image_url. The renderers do check the
value, but only with startsWith('/uploads/') — a payload like
/uploads/x" onerror="…" y="
satisfies that and then breaks out of the src="…" attribute.
category.color is the same shape one line up, reachable by an admin. It also
reaches SharedTripPage, which builds its own divIcon and answers without a
guard, so that payload lands in front of people who have no account here.
Colours are allow-listed rather than escaped. Escaping stops the attribute
breakout but leaves a working CSS value behind, and `url(https://…)` in a
background is a tracking pixel. Anything that is not #rgb/#rrggbb falls back.
image_url is a URL, not an enum, so that one is escaped; the prefix check stays.
Server side, the category contract now demands a hex colour, and the endpoints
validate through DTO classes instead of @Body('color') — a property read carries
no metatype, so the global pipe never saw it. That retires two entries from the
body-contract allow list. It replaces the hand-rolled 400 for a missing name
with the pipe's envelope, the same trade the places and trips migrations made.
The response schema deliberately still accepts any string: the write path was
open for a long time, so tightening it would make an instance fail on its own
stored rows. The client allow-list is what covers those.
Verified against the unpatched renderer — the three new cases fail without it.
* perf(client): serialize the map icons without react-dom/server
Six modules imported react-dom/server statically to turn a lucide icon into an
SVG string. That put Fizz — roughly 190 kB raw, 57 kB gzip — into a chunk three
lazy routes share, the Leaflet renderer among them, which is the default. For
output that is always a single <svg> with a flat list of shape children.
mapViewport chunk 189,532 -> 3,710 B
Tooltip chunk 194,333 -> 1,970 + 4,860 B
precache total 17,856 -> 17,675 KiB
utils/iconMarkup.ts serializes React elements directly. It is generic over
elements rather than over lucide's icon tables on purpose: iconNode is closed
over inside createLucideIcon and never exported, so calling the forwardRef
render function is what survives a lucide upgrade.
Byte-compatibility is checked, not asserted. iconMarkup.parity.test.ts diffs the
output against real Fizz for every exported icon in three prop shapes — the test
file is the one place still allowed to import react-dom/server, and it never
reaches a bundle. Two traps it covers: viewBox must not be kebab-cased into
view-box, and Fizz emits <path …></path> rather than a self-closing tag.
TripPDF loses its lazy renderer entirely — ensureRenderer, the module-level
handle and the two "not ready yet" guards that returned an empty string. There
is nothing left to load.
An eslint rule keeps the import from creeping back. Without it the regression is
invisible: the build stays green, the app keeps working, and the chunk quietly
grows by 190 kB again.
* perf(map): load one GL engine instead of both
MapViewGL, JourneyMapGL and the settings preview each imported mapbox-gl and
maplibre-gl statically, so rollup emitted one 2.83 MB chunk carrying both.
Every map user downloaded 1.8 MB of mapbox plus 1.03 MB of maplibre and then
ran exactly one of them. The 110 kB stylesheet was merged the same way.
mapbox users 2,827,647 -> 1,797,952 B (-1.03 MB, CSS 110.6 -> 40.8 kB)
maplibre users 2,827,647 -> 1,027,635 B (-1.80 MB, CSS 110.6 -> 69.8 kB)
The engine is a prop now rather than a module import, and the pairing happens
behind the lazy boundary in glLazy.tsx. No module reaches both SDKs statically
any more, which is the whole mechanism — the components themselves stayed
engine-agnostic, they already guarded on the provider everywhere the two
genuinely differ.
Both imports of a pair start together instead of in sequence: the engine is by
far the larger half, and fetching the component first would put its round trip
in front of the download that actually costs time.
check:gl-split guards it. A stray static import anywhere in the map tree brings
the merged chunk straight back with a green build and a working app, so the
script greps the built chunks for a marker of each SDK and fails when one file
holds both. Not on the bare string "maplibre" — the provider value and the
maplibre_style setting key appear in app code too, so that matches the shared
core chunk as well.
Two boundaries were missing and are added here, because the split makes them
reachable: the GL boundaries had no resetKeys, so a chunk failure under one
provider kept showing Leaflet after switching to the other, and the settings
preview had a Suspense but no boundary at all — that is the one place a user
flips providers live, so it is the likeliest chunk to fail.
Worth being precise about the gain: the service worker precaches every chunk,
so an installed PWA still downloads both engines. What this buys is the
critical path on a cold cache, and parse/compile/execute always — a megabyte
of unused engine ran on every map open.
* perf(client): split out the video player and the ISO subdivision data
Two heavyweights that shipped with routes that rarely use them.
plyr (112 kB raw / 33 kB gzip, plus 32 kB of CSS) hung off three lightboxes and
through them off the FileManager, JournalBody and MTripShell chunks. The
lightboxes only render the player on their video branch anyway, so the split
costs one frame of placeholder — a black rectangle in the target geometry, so
the lightbox does not jump when the chunk lands.
iso-3166-2 is a single 238 kB data blob with no tree-shaking to be had, and
only fetchRegionOptions touches it. It now loads after the hasRegions check,
so it is fetched only for countries that get a region picker at all:
holidayRegions drops from 257 kB to 14 kB. The import sits inside the existing
try — a failed chunk yields the same empty array as a failed request, rather
than an unhandled rejection in the two callers that only do .then(setRegions).
tz-lookup is deliberately left alone. TransitSearchPanel builds the departure
time it sends to MOTIS out of tzAt(), so a not-yet-loaded window there is a
wrong query rather than a late pixel — that one needs its own change with its
own care.
Everything that changes with a release currently drags the entry chunk with
it, and 63 of 141 chunks import the entry statically. Measured with a
one-token edit to a leaf route module and a rebuild, 70 of 141 chunks rehash —
just over 1 MB gzip that every installed PWA re-downloads per deploy, since
the service worker registers with autoUpdate.
Pulling react and the core libraries out cuts that to 847 kB:
rehashed 70/141 -> 60/140
gzip 1,038,497 B -> 847,688 B (-18.4%)
entry 580,672 B -> 258,105 B
eager JS 1,482,722 B -> 1,482,523 B (unchanged)
eager CSS same file, same hash
No first paint win, then — the gain is a smaller update delta and a halved
entry to parse on weak devices.
Three of the five groups the plan called for are left out, and not silently:
mapbox-gl/maplibre-gl, leaflet and react-markdown already sit in async chunks
whose hashes survive a release, so a group would only rename them. Worse,
without tags: ['$initial'] a group also collects modules that hang behind
React.lazy today and turns its chunk into a static import of the entry —
measured, that put the 2.8 MB GL chunk into the index.html modulepreload and
took eager JS to 4.61 MB, undoing the route split for maps.
Note for the next person to touch this: vite 8 bundles with rolldown, so
rollupOptions is only an alias and both manualChunks and advancedChunks are
deprecated in favour of codeSplitting.groups. A config using the old names
builds green and has no effect.
Also refreshes two comments that still quoted pre-split bundle numbers.
Every page imported its mobile screen statically and chose while rendering, so
each route chunk carried both trees and the viewport only decided which half
stayed dark. Worse, because ten route chunks shared the mobile tree, rollup
hoisted it into the entry chunk — so every desktop user downloaded the whole
mobile shell before seeing a single page.
The branch now decides the chunk. A phone never fetches the desktop planner, a
desktop never the mobile shell.
entry chunk 1,243 -> 581 kB (gzip 349 -> 177)
TripPlannerPage 914 -> 340 kB
VacayPage 361 -> 57 kB
AdminPage 413 -> 225 kB
SettingsPage 259 -> 155 kB
Splitting inside each page instead was measured and rejected: it leaves the
desktop code in the same module as the switch, so a phone would still download
it and then wait for a second, serial request on top — 914 kB and a waterfall
against 550 kB in parallel.
The pass-through wrapper in each page stays deliberately. check-page-pattern
scans from the default export to the next top-level function and rejects hooks
there; making the desktop component the default export fails it immediately on
dashboard, planner and atlas.
RouteFallback gets a mobile face. It renders above MobileShell, so the --m-*
tokens do not resolve for it and the desktop spinner on bg-surface would be a
foreign white sheet in front of the mobile screen.
The five scattered viewport spot checks are replaced by one contract covering
all eleven routes, which is where the switch now lives.
* refactor(costs): move the pure split maths out of the panel module
splitEqualShares, calculateTicketShares and TicketItem sit next to the payer
maths they belong with. They are pure functions with no module state and no
imports from the panel, so this is a move and nothing else.
The reason it matters: the mobile cost sheet and its model import these two
functions from CostsPanel, and MTripShell is a static import of the trip
planner page. Any attempt to load the costs panel on demand would therefore
pull the whole 97 kB module straight back in through the mobile branch.
* perf(planner): load the tab panels on demand
Each tab panel only mounts while its own tab is active, yet all six shipped in
the planner chunk. The page was 1,203 kB after route splitting — larger than
the entire entry chunk. It is now 914 kB, with the panels in six chunks of
their own (files 99, collab 87, packing 71, costs 62, reservations 32,
todo 31 kB).
The panels also get the error boundaries they never had. fe-error-boundary
covered the maps and the plugin frame but not these, so without them a dead
panel chunk would throw past to the route boundary and take the whole planner
with it. The boundary sits outside the Suspense, since Suspense only owns the
pending promise and a rejected one throws straight past it.
One boundary per panel rather than one around the content area: a shared one
would already be mounted with the visible tab, so switching tabs would swap
the entire planner for the placeholder instead of just the part still loading.
PluginFrame stays static — DayDetailPanel and PlaceInspector import it from
the plan tab anyway, so splitting it here would move nothing. The two import
modals stay static for the same reason, via the mobile sheets.
The wiring tests now await their panel. That includes the ones still passing:
React.lazy caches the resolved module for the whole file, so they were only
green because an earlier test had already pulled it in.
* chore(client): tighten the precache ceiling now the entry chunk is split
The 6 MB ceiling was set against the pre-split bundle, where the entry chunk
alone was 5.2 MB. It is now 1.2 MB and the largest precached entry is the
heic-to chunk at 3.0 MB, so the old limit had nothing left to catch.
Worth being explicit about what the limit does: a file above it is dropped
from the manifest and the build still succeeds, printing only that the file
will not be precached. The PWA is then quietly broken offline for that file,
which is why the ceiling has to stay close to the real bundle.
Also records why every route chunk is precached rather than pulled in at
runtime, since that is the reason route splitting did not shrink the install.
* docs(wiki): correct the offline cache table
The table still described an api-data cache holding API responses for 24
hours under NetworkFirst. That cache is gone: API requests are NetworkOnly
because Workbox keys entries by URL and cannot vary them on the session
cookie, so on a shared device one account's data could be served to the next.
Documenting a cache that deliberately does not exist is the wrong way round
for a privacy decision.
Map tiles were listed at 1 000 entries; the actual cap is 12 288 and has to
stay in step with MAX_TILES in tilePrefetcher.ts. The two GL tile caches were
missing altogether, and the precache line now says that it covers every page
rather than just the shell.
* feat(client): retry a failed page chunk once before giving up
A dynamic import fails for two different reasons, and only one of them is
worth a second attempt. If the file is gone because a deploy happened while
the tab was open, nothing but a reload helps and reloadOnceForChunk already
owns that path. If the fetch itself went wrong — hotel wifi, an aborted
request, a service worker handing back a truncated entry — one retry fixes
it.
The retry cannot reuse the original specifier, since Vite rewrites it into a
hashed asset URL at build time; the only usable URL is the one the browser
names in the failure message, and a fresh query on it misses both the HTTP
cache and the Workbox precache.
This is also where a healthy load is acknowledged. Clearing the reload marker
used to sit in main.tsx right after render(), which runs before React commits
and long before any chunk is fetched, so the marker was gone before it was
ever needed.
* perf(client): load the page chunks on demand
Every one of the twenty pages was a static import, so the entry chunk carried
the planner, the journal, the atlas and the vacation planner for anyone who
only opened the dashboard. They now load through lazyWithRetry; the entry
chunk drops from 5,183 kB to 1,243 kB (gzip 1,300 -> 349).
LoginPage stays static on purpose: anyone logged out lands there, and every
other route redirects there first.
The Suspense boundary sits above <Routes> rather than inside ProtectedRoute
so it stays mounted across navigations. react-router runs location updates
inside a transition, which keeps the current page on screen while the next
chunk arrives instead of flashing a spinner on every jump.
The precache is unchanged by design: workbox picks up the new chunks, so the
install size stays where it was (107 -> 178 entries, +31 KiB) and every route
remains available offline. This buys first paint, not install size.
* feat(client): catch render errors instead of blanking the page
There were no error boundaries anywhere, so one throw during render unmounted the
whole tree and left the user staring at a white page with no way back.
ErrorBoundary holds the state in a class — getDerivedStateFromError has no hook
equivalent — and renders a function component for the visible half, so the panel
can translate normally. TranslationContext isn't exported, so contextType was not
an option anyway.
Two things it does beyond showing a message. A chunk that 404s after a deploy gets
one reload rather than a retry button: React.lazy caches the rejected promise, so
retrying re-throws immediately. And the reload happens once per session, or a
genuinely missing chunk would spin the tab.
The global handlers in main.tsx cover what no boundary reaches — event handlers
and async code. They filter axios errors: the interceptor already redirects on
401/403 and there are 600+ toast.error call sites, so reporting again would mean
double toasts and, for the redirects, loops.
* feat(client): isolate routes, widgets and the shell chrome
The route boundary goes on the single line where all 16 protected routes render
their content, inside MobileShell rather than around it: the navigation survives,
so a broken page is something the user can walk away from instead of a dead end.
The mobile --m-* tokens live on the shell's .m-root too, so a fallback outside it
would render invisible.
key={location.pathname}, not resetKeys — ProtectedRoute is the same component at
the same position for every protected route, so otherwise a failure could follow
the user across a navigation.
The public routes render outside ProtectedRoute and were therefore uncovered,
including /login (where people land when everything else failed) and the two
anonymous share pages. PublicRoute wraps those seven.
The five global widgets and the shell chrome get fallback={null} each: a broken
widget should disappear quietly, and if BottomNav dies the user loses the very
thing they need to leave the route.
* feat(client): isolate the GL maps and the plugin frame
Both map wrappers put the boundary outside their Suspense, not inside: Suspense
handles a pending promise, a rejected one throws straight past it. Falling back to
the Leaflet renderer keeps a usable map rather than showing an error card — the
same downgrade the Suspense fallback already does while loading.
PluginFrame is third-party code rendering in our tree; a throw there should cost
the plugin, not the page. resetKeys={[pluginId]} so switching plugins clears it.
The locale loader had no .catch (TranslationContext.tsx). After a deploy the chunk
can be gone, which produced an unhandled rejection and left the UI silently on the
previous language. It now keeps the strings it has and says so in the console.
* refactor(client): opt into the v7 router behaviour while still on v6
Turning the behaviour on separately from the version bump keeps the two failure
modes apart: red here means a behaviour change, red on the bump means the import
rewrite. The suite is unchanged either way — 527 files, 11099 tests.
Only the two flags a non-data router understands are set; the rest belong to
createBrowserRouter, which this app doesn't use. Both are applied at all six
router mounts, so the tests exercise the same behaviour as production rather than
validating v6 while prod runs v7.
* fix(client): move to react-router 7.18.2
6.30.4 is affected by GHSA-wrjc-x8rr-h8h6 — open redirect through a backslash in
Link and useNavigate, which this app uses in 41 places — and the 6.x line has no
fix: react-router-dom 6.30.2 to 6.30.4 carries GHSA-jjmj-jmhj-qwj2 with no patched
version at all. Staying is not an option, and 7.0 to 7.17 only swaps it for
GHSA-chx6-hx7r-mcp5.
The package rename comes with it: react-router-dom is a re-export shim in v7, so
using it would only defer the same rewrite to the v8 jump. 77 occurrences across
60 files, all of them import specifiers and vi.mock strings.
The surface is small — useNavigate, useParams, useSearchParams, useLocation,
Routes, Route, Link, Navigate, useMatch, MemoryRouter, BrowserRouter — no data
router, no loaders or actions. Nothing else had to change; the future flags from
the previous commit are simply how v7 behaves, so they come back out.
npm audit still reports GHSA-qwww-vcr4-c8h2 on react-router (RSC mode CSRF, fixed
in 8.3.0). It needs RSC mode, which needs a server; this is a static SPA. Closing
it means the v8 major, which is its own piece of work.
* chore(nest): register LlmParseModule in AppModule
The module was only in the container because BookingImportModule and
PluginsModule happen to import it, so /api/admin/llm/local depended on those two
keeping that import. Register it where every other module is registered.
* refactor(nest): give the rate limiter one home and one instance
RateLimitService keeps its counters in a Map on the instance, but it was listed
in the providers of auth, oauth, transit and trip-invite, so each of them got its
own. Nothing is broken today — the bucket names in use (login/mfa/forgot/reset,
oauth_*, transit_*, trip_invite) don't overlap — but two modules picking the same
name would have quietly allowed twice the configured attempts.
Move it to nest/common, where it belongs now that four domains use it, and hand
it out through a small module they import. Not @Global: the e2e suites build a
container around one domain module, so a global AppModule never pulled in isn't
there at all.
transit.mcp.ts keeps its own instance on purpose — it keys by userId under
mcp_transit_* while the REST path keys by IP.
* test(ws): pin the wire envelope before touching the transport
Every frame on the socket is flat: {type, tripId, ...payload} out, {type, tripId}
in. Nest's WsAdapter dispatches on message.event and answers {event, data}, so
moving this onto a gateway without a custom serialiser would leave join silently
unhandled — no welcome.socketId to send back as X-Socket-Id, and every client
would then receive its own writes.
Nothing catches that today: the 100+ tests that touch realtime mock src/websocket
and never look at the bytes. WS-032 sends the exact shape the stock adapter would
and asserts the server ignores it.
* docs(nest): drop the parts of the README that no longer exist
strangler.ts, DEFAULT_NEST_PREFIXES/NEST_PREFIXES and tests/parity/ are all gone —
every request goes through Nest now — but the README still walks a new contributor
through adding a prefix and writing a parity test against the Express route.
Also spell out that the 80% gate is an average over src/nest/**, not a floor per
file, so a large untested module can sit behind the well-covered small ones.
* refactor(settings): load the GL map preview on demand
MapViewGL and JourneyMapGL are both behind lazy() so a Leaflet-only install
never downloads the GL engines. MapboxPreview imported mapbox-gl and maplibre-gl
statically, which put them back in the entry chunk and cancelled that out.
Wrap it in lazy()/Suspense on both the desktop tab and the mobile screen. The
entry chunk drops from 7987 kB to 5157 kB (2057 kB to 1292 kB gzipped); the
engines now ship as their own chunk, loaded when a GL provider is picked.
The preview no longer resolves synchronously, so the tests that read it right
after switching provider await it now.
* chore(client): drop unused deps, self-host MuseoModerno, add a bundle baseline
@react-pdf/renderer, react-window and @types/react-window have no importers
anywhere: TripPDF builds HTML for the browser print window and JourneyBookPDF
goes through marked. RegisterPage is dead too, /register renders LoginPage.
MuseoModerno was the last font still coming from the Google CDN through a
render-blocking <link>, while main.tsx already self-hosts Poppins and Geist for
exactly the reason stated in the comment there. Same weights, now bundled.
Add rollup-plugin-visualizer behind `npm run build:analyze` so the splitting work
has numbers to argue with, pin build.target instead of inheriting Vite's default,
and bring the workbox size ceiling down from 10 MB to 6 MB — anything above it is
dropped from the precache manifest silently, which only shows up as a broken
offline app.
PlaceInspector returns null when no place is selected, but the useCallback
for handleFileUpload sat below that return. Clearing the selection therefore
changed the hook count and React threw "Rendered fewer hooks than expected",
which takes the whole tree down since nothing catches it.
Move the callback above the bail-out and read the id through place?.id.
The FourZero show was built as a one-off for the 4.0.0 release and was
always meant to come back out afterwards. It lived in its own folder
plus two marked lines on the dashboard, so removing it is exactly that:
the folder and the two lines.
The beacon carried its own cut-off date of 23 August, so nothing about
this changes what users see between now and then beyond taking it out
early.
The new-trip forms hard-coded EUR and ignored settings.default_currency,
so creating a trip always opened on EUR even with USD set in
Settings > General > Currency. Read default_currency (falling back to
EUR when unset) in both the desktop TripFormModal and the mobile
MNewTripSheet; edit mode still pre-fills the trip's own currency.
Fixes#1784
- reverse() now receives the UI language
- error assertions follow the translated journey.editor.* strings
instead of the raw error.message and the generic common.error label
- a late error callback after a resolved one-shot fix is asserted to be
a no-op, since a settled promise cannot reject anymore
- add tests for the new location button: reverse-geocoded fill and the
translated inline denial message
- MJourneyEntrySheet renders the same crosshair button in its location
field, with the coordinate fallback and reverse-geocode race guard
- Quick-capture auto-locate now goes through getCurrentPositionOnce(),
replacing the raw navigator.geolocation call and the untranslated
error.message it surfaced
- Error code to i18n key mapping extracted to geoOnceErrorKey(), shared
by both editors
- Test for the race guard: a search result picked while the reverse
geocode is in flight must survive
- One-shot getCurrentPositionOnce() helper in useGeolocation with typed
error codes (unsupported, insecure context, permission denied, timeout)
- Button inside the entry editor Location field fills coordinates from the
device and reverse-geocodes the place name via /api/maps/reverse,
falling back to formatted coordinates when no name is available
- New journey.editor.* i18n keys in all 23 locales
- Unit tests for the helper and component tests for the button flow
Four tones cannot keep a twenty-stop trip's legs apart, which is most of
the point of colour-coding one. A contribution can now carry `#rrggbb`
alongside `tone`, per region or as a shorthand.
The split of control is: the plugin picks the hue, the host picks the
weight. A colour rides the same per-theme, per-region alphas as a tone,
and its OKLCH lightness is clamped into a per-theme band first, so no
contribution can hand a user a day card they can't read. The band
brackets the built-in tones rather than tightening around them, so an
ordinary colour passes through untouched and a hex equal to a tone
renders exactly like that tone. Hue and chroma are never touched - it
bounds a colour, it doesn't boost one, so a near-white grey still lands
as a faint tint.
Only `#rrggbb` survives normalization, and that is a security boundary
rather than tidying: the value ends up inside a `color-mix()` in an
inline `background`, a shorthand that takes a layer list, so a string
closing the paren early could append a `url(...)` and beacon everyone
viewing the trip.
A region gets exactly one paint instruction - what it names beats the
shorthand, a colour beats a tone at the same level - so the client never
has a tie to break. A region naming a colour the host can't use degrades
to `default` like a bogus tone already did.
Lets a plugin colour-code days in the Plan sidebar. trip-segments can
already announce "Kanazawa begins" in the day plan, but it can't make
day 12 look like it belongs to Kanazawa.
dayScheduleProvider can't cover this: it renders a fixed row, and it's
capped at 60 items per provider, so one item per day would tint the
first 60 days of a six-month trip and silently stop. This hook is
bounded by day count instead.
GET /api/day-tints/:tripId returns per-day tones for three regions of a
day card - the number badge (also the mobile day chip), the header row,
and the expanded activity list. `tone` is a shorthand filling every
region a contribution doesn't name; unnamed regions render as they do
today. Tones only, never hex, so the host can pick the alpha per theme
and per region.
One contribution per day, resolved whole: first entry wins within a
provider, first granted provider wins across plugins.
Newer OpenAI models (gpt-5.x) reject max_tokens with a 400 and demand
max_completion_tokens, so AI booking import failed and no reservations
were created. Send max_tokens first (understood by the classic chat API
and every local server), and only on a 400 naming max_completion_tokens
resend the whole request with that parameter instead — mirroring the
existing json_schema->json_object fallback beside it. Detection is by
API response, not a model-name allowlist.
Fixes#1760
Add a corner button on the Leaflet trip map that switches the base layer
between the default street tiles and a tokenless satellite view (ESRI
World Imagery). The choice is a per-user setting (map_base_layer) that
persists and applies instantly, including offline. GL-renderer satellite
styles are unchanged.
Closes#993
The web expense modal only sent the payer when at least one split
participant was selected, so a personal expense (e.g. a flight paid by
one person, not shared) was stored with an empty payers[] and lost its
payment attribution. Drop the participants.size guard so a picked payer
always goes out, matching MCostSheet. "Nobody paid" stays expressible via
the explicit "No one paid yet" option (payerId 0); the #1286 test is
updated to use it.
Fixes#1766
The AI Parsing admin panel kept the Base URL in state after switching provider
and saved it unconditionally, so moving a configured Local/OpenAI provider to
Anthropic left the old host in the stored config and misrouted every Anthropic
request away from api.anthropic.com. Clear baseUrl on save when the provider is
Anthropic, matching the per-user LlmConnectionSection behaviour.
Fixes#1748
An item shared through the sharing tier is visible to its recipients but
carried by its owner. The bag totals filtered on bag_id alone, so every
recipient saw the owner's weight added to their own load — which is exactly
the situation that tier exists for (one person brings the first-aid kit).
Filtering by owner_id, as the issue suggests, would have broken more than it
fixed: owner_id is stamped on every item including Common ones, so the group
pool would have shrunk to "only what I entered myself" even on trips where
nobody shares anything. The predicate keeps the Common pool intact and drops
only private items owned by someone else.
Closes#1767
The column, the contract, the API route and its tests have all been there
since v2.9.0 — there was simply no field to enter a limit in, so the only
way to record one was to put it in the bag name, where nothing can act on
it. Both the desktop bag card and the mobile bag row now take a limit in
kilograms (how airlines state them) and store it in grams. Emptying the
field clears the limit instead of writing a zero.
While adding it: the fill-bar formula existed in three places, and the one
in the bag modal had been copied without the limit, so a limit set on a
window narrower than xl did nothing. It now lives once in
packingListPanel.helpers as bagFillPct, next to itemWeight.
createItem never accepted weight_grams or bag_id — not in its signature, not
in the INSERT — so cloneItem had nothing to pass on and every copy arrived
empty. For a group trip that meant re-entering every weight by hand for each
traveller, which is what people were doing.
The bag needs more care than the weight. Bags belong to a trip, but they can
be assigned to a person, so inheriting the original bag would drop your copy
into somebody else luggage and inflate their total. A copy keeps the bag only
when nobody owns it or when the copier is one of its members.
The fire-and-forget notification sends guard the send() call but not the
dynamic import around it, so a rejected import escapes as an unhandled
rejection. In CI that surfaced as an EnvironmentTeardownError from vacay
after the suite had already passed, failing the run with every test green.
trips, packing and collections already catch on the import chain; collab and
the two vacay sends were missing it.
The mobile screen shares useAtlas, so it inherits the switch state; it gets
its own compact toggle next to the bucket-list button. Search now labels a
planned country as such instead of calling it visited, the stats card hangs
the planned count off the country total, and the suggestion list keeps
places you have actually been at the top.
Planned countries are off the map by default and come back through a switch
above the globe, drawn dashed so they never read as somewhere you have been.
The switch only appears once there is something planned to show, and it
remembers its state.
Three things that needed care:
The colour palette is built from the visited countries alone. It assigns
colours by index, so letting planned ones join would reshuffle the whole map
every time the switch is flipped.
countryMap decides both the tooltip and what a click does, so it is built
from the visible countries — a hidden country has to fall through to the
mark-as-visited branch or it becomes unclickable.
The region layer read `data` without listing it as a dependency, which meant
region colours already went stale on their own; adding the switch would have
made that visible.
Five copies of the optimistic mark-as-visited update have moved into
withCountryMarkedVisited, which also promotes a planned country instead of
adding it twice. Antarctica joins the continent row when someone has been.
getTravelStats derives its countries independently of the atlas service, so
without the same date filter the dashboard would have kept counting trips
the map no longer treats as visited — the drift #1490 closed.
Manually marked countries and the hidden-country tombstones are untouched.
Every country now carries the status of the strongest source it came from:
a country reached by both a past and a future trip is visited, one that only
appears in upcoming trips is planned. Marking a country by hand still wins
over everything, since that is a statement of fact rather than a date.
totalCountries drops to the visited ones and totalCountriesPlanned sits
beside it; continents and mostVisited follow the same rule. countries[]
stays a superset of what it returned before, so an older client keeps
painting the map exactly as it did.
The endpoint query has to select trip_id to know which trip a leg belongs
to, which makes its DISTINCT per trip — the coordinates are collapsed again
in JS so the point-in-polygon lookup still runs once per location.
The atlas counted every trip as a visit, so a country you had merely booked
a flight to was stamped the same as one you came back from.
tripVisitStatus buckets a trip by its dates: already started (or running)
counts as visited, starting in the future is planned, and a trip with no
dates at all is an idea that must not reach the visit stats. The comparison
is UTC, matching the dashboard and journey lifecycles.
Requested in discussion #1048.
Records the fold and corrects two long-carried errors: there is no admin MCP
surface (the '11 MCP consumers' figure predated the Phase-0 addons extraction)
and systemNotices/conditions.ts was never a consumer. The memories/admin edge is
documented in the right direction for the first time. journeyService and
oauthService now head the frontier.
The 851-line services/adminService folds into the wrapper AdminService over
DatabaseService plus injected Settings/Addons/Passkey/Packing/Auth/Permissions/
Notifications services; the auth, notifications and permissions bridge imports
all become injections, PERMISSION_ACTIONS stays a plain const, and the
mcp/sessionManager deep import keeps its anti-cycle comment. Ahead of the fold
the 11 packing-template functions move to PackingService, which already owned all
three template tables — resolving the admin-2 residual with no bridge. The pure
and module-scoped half moves to admin.helpers.ts, the version cache staying
module-scoped so admin.bridge's instance and the container singleton share one
GitHub fetch. A 1-export admin.bridge.ts serves scheduler.ts's cron, the only
out-of-container consumer. Recipe steps 3 and 4 are no-ops: no src/mcp/tools/
admin.ts has ever existed and the plugin host never imported the domain.
Four lines are non-verbatim, all path re-anchoring one directory deeper, both
resolved paths verified against the emitted dist layout.
admin.dto.ts clears all twenty AdminController allow-list entries, retiring the
three 'enabled must be a boolean' checks plus 'permissions object required' and
'Object body required' for the pipe envelope. The module e2e goes DI-native: the
3-method whole-module mock dies, 6 cases become 15 over real SQL.
Quirk fixes on top, each pinned: the three places toggles go fail-closed with an
append-only backfill migration; the template item routes honour :templateId;
updateUser rejects an empty username/email; createInvite 404s an unresolvable
trip_id; listOAuthSessions guards its scopes parse; GitHub fetches gain a timeout
and size cap and failures cache briefly; updateOidcSettings/updateAddon/updateUser
become transactional; admin.oauth_session.revoke is renamed to match its
siblings; two mutations gain their missing audit rows; getAuditLog drops the
_parse_error sentinel.
Tests move with IDs preserved (ADMIN-SVC-001..069 incl. the pre-existing 029/030
gap and duplicated 069, VNOTIF-001..007), plus ADMIN-BR-001 and ADMIN-SVC-070..076.
Twelve schemas covering every AdminController body, replacing the four that were
imported nowhere. Deliberately permissive wherever AdminService already owns a
bespoke 400 ('Invalid role', 'Name is required', 'Username, email and password
are required'), so those errors survive byte-identically: role is z.string() not
an enum, template name and the user-create fields stay optional. savePermissions
keeps z.unknown() values so unknown levels still land in the 200 response's
skipped list; addon config and the notification-preference matrix stay records so
per-addon keys and plugin channels are not stripped; devTestNotification's inApp
accepts the object form the dev panel actually sends.
'Abflug'/'Abflugzeit' are flight-specific, but the reservations.departureDate/
departureTime keys are shared across all transport types in TransportModal
(flight and train waypoints). Switched to 'Abreise'/'Abreisezeit', which read
naturally for any mode of transport.
Import-scan regeneration: notificationService and the folded
inAppNotifications leave the graph, the notifications cluster is consumed as
plain infra by the DI service, and the unblock cascade over-delivered —
repointing memories' sends erased that cluster's last domain edge, so BOTH
adminService and journeyService join the ready frontier (adminService heads
the order). Adds the dated correction bullet and the 'Quirks fixed after the
notifications fold' section (fixed vs deliberately-preserved).
notifications.dto.ts wraps the @trek/shared request schemas via createZodDto
and types all five @Body() params; the five NotificationsController entries
leave the body-contract allow-list (ratchet gate enforced). Contract catch:
the client sends server/token: null on test-ntfy to mean 'use the saved
value', so testNtfyRequestSchema gains .nullable() there (spec pinned). The
inline response-enum and url-type checks die — malformed bodies now get the
pipe's standard envelope; valid bodies behave byte-identically. Two
integration cases pinning pre-ratchet tolerances (flat notify_* preferences
body, body-less test-smtp POST) update to the current wire contract.
The send() dispatcher (services/notificationService.ts) and the in-app store
(services/inAppNotifications.ts) fold together into the DI-native
NotificationsService over DatabaseService + RealtimeService; the prefs
matrix, transports, channel registry and inAppNotificationActions stay plain
infra modules. A 1-export notifications.bridge covers scheduler, legacy
adminService/memories and the six lazy fire-and-forget sends; AdminController
and the plugin RPC host (24th factory dep) inject. The 5-tool registrar +
notifications-in-app resource move to notifications.mcp.ts. Tests move with
case IDs preserved (NSVC-*, INOTIF-*, + NSVC-020 bridge pin); ~18 suites
repoint path mocks/warm-ups; the module e2e goes DI-native.
Also carries the verified-quirk fixes that would normally trail as a
separate fix(server) commit (file-overlap with the fold): toUtcIso on all
three created_at paths (INOTIF-013/014), respond claims the response CAS
before running the action handler with release-on-failure (INOTIF-011/012),
synology's fire-and-forget send gains .catch, the dispatch-failure log
unwraps Error messages (NSVC-021), and the MCP resource description drops
its false 'unread first' claim.
Verified defects the relocation carried byte-for-byte, each pinned by a
regression test (AUTH-DB-089..093):
- getTravelStats kept dropping lat/lng of exactly 0 (equator / prime
meridian) through its falsy check — now explicit != null.
- registerUser's INSERT user -> invite used_count bump -> joinTripAsMember
ran statement-by-statement; the whole signup now runs in
db.transaction() so a mid-sequence throw rolls everything back.
- verifyMfaLogin's backup-code burn and last_login/login_count bump now
commit as one atomic pair.
- deleteMcpToken's revokeUserSessions call is best-effort (try/catch),
matching the changePassword/resetPassword revocations.
- updateApiKeys' current! non-null assertions became current?. ?? null —
a user row deleted mid-request degrades to a 0-row UPDATE, not a
TypeError/500.
Deliberately preserved (documented in migration-graph.md): the two
divergent "Password authentication is disabled" strings and enableMfa's
15-min pending-secret TTL after a wrong code.
Re-ran the three-pattern import scan over server/src: authService row
deleted, thirteen helper rows updated (nine lose their last legacy
importer), oidcService/passkeyService join the ready frontier as
auth.bridge repoints, adminService is blocked on notificationService
alone, atlas.bridge's death recorded, dependency-honest order step 5
ticked, a post-auth-fold corrections entry added (bridge tax held exactly;
the reverse cycle-break, the in-container module fan-out and path-mock rot
are now tracked shapes), and the preserved auth quirks listed as trailing
fix(server) candidates.
DTO ratchet for the auth domain (collections sibling-commit precedent):
- shared/src/auth/auth.schema.ts gains the six missing request schemas —
mapsKey (nullable: the client clears with an explicit null), apiKeys and
settings (per-key partials, nullable key values), appSettings (open
record — the server's ADMIN_SETTINGS_KEYS allow-list and its bespoke
lockout/self-MFA 400s stay service-owned), mfaDisable, resourceToken
(the 'Invalid purpose' 400 stays service-owned) — plus spec cases.
- New nest/auth/auth.dto.ts createZodDto wrappers; all 14 raw @Body()
params typed (9 AuthController + 5 AuthPublicController), validated by
the global ZodValidationPipe.
- All 14 auth entries removed from body-contract-allow-list.ts
(PasskeyController's 4 stay until passkeyService migrates).
- auth.e2e mirrors the production pipe and pins the { error: 'field: ...' }
envelope on a shapeless login body; auth.controller.ts drops its
duplicated avatarDir const in favour of the auth.helpers export.
Fold the 1497-line services/authService.ts into the DI-native nest/auth
domain (Wave-5 chain opener, unblocked by the atlas fold):
- Pure crypto half (backup-code hash/match/generate, stripUserForClient,
key masking, the import-time DUMMY_PASSWORD_HASH timing equaliser and
avatarDir mkdir — documented parity exceptions) moves verbatim to the
plain module nest/auth/auth.helpers.ts.
- DB half folds into AuthService over DatabaseService + injected
PermissionsService (ex permissions.bridge) + AtlasService (ex
atlas.bridge, deleted on schedule). mfaSetupPending and the per-email
reset throttle stay module-scoped so the bridge and container instances
share them; avatarDir and require('../../../package.json') re-anchor one
directory deeper — with the two injection swaps, the only non-verbatim
lines. Controller-facing method surface unchanged.
- New 8-export auth.bridge.ts for out-of-container consumers: mcp/index.ts
token verification, the journey/notifications/transports registrars'
isDemoUser, and the still-legacy adminService/oidcService/passkeyService.
- 15 domain *.mcp.ts demo guards now inject AuthService (their modules
import AuthModule); atlas.mcp.ts alone stays on the bridge because
AuthService injects AtlasService and the reverse module edge would close
an AuthModule<->AtlasModule cycle (places.mcp precedent). OidcService and
PasskeyEnabledGuard inject AuthService too.
- Tests move with case IDs preserved: authService.test.ts ->
auth.helpers.test.ts, authServiceDb.test.ts -> auth.service.test.ts
(+ AUTH-BR-001..007 bridge delegation, + AUTH-DB-050..088 pinning the
previously untested getAppConfig/login/MFA/reset/admin-settings branches
now under the src/nest coverage gate). auth.e2e converts DI-native (real
bcrypt, real audit rows); oidc.e2e swaps its dead path mock for an
instance spy; websocket/integration suites repoint to bridge/helpers.
Quirks preserved byte-for-byte (0-coord drop in getTravelStats,
un-transactioned registerUser/verifyMfaLogin multi-writes, unguarded
revokeUserSessions in deleteMcpToken, current! assertions in
updateApiKeys) — fixes land separately.
Fold the 1612-line services/atlasService.ts into the DI-native AtlasService:
the stats aggregation, visited countries/regions with the #1490 tombstone and
cascade logic, and the bucket-list CRUD become methods over the injected
DatabaseService. The ~750-line pure-geo half — the bundled admin0/admin1
boundary stores with their #1576 streaming builders, the point-in-polygon
indexes, Nominatim geocoding with its shared throttle, and every module-scoped
cache including the import-time unref'd cleanup interval — moves verbatim to
the plain module atlas-geo.ts so the caches stay process-global across the
container instance, the bridge instance and test helpers (assetPath
re-anchored one directory deeper, the only non-verbatim line).
The 10-tool mcp/tools/atlas.ts registrar plus all four atlas resources in
mcp/resources.ts move onto the decorator registry as atlas.mcp.ts — the when:
atlas-addon gate is parity here (legacy tools and resources both gated; the
REST controller deliberately does not). resources.test.ts retires with its
last two cases, which move to tools-atlas-expanded.test.ts. The plugin RPC
host swaps its 9 atlas imports for the injected AtlasService (23rd constructor
dep); a minimal 2-export atlas.bridge.ts serves the one legacy consumer,
authService.getTravelStats, and dies when authService migrates.
Adopts atlas.dto.ts over the existing shared schemas and clears all three
AtlasController entries from body-contract-allow-list.ts: the hand-rolled
'name and country_code are required' 400 becomes the ZodValidationPipe
envelope (todo/places trade) while the whitespace-only bucket name keeps its
legacy 'Name is required' trim guard.
Fixes the verified defects the relocation had faithfully carried, each with a
regression test (ATLAS-SVC-031..036 plus two MCP casing cases): the four
multi-statement mark/unmark writes now run in db.transaction() (the region
cascade nests as a savepoint), the trip-less countryPlaces early return
honours manually_marked, the '|| null' bindings that dropped lat/lng 0 and
empty-string notes on bucket update become '?? null', the mutating bucket SQL
is user-scoped, and the MCP region/country-places tools uppercase their codes
to match REST.
Quirks deliberately preserved: the detached GET-path place_regions backfill
writes, getStats' two divergent return shapes, the import-time cache-cleanup
interval, getCountryFromAddress's unvalidated bare 2-letter codes, and the
whitespace-only-name silent no-op on bucket update.
Tests move with their case IDs: atlasService.test.ts becomes
unit/nest/atlas.service.test.ts (pure-geo imports from atlas-geo, DB calls
through the constructed service, plus bridge-delegation cases), the atlas e2e
converts to the DI-native pattern (real SQL against the full temp-db schema),
and the plugin-host path mock becomes a constructor stub.
The set of uploads subdirectories is hand-mirrored between the image build
and the boot-time mkdir in server/src/index.ts, and drift is invisible
until a user whose bind-mounted uploads dir is not writable by node hits an
EACCES on the first upload to the missing dir (#1762).
Parse both lists and assert they match, with a guard case so the regexes
cannot silently match nothing and pass vacuously.
Both were created lazily on the first upload, which needs write permission
on the uploads mount point itself at request time. On hosts whose
bind-mounted uploads dir is not writable by the runtime user, every feature
writing into an already-existing subdir keeps working while journey photo
and place image uploads fail with EACCES (#1762). Creating them in the
image and at boot turns that into one loud startup failure instead of a
stray 500 on upload.
Also drop the recursive chown over all of /app from the final layer: it
copied up every inode it touched and duplicated node_modules, dist and the
client bundle into a 518 MB layer. The chown now rides the layer that
installs node_modules, later copies carry --chown=node:node, and the last
layer chowns only the paths it creates. Everything under /app stays
node-owned; image size 2.36 GB -> 1.68 GB.
The place fold capped the Google list-import body at 8 MB but left the Naver
importer reading its pages unbounded — same attacker-influenced URL (the folder
id), and worse in practice because that path is a pager that buffers a fresh
body on every iteration.
fetchPage now checks the declared content-length before reading and the actual
length after, returning the existing 'Failed to fetch list from Naver Maps' 502
on either. Reading the body as text and parsing it explicitly (rather than
apiRes.json()) is what makes the post-read check possible; the malformed-payload
branch keeps its 'Invalid list data received from Naver Maps' 400 because the
JSON.parse stays inside the same try.
The post-read check is the one that matters for a hostile server: a chunked
response carries no content-length, so the declared check alone is evadable.
Covered for both providers now — PLACE-SVC-073b (Google, chunked) and
PLACE-SVC-074..077 (Naver: declared, chunked, malformed, happy path).
The five Naver fetch mocks in the integration suite move from json() to text().
Fold the 1029-line services/placeService.ts into the DI-native PlacesService:
the CRUD + ratings SQL, the GPX/KML/KMZ importers and the Google/Naver list
importers become methods over the injected DatabaseService, byte-identical
statements, COALESCE semantics and the If-Match conflict protocol (#1135).
The pure half — frozen XML parsers, the KMZ unpacker, the dedup predicates,
the Google hex-id parsers, reclaimPhotoCache — moves to places.helpers.ts
(maps.helpers precedent).
The 10-tool mcp/tools/places.ts registrar plus the trek://trips/{tripId}/places
resource move onto the decorator registry as places.mcp.ts. search_place travels
with them because its gate is places:read, not the read-only geo group, and now
injects MapsService. TripsService, DaysMcp, BookingImportService and the plugin
RPC host (PluginHostDepsFactory's 21st constructor dep) all inject PlacesService,
so the domain needs no bridge at all. The two assignments.bridge imports that
places.mcp.ts keeps are a deliberate cycle break: AssignmentsModule imports
DaysModule and DaysModule now imports PlacesModule for days.mcp.ts's place
creation, so injecting AssignmentsService there would close
DaysModule -> PlacesModule -> AssignmentsModule -> DaysModule. Same seam
reservations.mcp.ts uses, same trade trips.bridge.ts documents.
Also folds services/placeEnrichment.ts in — its DB/websocket/Maps half becomes
PlacesService methods over the injected DatabaseService/RealtimeService/
MapsService, its pure match selector joins places.helpers.ts — which retires
nest/maps/maps.bridge.ts with its last consumer.
Adopts places.dto.ts for the seven grandfathered body contracts and clears every
PlacesController entry from body-contract-allow-list.ts. Two consequences worth
knowing: the bespoke 400s 'Place name is required', 'ids must be an array of
numbers' and 'URL is required' are now the ZodValidationPipe envelope, and
because a pipe runs before the handler a malformed body 400s ahead of the
trip-access 404 it used to follow — the same trade the todo and trips migrations
took. placeBulkUpdateRequestSchema.ids drops .min(1) so the endpoint's
empty-list short-circuit stays reachable.
Fixes four defects the relocation had faithfully carried, each with a
failing-first regression test:
- create dropped lat/lng of exactly 0, so a place on the equator or the prime
meridian lost its coordinates (`x || null` cannot tell absent from zero).
- duration_minutes: 0 was unsettable — replaced by the 60-minute default on
create, and read as "absent" by COALESCE on update.
- the journey delete hook fired on unscoped ids: DELETE /:id, POST /bulk-delete
and the delete_place MCP tool could detach another trip's journey entries for
an id they then refused. All three scope first via the new
PlacesService.scopedIds, mirroring the guard PluginHostDepsFactory already had.
bulk_delete_places also moves its hook ahead of the DELETE —
journey_entries.source_place_id is ON DELETE SET NULL, so running it afterwards
left the entries as orphans.
- the place search treated % and _ as LIKE wildcards; a bare % returned the whole
trip. The three clauses now carry ESCAPE with the term escaped.
Plus hardening on the Google list fetch: the response is capped at 8 MB
(declared content-length and post-read length, the transit.service precedent)
and its JSON.parse is guarded, so a malformed provider payload produces the
existing 'Invalid list data received from Google Maps' 400 instead of throwing.
Quirks deliberately preserved: the non-COALESCE route_color (#776),
currency/transport_mode staying unclearable, importGpx returning null for a file
that parses but yields no usable geometry, update skipping the If-Match check
when the stored updated_at is null, and every string-valued `x || null`.
Tests move with their case IDs: placeService.test.ts and the wrapper suite merge
into unit/nest/places.service.test.ts, kmzUnpack into places.helpers.test.ts,
placeEnrichment into places.enrichment.test.ts. The places e2e converts to the
DI-native pattern (real SQL against temp-db DDL), tools-places.test.ts is kept
and extended so it exercises the decorator path through the test registry, and
the plugin-host and booking-import path mocks become constructor stubs.
The desktop navbar's centred tab pill is absolutely positioned, so on the
md-range (tablet, ~768-1024px) its right edge slid underneath the right-hand
action cluster (notification bell + user menu) once the tab text labels made it
wide enough.
Hide the tab text labels below the lg breakpoint (icon-only on tablet, icon +
label on desktop ≥1024px), which keeps the pill narrow enough to clear the
actions. The icons gain title/aria-label so they stay accessible without text.
Co-authored-by: phreaky <phreakydev@gmail.com>
Keeping a stay whole moved its cards to the next sheet, and the header bar
stayed behind on the old one — empty, above a stretch of white, with the
same day printed again overhead. The repeat itself is the continuation
marker from #1471; what was wrong is that there was nothing left to
continue from.
break-after on the thead and break-before on the tbody are both ignored by
Chromium here, measured rather than assumed. Holding the whole day together
is what works: a day that fits moves as one, and a day too long for a page
still breaks and still repeats its header. Scoped to the flowing layout,
since a day that starts its own page can never strand one.
An accommodation could be cut in half by a page break — the check-in time
at the foot of one sheet, the hotel name and address at the head of the
next, under a repeated day header that read like a duplicate day. Place and
note cards have carried break-inside: avoid all along; this one never did.
It could always happen to a day that overflowed its page. Letting the days
flow means content meets a page edge far more often, which is how it turned
up. Printing the same trip through Chromium: two torn stays before, none
after.
The export button is the one toolbar button with a hand-rolled tooltip
instead of the shared one. It was anchored with `right: 0` and never
wrapped, so its label grew leftwards out of the day plan — the leftmost
pane — and off the window.
The shared tooltip is portalled and clamped to the viewport, which is what
every neighbouring button already uses. The hover state it needed was
threaded down from DayPlanSidebar and has no other reader, so it goes too.
The preview shipped a bare checkbox, which is not what a toggle looks like
anywhere else in TREK. The overlay is imperative DOM, so the switch from
Settings/ToggleSwitch cannot be rendered into it — as static markup it would
carry no behaviour — but its geometry and its tokens can be, and the control
now reads as the same one.
Every day started a new page, so a plan of short days printed one sheet per
handful of lines and looked nothing like the view it was exported from.
The preview now carries the choice, next to Save as PDF, and remembers it.
The two layouts differ by a single class on <body>, so ticking the box
re-lays the open preview out immediately instead of rebuilding the document
and re-fetching every photo.
Breaking per day stays the default — this only adds the other option.
The popup is a modal, so it portals out of the sidebar and misses the
opt-in its rows sit inside everywhere else. Its rows already carried a bare
draggable attribute, which on a tablet meant a long press selected the row
label and nothing moved.
Dragging a place from the right pane to the left one did nothing on an
iPad or a Samsung tablet — the row was selected as text instead. Phones
never reach this layout, they get the mobile shell below md, so the coarse
pointer check in dragDisabled was only ever switching tablets off.
It was also doing that for a reason that had gone away. What swallowed the
scroll gesture in #1432 was the drag-drop-touch polyfill, and that has been
confined to hybrid laptops since. Width is the gate that still means
something: below lg the plan and the places are separate tabs, so there is
nothing to drag between.
So touch is no longer a reason to disable the drag, and the planner arms
the long-press bridge instead where the pointer is coarse. Rows get
-webkit-touch-callout and user-select off there too, otherwise the press
that starts the drag brings up the selection first. The isTouch prop had no
other use and is gone, along with a comment claiming arrow reorder buttons
take over on touch — there are none.
A finger never starts an HTML5 drag on Android, and the answer used to be
drag-drop-touch: a document-wide polyfill that turned every touch over a
draggable element into a drag. That cost the places list its scrolling
(#1432) and the map its one-finger pan (#1440), so it now only loads on
hybrid laptops.
This is the narrow version. It only watches touches that begin on a
draggable row inside a container that opted in, and it waits out a long
press before claiming the gesture, so a swipe is still a swipe. Once armed
it replays the real drag sequence on the element under the finger, which
leaves every existing drop handler working as it is.
Nothing is wired up to it yet.
WAL was hardcoded, and on network-backed storage — Azure App Service Linux,
SMB/NFS volumes on NAS and PaaS — SQLite documents it as unsafe, because the
coordination runs through the -shm file and mmap. Operators had no way out.
TREK_DB_JOURNAL_MODE and TREK_DB_SYNCHRONOUS now set both pragmas, defaulting
to today's behaviour. The mode lives in the file header, so all three
processes that open travel.db read the same configuration — the server,
reset-admin.js and migrate-encryption.ts, which used to hard-set WAL and would
have silently reverted a deliberate DELETE setup on the first key rotation.
An invalid value falls back and warns rather than refusing to start, and the
mode that actually took effect is read back and logged, so an operator can see
whether the setting landed.
The scheduled backup archived travel.db directly. The archiver reads its
entries during finalize(), so a WAL checkpoint writing pages back mid-stream
tore the copy — and the -wal that would make it recoverable isn't in the zip.
The manual path was moved to VACUUM INTO for exactly this reason; the
scheduler never followed. It calls the same createBackup() now, which also
gets it the encryption key and plugin data it was missing, so a scheduled
backup is finally restorable onto another install.
The pre-rotation copy in migrate-encryption.ts had the same problem: a plain
file copy of a live WAL database, and it is the operator's only rollback if a
key rotation goes wrong.
Auto-backups now exclude the re-derivable photo caches, matching manual ones.
Transitous exposes routeShortName, which many operators — German long
distance among them — do not fill. displayName carries the identifier
passengers actually see, with routeShortName as the fallback.
MSheet binds the Escape listener in an effect, so a key fired the moment the
sheet content appears can land before the listener exists. Under a loaded
full-suite run that made the test fail on its own timing. It retries until
the sheet is gone rather than pressing once.
The ring was checked against the viewer's clock, so a trip in Europe planned
from Singapore showed the wrong state. It evaluates in the timezone of the
place now, from Google's structured opening periods rather than the localised
description text — parsing display strings breaks as soon as the account
language is not English. Midnight-spanning hours, closed days and 24/7 are
covered.
assign broadcast, unassign did not, so the place stayed on the Plan tab until
a manual refresh. It emits the same event as the REST route now. createPlace,
updatePlace and deletePlace in the same file had the identical gap.
The 12h/24h rules lived in three copies and two of them ignored the setting.
They are one function now. splitReservationDateTime also truncated the
meridiem instead of converting it, so 3:00 PM became 3:00 — three in the
morning. Values already stored that way normalise when a form renders the
field, without changing the payload or the stored format.
Attachments were treated as ready-made HTML, but a real mail is multipart
with base64 or quoted-printable transfer encoding. What reached the model was
the plaintext headers followed by a wall of base64, truncated before any
content — which is why a sender name from the Subject line still showed up
while the booking never did.
The html part is preferred, plain text is the fallback, transfer encoding and
charset are decoded and entities resolved before the text goes out. No new
dependency. If no text part is found the previous raw path still runs.
Opening one trip fires a 404 per place without a photo, 30 to 60 of them from
one IP in seconds. CrowdSec and fail2ban read that as scanning; the reporter
had three users banned within five minutes. A place having no photo is not an
error and no longer answers as one.
OSM places were also sent to Google as node:/way:/relation: ids, which can
only return 400 INVALID_ARGUMENT — a billed round-trip and about 1.3s per
place. getPlacePhoto already guarded coords: this way; the OSM prefixes now
get the same treatment, as does getPlaceDetailsExpanded, which had no guard.
The negative cache separates a place having no photo from a failed provider
call instead of expiring both after five minutes and replaying the storm.
The a/b/c shards stopped being necessary in 2022 and d has since lost its
DNS record. The map never asked for d — Leaflet defaults to abc — but the
offline prefetcher rotated over a to d, which is where the failing requests
came from, and it cached most tiles under a host the map never requests.
Presets and the prefetcher use the apex host now. A URL already saved in
user_settings is rewritten where it is read and written back, so existing
instances heal without a migration touching what someone typed themselves.
The CSP only allowed https://*.tile.openstreetmap.org, and a wildcard host
never matches the apex domain, so the prefetch fetch would have been blocked.
The service worker rule had the same gap and would have stopped filling the
tile cache.
Treats the gate as a floor rather than a target — the suite runs well above it,
and a PR that legitimately trades a few points should not have to move the
numbers to land.
214 items came out of writing the tests. Real defects are corrected, dead code
is removed rather than covered, and the tests that pinned the old behaviour
move with them.
Also drops two agent coverage scratch files that slipped into the coverage PR
and adds the ignore rule that keeps them out.
These are the only two test files that call request.formData() inside an MSW
handler, and they hold exactly the three tests that failed on CI while passing
locally. The body axios' XHR adapter produces under jsdom is one Node 24's
multipart parser rejects, so the handler threw before reading a field and the
component reported a failed upload. Reading the raw payload asserts the same
thing without depending on the parser.
The helpers built a one-byte File and then redefined size to 100, so the
multipart body and the length it announced disagreed. Node 24 rejects that
while 22 lets it through, which is why only CI saw it: every test that
uploaded such a file and expected success failed, the one expecting an error
passed for the wrong reason, and the KML case with a truthful File was fine.
The client suite is ~11 minutes now. With typecheck and lint sitting in front
of it in the same job, a single lint finding threw away the whole test signal
for that run — twice today. Both jobs pay the install and the shared build,
which is cheap next to running the two in series.
waitFor and findBy* default to 1s. That holds on a dev machine but not on a
4-core runner executing 527 test files in parallel forks, where three import
dialog tests timed out waiting for their upload to settle.
A literal U+FEFF inside a regex trips no-irregular-whitespace, which is an
error rather than a warning, so the client lint job failed. The escape is
equivalent.
Pushes lines from 95.2% to 98.1% and branches from 85.2% to 89.9%, mostly
in the release-notes modal, the reservation overlay, the journey entry sheet
and the trip dialogs.
Swap the factory's six tripService imports for the injected TripsService; delete services/tripService.ts and the collab/vacay bridges with it; prune the days/budget bridges to their surviving exports; update the migration docs and regenerate migration-graph.md.
Fold the legacy tripService SQL into TripsService; port the trips MCP registrar/resources/trip-summary prompt to trips.mcp.ts and the share-link tools to share.mcp.ts (predicate access for the trips:delete/trips:share/canReadTrips gates, deprecation notice via the attach ctx); adopt trips.dto.ts and clear the seven allow-list entries; delete todo/share bridges with their last consumers. Includes the trailing quirk fixes: transactional deleteTrip/deleteGuest, owner display_name in listMembers.
assignment:moved requires newDayId again; assignment:reordered and
day:deleted drop their drift-union variants now that every emitter
sends the canonical shape.
move_assignment omitted newDayId (moved assignments vanished from
collaborator views), reorder_day_assignments sent assignmentIds where
the client reads orderedIds (reorders emptied the day remotely), and
delete_day sent { id } where the client reads dayId (days never
disappeared). Registry drift findings from the WS contract work.
remoteEventHandler's two switches become enumerable lookups keyed by
TrekWsTripEventName (bodies unchanged); WebSocketEvent.type derives
from the registry. wsEventPolicy.ts makes the client's non-reactions
explicit (HANDLED_OUTSIDE_TRIP_STORE + IGNORED_WS_EVENTS) and the
registry-parity test turns an unclassified server event into a red
test instead of silence. Zero behavior change.
broadcast/broadcastToUser gain registry-typed overloads (event name
must be a TREK_WS_EVENTS key, payload must match its schema) with the
plugin: namespace as the deliberate escape hatch; the 13 service
wrappers forward the generic signature. Compile-time only — the
runtime pass-through and all call sites are unchanged.
TREK_WS_EVENTS: 65 trip-scoped + 29 user-scoped events with Zod
payload schemas transcribed from the emitting call sites; derived
name unions and payload types; plugin: namespace reserved as the
escape hatch. 8 payload-drift shapes modeled as unions and flagged.
Nest services inject RealtimeService (call-time delegation to the
websocket module, DatabaseService pattern) instead of importing
broadcast/broadcastToUser globals. Zero behavior change; test edits
are constructor-arg additions plus RealtimeModule in e2e harness
imports only.
MPlaceSheet had no relation to route_geometry at all, so the section is
new. It saves through the single-field patch the sheet already uses for
the image, which keeps it clear of the edit form and its prefill.
The browse list gets the same colour stroke as the desktop sidebar. It
matters more here: the mobile map is full-bleed with no sidebar next to
it, so without the stroke a coloured line has nothing to tie it to.
New shared component rather than a ninth copy of the swatch grid that is
already inline in eight places. Selection is drawn with var(--accent) so
it follows whatever accent the user runs.
The custom-colour cell listens for the native change event instead of
React's onChange: for a colour input React maps onChange to input, which
fires on every drag inside the OS picker — dozens of writes, each one a
409 candidate against the base-version header. The auto cell previews the
colour you would get back, not the one currently set, which is a
different value as soon as anything is picked.
The row sits next to the track stats block, not inside it: that block
returns null on unparsable geometry and the colour control has no
business disappearing with it. The elevation chart and the distance pin
follow the track colour too, otherwise the panel and the map read as two
unrelated objects.
In the places list the track icon becomes a stroke in the colour the line
is drawn in. The map has no legend of its own, so that stroke is the only
thing tying a line back to a row. Undo after deleting a track now
restores its geometry and colour instead of bringing back a bare point.
The MapLibre layer was already data-driven, so it only needed the feature
property filled. Leaflet needed more work: the GPX polyline passed its
style as bare color/weight props, and react-leaflet only calls setStyle
when the pathOptions reference changes — a recoloured track would have
kept its mount-time colour until the next reload. Every other vector
layer in here already uses pathOptions.
A track with a colour gets a white casing so it survives satellite
imagery and dark basemaps. The casings live in their own pane below the
lines: Leaflet stacks paths in insertion order, so drawn inline they end
up over the line of a track that was added earlier. They are always
mounted and hidden by opacity, otherwise giving a track a colour remounts
the path into the wrong position.
Tracks are clickable now — 3.5px is not a target, so the hit area is a
transparent 14px line on top. bubblingMouseEvents is off on it: paths
bubble to the map by default (markers do not), and the map's own click
handler would clear the selection the track just made. The GL side needs
the mirror image of that, a guard so a click on a cluster bubble over a
track still belongs to the cluster.
Imported tracks all render in the same blue: the GPX/KML importers never
assign a category, so category_color is null for every one of them and
several walks in the same area are impossible to tell apart.
Adds a nullable route_color on places, kept out of the COALESCE group in
updatePlace so an explicit null can reset a track back to inheriting its
category colour. The hex is validated in the Nest controller rather than
in the shared contract: the update body is an open record on purpose, and
tightening it would mean removing PlacesController.update from the
ratchet-only body-contract allow-list.
The importers now hand each new track a colour of its own, picked from
the palette entries the trip is not already using. That happens in the
Nest service over the injected DatabaseService — the colour assignment is
new logic, and new logic does not belong in the legacy place service (see
server/src/nest/README.md). The field itself is only threaded through the
existing update/create statements, which the proportionality rule covers.
Also fixes duplicateTrip, whose explicit column list has been silently
dropping route_geometry all along, so a duplicated trip lost its tracks.
- the Frankfurter fetch now carries an AbortSignal timeout (10s) and a 1 MB
response-size cap, and the body is boundary-validated through a type guard
instead of an as-cast
- fetch failures are logged instead of silently swallowed (callers still get
the graceful null)
- the inflight-coalescing cleanup moved into .finally so a rejected fetch
can no longer leak the key
- the dead convertWithRates export (zero call sites) is dropped from the
service and the bridge
Kept on purpose: the beyond-TTL stale-cache fallback, the >1-keys
empty-response heuristic and the || falsy-coercion defaults.
Permissions: stored levels are validated at load (corrupt/out-of-range rows
fall back to the default consistently across all readers), the load-error
swallow is narrowed to missing-table (anything else logs), and an
all-skipped save no longer opens a transaction or flushes the cache.
The module-scoped cache stays by design (shared DI+bridge invalidation),
and the checkPermission 'admin' case stays — it is reachable (correct deny
for non-admins on admin-level actions), not dead code.
Audit: data/logs creation is lazy (no import-time disk side effect; the
LOG_LEVEL freeze remains — tests/setup.ts timing contract), the log helpers
gate on a real severity threshold (error < warn < info < debug; prod
default 'info' unchanged), file-IO failures leave a console.error trace
instead of vanishing in bare catches, and resolveUserEmail treats only
null/undefined as anonymous so a real id 0 resolves via the DB.
Two verified defects carried through the parity fold:
- The create-path accommodation metadata sync gated on the raw
accommodation_id instead of the resolved one, so a hotel whose
accommodation was just auto-created never received its metadata
check-in/out times or confirmation. Now keyed off resolvedAccommodationId
(RESV-SVC-006 re-pins the fixed behavior).
- The multi-statement writes ran outside transactions: create (accommodation
insert + reservation insert + endpoint save + metadata sync), update
(accommodation upsert + update + endpoint replace + sync), remove (the
3-delete cascade), setReservationTravelers (delete + inserts) and
resyncReservationDays now run in db.transaction() — a mid-write failure no
longer leaves partial state (RESV-FIX-001/002 pin the rollbacks).
Left as-is on purpose: the truthy updatePositions dayId check, the
empty-string COALESCE keeps on title/status/type, the TEXT accommodation_id
normalization, the optional day_plan_position wire tolerance and the
swallowed notifyBookingChange catches — all contract or intentional
behavior, not defects.
- nest/README.md: reservations joins the DI-native list; recipe history gains
the residue-fold paragraph; next up is dayService per the dependency-honest
order.
- migrate.md: wave-4 entry ticked, with the migration-graph correction borne
out — reservationService imported neither budgetService nor dayService.
- migration-graph.md: regenerated for the post-fold import graph (row/node
removed, frontier and order updated, dayService is the next pick).
- plugins/DI-MIGRATION.md: the factory's last plain-function reservation
import is drained.
Folds the 626-line services/reservationService.ts into the DI-native
ReservationsService — byte-identical SQL, statuses, bodies, error strings
and broadcasts; the legacy file is deleted.
- reservations.mcp.ts: the 5 legacy tools (create/update/delete_reservation,
reorder_reservations, link_hotel_accommodation) + the trip-reservations
resource move to the decorator registry (identical names, schemas,
annotations, error strings, broadcasts); the canWrite/canRead registration
gates become declarative access markers. Legacy registrar and resource
entries removed.
- reservations.bridge.ts: 9 legacy-named exports + 3 type re-exports for the
outside-container consumers (tripService, airtrail import/sync, the
still-legacy transit + transports registrars, transitItineraryService's
type-only import).
- notifyBookingChange folds to the legacy (tripId, actorId) signature; the
controller passes user.id and the plugin-host factory drops its last
plain-function reservation import for a private method over the injected
service (constructor unchanged).
- TripsService and BookingImportService inject ReservationsService instead of
importing legacy functions.
- Tests: reservations.service.test.ts rewritten as a real-SQL suite
(RESV-SVC-001..030, RES-TRAV-001..005 moved verbatim,
RESV-BRIDGE-001..009 pinning the bridge); both e2e suites converted to the
DI-native pattern; path mocks repointed to the bridge or constructor stubs;
tools-reservations kept and extended with the moved resource cases.
Quirks preserved on purpose (parity first — see the follow-up fix commit):
the raw-accommodation_id metadata-sync gate, the un-transactioned
multi-statement writes, the truthy updatePositions dayId check, the
empty-string COALESCE keeps, and the TEXT accommodation_id normalization.
ReservationsController's 4 grandfathered @Body() params now validate through
the global ZodValidationPipe (createZodDto over the existing @trek/shared
schemas); their allow-list entries are removed (the boot gate rejects stale
entries). The bespoke 400 bodies (Title is required / positions must be an
array / user_ids must be an array) become the pipe's field envelope, and
non-numeric traveler ids now 400 instead of being silently dropped.
reservationPositionsRequestSchema is loosened to the real wire contract:
day_plan_position is optional (the legacy route never validated position
items and an absent value binds NULL) — pinned by RESV-006.
AccommodationsController stays grandfathered until the dayService migration.
Folds the legacy services/fileService.ts into the DI-native FilesService
(injected DatabaseService, byte-identical SQL). Load-time constants move to
files.constants.ts; files.bridge.ts survives with a single export for the
request-time getAllowedExtensions read the module-scope multer configs need
outside DI. TripsService and PluginHostDepsFactory inject FilesService.
Also rides the nestjs-zod ratchet (FileUploadDto/FileUpdateDto/FileLinkDto
over the @trek/shared contracts, FilesController entries dropped from the
body-contract allow-list) and fixes the verified legacy quirks: file-link
insert failures now surface instead of being swallowed, updateFile coerces
an empty-string description to NULL like createFile, and formatFile's
misleading optional trip_id annotation is gone.
PluginRuntimeService's operatorEgressHosts/pruneErrorLog and PluginsService's
egressHostCount were module-level helpers on the raw db proxy; they fold into
their classes as methods over the injected service. (PluginHostDepsFactory's
equivalent conversion rode the settings commit — same files.) No injectable
class under src/nest imports the db global anymore; the remaining importers
are the documented module-level seams (bridges, platform shell,
plugin-host-state, signature-status, the bare plugin user-setting readers).
SettingUpsertDto/SettingsBulkDto wrap the existing @trek/shared schemas; both
SettingsController entries drop off the body-contract allow-list (boot
ratchet). Per the todo precedent the dead bespoke 400 guards go: missing/empty
key and a non-record settings body now answer with the pipe envelope
({ error: 'key: …' } / { error: 'settings: …' }) instead of 'Key is required' /
'Settings object is required', and arrays for settings are rejected. The local
masked-value literal now imports MASKED_SETTING_VALUE from @trek/shared.
The settings e2e registers ZodValidationPipe and pins the DTO 400s plus the
bulk masked-sentinel skip.
Folds the legacy services/settingsService SQL into the Nest SettingsService
(byte-identical statements; BEGIN/COMMIT blocks become db.transaction()).
AdminService and ShareService inject it; llm-config.resolver becomes the
injectable LlmConfigResolver (SettingsService + DatabaseService), injected by
LlmParseService and PluginHostDepsFactory. No MCP surface, no bridge.
Whole-file commit, so this also carries the post-migration quirk fixes that
live in the same files: null serializes as '' (a stored "null" leaked out of
getDecryptedUserSetting as the literal string), bulk upserts skip the masked
sentinel (mirroring the single-upsert no-op) and return the written count.
Unit suite moved to tests/unit/nest with SET-SVC-001..026 preserved,
027..030 pin the fixes; settings/admin e2e converted to DI-native.
New tripInviteLinkCreateRequestSchema in @trek/shared (single source of
truth), a createZodDto wrapper on the controller's raw @Body(), and the
TripInviteLinkController.create entry dropped from the body-contract
allow-list (the boot ratchet enforces it).
expires_in_days accepts number | digits-only string | null | absent — the
string branch is digits-only so garbage like "7abc" is a 400 instead of
silently parseInt-ing to 7; the empty string (blank form input, meaning no
expiry) stays accepted. The e2e registers ZodValidationPipe and pins both
the bounded-expiry and 400 paths.
Fold the legacy tripInviteService SQL into the Nest TripInviteService over
the injected DatabaseService (byte-identical statements; the JS-side expiry
check and non-positive-means-no-expiry coercion preserved) and delete the
legacy module. No bridge, no MCP port and no plugin-host change needed —
nothing outside the container consumed this domain.
Also wraps the rotation's probe + write + re-select in db.transaction()
per the repo rule that multi-statement writes are never bare.
Unit suite moved to tests/unit/nest preserving TRIP-INVITE-001..006 and
adding 007 (expiry pinning); the e2e converted to the DI-native pattern
(real SQL against temp trips/trip_invite_tokens tables).
- One global ZodValidationPipe (APP_PIPE, createZodValidationPipe) validates
any @Body() typed with a createZodDto class; the custom exception factory
keeps TREK's { error: 'field: message; ...' } envelope byte-identical.
- The 7 piped controllers (collections, places, vacay, todo, health,
airtrail x2) now type bodies with colocated <domain>.dto.ts wrappers —
the Zod schemas in @trek/shared stay the single source of truth.
- OpenAPI: the hand-rolled api-zod.ts enricher is replaced by nestjs-zod's
cleanupOpenApiDoc; request bodies surface as $refs to named DTO schemas.
- validateBodyContracts() runs in buildApp() after init and refuses to boot
when any POST/PUT/PATCH @Body() (including @Body('field') sub-reads) lacks
a ZodDto metatype. The 175 pre-existing legacy handlers are grandfathered
in body-contract-allow-list.ts; the list is ratchet-only — the gate also
throws on stale entries, so migrating a domain forces their removal.
The optional metadata fields (category, due_date, description,
assigned_user_id — plus priority on update) now accept explicit null: the
client clears fields by sending null, which the schemas rejected. checked
accepts the legacy 0/1 numeric form alongside boolean (the integration net
pins it). Drops the client-side casts that papered over the too-narrow
types.
updateCategoryAssignees ran its DELETE + INSERT OR IGNORE loop unwrapped, so
a mid-loop failure could leave a category half-cleared. Also widens the item
params to the nullable shapes the wire contract actually sends (prep for the
todo body validation).
The hero picks the trip that is running, else the next one coming up, else
just trips[0] — and that pick was then removed from the grid below. The first
two cases are deliberate: the trip is already on screen, showing it again
right underneath would be noise. The trips[0] fallback is a different thing.
It fires precisely when nothing qualifies, so it borrows a trip that belongs
in the grid, and removing it there leaves the user with a hero and an empty
list. Someone whose trips are all finished sees "No trips yet" under a card
of the trip they just came back from.
Split the two: only a trip that earned the hero is taken out of the grid.
Undated trips were caught by this too, and worse — getTripStatus returns
null rather than 'past' for them, so a lone dateless trip appeared in no
tab at all.
The empty state was the second half of the report. It equated "planned is
empty" with "you have no trips" and told people to create their first one.
Gate it on there being no trip at all, which is the condition the mobile
dashboard already uses.
Mobile shares useDashboard, so it is fixed along with the desktop.
Closes#1706
sessionStorage outlives a logout inside the same tab, so the keys stay readable
after the next user signs in. They are namespaced by user id, which stops the new
user's plugins from reading them, but the data is still sitting in the tab — the
same reason logout already drops the appearance snapshot, the SW caches and the
user-scoped offline DB. Purge them there too.
Three things around it, from reading the handlers next to the new ones:
- The inactive-plugin purge ran before the store was updated, so a sessionStorage
failure cost us the plugin list we had just fetched. Commit first, purge after.
- The session handlers answered without checking requestId, unlike confirm and
geolocation — an answer the frame can never match is worth skipping.
- The mock host threw plain Errors while the real bridge rejects with .code set,
and checked the key before the trip scope where the host does the reverse, so a
plugin's error handling could pass its tests and still break in the app.
The context table listed the new field without saying what a plugin is meant to
do with it. It is a display hint — booking codes still arrive in full over
trek:invoke — and reading it as redaction is the one mistake worth preventing.
The dev preview hardcoded it to false, so the blurred path was the only host flag
an author could not try out; it gets a toggle like reduce motion and the rest.
The widget's new healthApi.features() call left BackgroundTasksWidget.test.tsx
red — the api/client mock never exported healthApi, so both existing tests threw
on mount. The mock now carries it, pending by default like the poll backstop, and
the retry itself finally has tests.
Writing them turned up two gaps. The retry never called saveImportFiles, so the
files only lived in memory: reloading during the retry parse left the review with
no source document to attach to the bookings it creates, unlike every import
started from the modal. And the button stayed live while the request was in
flight, so a second click started a second force-ai job on the same files.
Resolving every waypoint day from endpoint.local_date also changed the edit path,
where it inverts the precedence the form uses everywhere else: local_date is
denormalised and the server only restamps it in the day_id branch of insertDay,
and only when reservation_time is set. Dragging a booking to another day rewrites
day_id/end_day_id and leaves the endpoints alone, so local_date points at the old
day — and reopening the form and pressing Update silently moved the booking back.
Gate the lookup on a prefill, where there is no saved day to lose, and the edit
path behaves exactly as it did before. Same change in the mobile sheet, which
shares the seeding code and would otherwise still show empty day selectors on an
imported booking.
syncAll runs straight off the back of a successful login, while the app
is still mounting its first screen. Kicking off a bulk background
download at that exact moment is the worst possible timing even with the
prefetch throttled, so hold the tiles back until requestIdleCallback
says the browser has nothing better to do.
Files keep their old behaviour — they are a handful of requests, not
thousands.
The prefetch dispatched every tile of a trip's bounding box in one
synchronous loop — up to MAX_TILES per trip. A trip spanning a few
countries enumerates thousands of tiles at zoom 10 alone, so logging in
fired ~8k no-cors requests back to back and kept them coming for half a
minute. All of them go through the Workbox fetch handler, which then has
to write a cache entry plus an expiration record per tile, so everything
the app itself requested queued up behind the burst and the UI just sat
there.
Run the queue through a small worker pool instead, and let it stop early
when the user goes offline or logs out.
Two things fall out of that:
- Tiles already in Cache Storage are now skipped before the request is
made, so a resumed or repeated prefetch over a warm cache is nearly
free.
- {s} has to be a pure function of the tile coordinates for that lookup
to hit. The rotating counter gave the same tile a different host on
every run, which also meant caching it up to four times.
The mobile create/edit sheet never showed the currency, so every trip
started from the dashboard silently took the server default of EUR and
could only be corrected from the trip planner, which reuses TripFormModal
and does have the field.
Same control as the desktop form — searchable select over currenciesWith(),
gated on trip_edit like the rest of the sheet.
Move the tag SQL from src/services/tagService.ts into TagsService over an
injected DatabaseService (identical statements, #10b981 default, COALESCE
semantics). Non-Nest consumers (MCP tools, plugin RPC host) go through the
new tags.bridge.ts, which exports the legacy function names 1:1 from a
module-level instance over the shared db Proxy — the settled pattern for
the remaining legacy-service migrations (recipe documented in
src/nest/README.md). Tests are DI-native: direct service construction in
unit (with bridge delegation cases for the coverage gate) and
DatabaseModule + real SQL in e2e.
The bump to sharp 0.35.0 left the @img/sharp-linuxmusl-* pins on 0.35.1,
so npm nested the 0.35.0 binary under node_modules/sharp while hoisting
libvips to the root. The nested .node then looks for libvips-cpp.so.8.18.3
next to itself, finds nothing, and the Alpine image build dies in the
client prebuild:
Could not load the "sharp" module using the linuxmusl-x64 runtime
ERR_DLOPEN_FAILED: Error loading shared library libvips-cpp.so.8.18.3
Neither docker workflow runs on push, so this only surfaces when a release
is cut. Pin both binaries to 0.35.0 so they hoist next to libvips again.
Used, carried over, comp days and remaining all fit a desktop sidebar; on a
phone they collapsed into four scraps of 9px text fighting for one row. Only
the balance is left, as a pill tinted by how much of it there is — the rest is
still a tap away in the entitlement view.
The toolbar chip says "Company" rather than "Company Holiday": it sits beside
two other buttons and the building icon already carries the meaning. The
allowance stepper reads "Days" instead of "days / year", reusing the label the
desktop entitlement tile has always used — the year is on the pill above it.
The panel dropped the eight feature tiles and the self-hosted line — they read
as a spec sheet on a screen whose job is to get you in. What is left is the
mark, the tagline and one line of copy, centred over the map, each with a soft
dark halo so the dots cannot eat the type. The tagline and subtitle never wrap:
they scale with the panel instead, since it is a fixed share of the viewport.
The takeoff after sign-in used to tell a different story than the screen it
came from — cloud ellipses, a lone plane, a contrail. It now picks the panel up
and escalates it: the same map, every route departing at once, the logo landing
as the network completes, then a soft hand-off into the app.
Replaces the drifting plane silhouettes with the thing TREK actually does:
routes lighting up between cities across a dot map of the world. Arcs leave a
city, a light travels the curve, the destination lands, and the arc retires to
make room for the next.
The map is drawn once into an offscreen layer and blitted per frame, so only
the handful of live arcs cost anything — no shadowBlur in the frame path. The
same component covers the sign-in moment via `variant="takeoff"`, where the
network fires as one cascade and stays lit instead of cycling.
Reduced motion gets the map as a still frame.
The login screen is unauthenticated, so it cannot call the Atlas endpoint the
rest of the app uses for country geometry. This is that same admin-0 bundle,
sampled along the coastlines onto a 300x150 grid and delta-encoded — 15 KB
instead of a 28 MB fetch nobody is allowed to make yet.
The page still described Vacay as strictly January to December and never
mentioned half days at all, so both new features would have shipped
undocumented.
Windows for all three year types, the fallbacks (anniversary without a hire
date, a day the month cannot have), usage counted across the calendar-year
boundary, carry-over chained over periods rather than years, deleteYear over a
shifted period, and the reads with and without a viewer.
For comp: excluded from usage, a half comp day still costing nothing,
converting a vacation day back, and comp_used reported on its own.
The half-day button carried a ½ glyph and comp/flex a clock, neither of which
appears anywhere on the calendar. Both now show what a click actually puts on
the day: the orange dot for a half day, a hatched disc in the selected person's
colour for comp.
The digit on an all-comp cell also needed help. The hatch lets the surface
through between its stripes, so a white number sank into it — a shadow carries
the contrast now and every logged day keeps the same digit colour. Mixed days
still have a solid segment under the number and get none.
The server has returned comp_used since the leave type landed, but nothing
showed it — comp days simply vanished from the sidebar. They now sit beside
the tiles rather than inside the used/left arithmetic, which is the point of
them.
Carry-over is labelled from the window the server counted that row over, not
from the viewer's own setting, so a fused plan with mixed leave years does not
tell one member their days came "from 2025" when they came from a period that
started in July.
The card also folds now. It is the longest thing in the sidebar once several
people are fused in, and the state belongs to the person looking at it, so it
lives in localStorage rather than on the plan.
Calendar, fiscal (month + day) or from the hire date, per user — it sits with
the plan settings but belongs to the person, so fused members each keep their
own and the grid follows whoever is looking at it.
Both grids render twelve months from the window start with year rollover
instead of a fixed Jan–Dec, and the mobile month navigation moves through slots
of that window, so stepping past December lands in January of the same period
rather than jumping back to the start of the calendar year. The month effect
keys on the window's shape rather than the settings object, which the store
replaces with an equal copy on every refresh — otherwise any sync would throw
away the month the user had open.
Known edge for a start day past the 1st: the cards are month-aligned, so the
first days of the start month are drawn but count toward the previous period.
Counting stays day-exact.
The server owns the arithmetic entitlements are computed with, but the grid has
to render the same twelve months and the overlays have to filter against the
same range, so the window resolves on both sides. yearWindow.ts is that mirror,
including the Feb-29 clamp and the anniversary-without-a-hire-date fallback.
The store loads the settings before the year list picks a default — with a
shifted year the current period is not necessarily the current calendar year.
Holidays are fetched per calendar year the window touches and clipped back to
what the grid draws; a school break straddling New Year comes back from both
fetches, so markers are de-duplicated per day. The label is part of that
identity, otherwise two calendars left on the same colour collapse into one.
Adds GET/PUT /addons/vacay/year-settings and makes the reads window-aware, so
the backend from the previous commit is actually reachable.
Entries and shared calendars now load over the caller's window instead of a
date prefix — a shifted year spans two calendar years and the prefix silently
dropped the second half. Reads are month-aligned (the grid draws month cards),
while entitlement stays day-exact; the two coincide for every window starting
on the 1st, which is all of calendar. deleteYear clears entries per author over
that author's period, and company holidays only over the range every member
shares, so one member deleting a period cannot reach into another's live one.
getStats reports the window it counted, and getOwnPlan seeds the period today
falls into rather than the calendar year.
A start day the month cannot have (Feb 29 from a hire date, a stored Feb 31)
is clamped to one every year has — otherwise the boundary string matches no
date and the whole window slides. Anniversary before a hire date is entered
resolves to the calendar year instead of borrowing a month a previous fiscal
setting left behind.
MCP reads pass their user through; without one the range stays the plain
calendar year, so plan-scoped callers are unchanged.
The kind column changed four things the tests still asserted the old way:
toggleEntry gained a parameter, the toolbar gained the comp modifier, shared
calendar rows carry kind, and a multi-person cell is now segment overlays
instead of one split gradient. splitBackground went with it — nothing but the
test still referenced it.
The toolbar tests picked the half-day button by index, which the new button
shifted; they select by name now so the next modifier does not break them.
Comp/Flex days (#1074) — complete. A second leave type that does not deduct from the vacation entitlement (flextime / overtime offset). New kind column on vacay_entries, excluded from usedDays via a CASE sum; toggleEntry carries kind through service/nest/store with change-detection; desktop and mobile get a comp toolbar toggle next to the half-day one (orthogonal, so a half comp day is free); comp segments render as a diagonal hatch of the person colour, per-segment for split days (Version B). comp_used is surfaced separately in getStats. Translated into all 23 locales.
Configurable vacation year (#737) — backend only. Per-user vacay_user_settings (calendar / fiscal / anniversary), resolveYearWindow, and usedDays / compUsedDays / getStats / carry-over now computed over the leave-year [start,end) window. 'calendar' resolves byte-identically to the old date-prefix behaviour, so nothing changes for existing users and the feature is inert until the settings UI lands. Frontend (settings selector + calendar window rollover), the settings endpoint, window-aware getEntries and holiday loading are still to do — see the PR description.
The always-on card picker is gone — travelers are assigned in the booking and transport modals now, matching mobile. The card shows assigned travelers read-only as pills in the grey field box, TravelerChips is replaced by a shared TravelerPicker, and the toolbar filter gets gradient avatars with an accent ring and check badge.
Gemini reached through the OpenAI-compatible endpoint returns
JavaScript-object-literal output — single-quoted strings, unquoted keys,
trailing commas — which strict JSON.parse rejects. The reservation list
then came back empty and the booking-import UI showed nothing (#1638).
Add a shared parseLenientJson() that tries strict JSON first and falls
back to JSON5 (which accepts exactly that relaxed superset), and route
both the OpenAI-compatible and Ollama-native parse paths through it,
replacing the two identical strict parsers.
A multi-leg flight imported "as one flight with a layover" is detached
from sync by design — AirTrail has no single flight to round-trip a
layover chain to, so it is created with sync_enabled=0 (#1535). The
reservations badge collapsed every sync_enabled=0 AirTrail row into "Not
synced" with the tooltip "This flight was removed in AirTrail", which is
false for a merged booking that was never removed (#1646).
Give the multi-leg case (metadata.legs.length > 1) its own state: show
the normal AirTrail badge with a truthful tooltip explaining it stays a
one-time import. A genuinely removed single-leg flight still shows "Not
synced". Adds reservations.airtrail.layoverHint across all locales.
The admin "user defaults" merge kept any user value that was merely
present, so an empty string shadowed the system-wide default. The
client's Settings save always writes every field — a blank Mapbox token
included — so a user who ever opened Map Settings got '' instead of the
admin token, and clearing their own token could never restore the
fallback (#1634).
Treat an empty/null user value for a defaultable key as "not set" when an
admin default exists, so it falls through. Non-defaultable keys and keys
without an admin default keep their exact stored value.
- The finale condenses out of light: ~1000 particles rise from the world and
gather into the anthem's letterforms before the crisp DOM text takes over;
the language cascade waits for the word to settle.
- Replaced fear-act lines decay letter by letter — the words of fear fall
apart instead of vanishing.
- The climax boom rocks the whole frame for half a second, in sync with the
sub impact and the border shatter.
- Hope-act lines reveal sentence by sentence, in speech rhythm.
- The standing anthem breathes — a barely-there scale and glow pulse.
- Beacon corners now use the dashboard tool widgets' --r-xl radius.
The show now ends standing: KEINE ANGST with the full 23-language cascade
stays up over the lit world, music carrying on until the traveler closes it —
no credits hand-off. Skip jumps straight to that standing frame.
Beacon: dismiss × removed, copy block centered on the canvas, the title gets
its quiet jewelry treatment — ivory-to-gold fill, a traveling sheen and a slow
breathing glow.
The overlay now portals to document.body: the sticky dashboard sidebar forms
its own stacking context, so the fixed navbar (z-200 at root) used to float
above the show regardless of the overlay's z-index.
The transition out of the fear act was too brutal: the silence now arrives as
a fast pull-away with the hall tail ringing out instead of a 100ms kill, the
pivot line trades its impact boom for a slow warm A1 swell (the first calm
breath), and the line itself breathes in over 3.4s.
Audio review fixes: the wind's swell LFO moves to a series tremolo stage so
the release can actually silence it before the cut (it used to warble through
the silence and click on stop); the heartbeat's noise layer becomes the
intended high-passed click instead of a dull lowpassed thud; dispose() pulls
the master to zero before the deferred context close so leaving the show
cannot pop.
Copy sits centered on the night-world canvas over a corner-pooling scrim; the
title grows monumental tracking, a gold sheen and a soft glow; the text CTA
becomes a slowly pulsing golden play circle that fills on hover. The version
eyebrow and the red live dot are gone — title, one line, play. Nothing else.
- Lines get a layered soft glow plus a radial pool of darkness under the text
zone, so they hold up over the burning border map without a hard box.
- Arc heads now interpolate fractionally along the curve with ease-out growth
(the whole-segment quantization read as stutter), in the show and the beacon
teaser alike.
- The score gains a convolution reverb with a procedurally generated impulse
response; the fear act trades raw saws for an organic bed (sine sub,
drifting fifth, breathing noise rumble, a semitone grind in the dread act),
the shimmer runs through a feedback echo, impact tails bloom in the hall,
the heartbeat stays close and dry.
The beacon is now a dark cinema card: a small canvas paints a breathing
night-world with golden travel arcs on loop, colored glows drift over it, the
title carries a gold sheen sweep and the copy sits on a bottom scrim. Core
colors ride inline so no dashboard widget CSS can wash the card out again.
Visual: city lights now come alive in the opening and die one by one while the
borders burn; three staccato words strobe with the heartbeat before the cut;
at the climax the borders no longer fade — they shatter into drifting sparks.
The traveler's own places ignite gold in the release act, with their real
numbers in the line ('{p} places. {c} countries. And not once did the world
hurt you.'). All heavy geometry is decimated to a vertex budget and pre-baked
into offscreen layers, so the frame path is a handful of drawImage calls — the
per-frame border stroke that dropped the show to ~1 fps is gone.
Sound: heartbeat with a click transient, wind textures, a doubled noise riser,
cue-driven sub-bass impacts and a gliding D–A–f#m–E progression with a sub
root under the anthem instead of the static pad.
Also: dashboard FAB hidden behind the show, suspended-AudioContext recovery
(Safari), tab-hidden pause for clock and audio, aborted geo fetch on close,
valid beacon markup (dismiss as sibling button), RTL-aware beacon, real
reduced-motion static frame, and native-speaker fixes for nl/sv/br/fr/it/es/
tr/gr copy.
A one-release statement against racism, hatred and fascism, told the way TREK
knows the world: through travel. A pulsing beacon in the dashboard sidebar
(desktop only, dismissible) opens a ~80s full-screen show — the world as a map
of dots whose borders burn red and then dissolve under a growing web of travel
connections, driven by a fully procedural Web Audio score (heartbeat, drone,
A-major release; no audio assets, no copyrighted material) and hard-cut lines
in all 23 TREK languages, closing on the anthem cascade and a credit to the
song that sparked it (Danger Dan — Keine Angst).
Self-contained in components/FourZero (own copy dictionary, no shared i18n
churn); wired into the dashboard with two marked lines. Removing the folder
plus those two lines retires the moment after the release.
* feat: add vacay school holiday overlays
* fix(vacay): remove SOH control character corruption in locale files
13 of 22 vacay locale files contained a 0x01 (SOH) control character
that corrupted the closing }; token. This broke the shared package
build (tsdown/rolldown PARSE_ERROR: Invalid Character) and cascaded
to all client tests failing to resolve @trek/shared.
Affected locales: ar, br, ca, cs, es, fr, gr, hu, id, it, ja, ko,
nl, pl, ru, sv, tr, uk, vi, zh, zh-TW
* fix(vacay): school-holiday migration order, i18n dupes, settings layout
- Move the school-holiday columns migration to the end of the migrations
array. It sat before the #1435/#1281 migrations, which every existing DB
had already run, so the runner skipped it and school_holidays_enabled /
vacay_holiday_calendars.type were never created — PUT /addons/vacay/plan
then 500'd with "no such column".
- Remove duplicate i18n keys in 13 locale vacay files (TS1117) that failed
the shared + client typechecks.
- Move the school-holidays block into the right column, directly under
public holidays, instead of full-width below everything.
- Add the school-holidays section to the mobile Vacay settings sheet.
* feat(vacay): modern school-holiday day styling + combined hover tooltip
- Replace the stacked 2px underlines with a single rounded accent band and a
soft background wash on plain school-break days (desktop + mobile — mobile
did not render school holidays at all before).
- Fold school holidays into the custom day tooltip under a divider instead of a
separate native title, and show the tooltip on every set day (whole vacation
days too), not only half days / shared calendars.
---------
Co-authored-by: Otto <uif86164@automotive-wan.com>
Co-authored-by: Maurice <mauriceboe@icloud.com>
* fix(dayplan): expose route tools on a hotel-to-hotel transfer day (#1297)
* feat(places): add a Planned filter to the places pool (sidebar, map, mobile)
* fix(places): stretch the mobile pool filter chips to fill the row
* feat(dayplan): persist per-leg and per-day travel mode (#1281)
The day-plan route used one transport mode for the whole day, held only in
ephemeral state. Store it instead: leg_transport_mode on an assignment is
the mode of the leg leaving that stop (null = inherit the day default), and
days gains a default_transport_mode. Both nullable, so existing itineraries
are unchanged.
Adds PUT /assignments/:id/transport and PUT /days/:id/transport — their own
endpoints so setting the day default can't wipe notes/title the way the
general day update would — mirroring the assignment_time path.
* feat(dayplan): choose the travel mode per segment (#1281)
Each leg between two stops now routes with its own mode, so a day can mix
walking, driving and plugin-routed legs. Tap the connector between two
places to change that leg's mode (desktop popup / mobile sheet); the
whole-day Foot/Car picker sets the persisted day default and leaves
explicitly-set legs alone. The connector and map redraw in the chosen mode,
and each route segment is tagged with the mode it was drawn in.
New strings across all 23 locales.
* feat(collections): add places by coordinates and sort the list
Raised in #1640: collections behaved differently from trips — you
couldn't add a place by typing coordinates, and the list had no order
control.
The add-place modal and the place detail sheet now take an editable
address plus latitude/longitude, with a "lat, lng" paste shortcut, so a
place can be saved by GPS without searching. The place update path
learned to persist coordinates, so an existing one can be corrected
later.
A sort control in the filter row toggles between the saved order and
alphabetical by name. The add-place button moved into that row to match
the other controls, and the add-label button dropped its dashed one-off
styling. New strings land across all 23 locales.
* feat(collections): make the add-place button icon-only
Keep the filter row compact — the button carries just the plus icon (with
an aria-label/title), squared off to match the dropdown height.
Trip members can rate a place 1-5 stars; the place shows the average,
with a who-voted custom tooltip and a star-sort toggle in the places
sidebar. Saved collection places are ratable too (plus a min-rating
filter), and ratings travel with a place copied between a trip and a
collection — carrying only the votes of users who belong to both sides.
Adds a full Collections MCP surface (list/get, CRUD lists + places,
status, rate_collection_place, labels, sharing, copy-to-trip) under a
new collections:read/collections:write scope, plus rate_place for trip
places; rating aggregates flow through the existing place outputs.
Mobile trip + collection place sheets gain the rating row. New UI
strings translated across all 22 locales.
* feat(reservations): add a parking booking type (#1444)
* feat(reservations): parking type option, icon and color across the UI
* test(reservations): cover the parking booking type
* docs(reservations): document the parking booking type
* feat(places): upload a custom place image on the server (#1136)
* i18n(places): custom place image labels
* feat(places): click a place thumbnail to set a custom image
* feat(places): custom place image on mobile place and collection sheets
* test(places): cover custom place and collection image upload
* docs(places): document custom place images
* fix(places): carry osm_id on the embedded place so day-plan thumbnails auto-fetch (#1136)
* feat(places): custom tooltip on the place image controls
* fix(places): show a custom place image on map markers, skip the redundant photo fetch (#1136)
* feat(vacay): read-only calendar shares on the server (#444, #667)
* feat(vacay): shared calendars card and ring overlay on desktop
* feat(vacay): share sheet and calendar rings on mobile
* test(vacay): cover the calendar share flows
* docs(vacay): document read-only calendar sharing
* feat(vacay): support half vacation days (#552)
Log a vacation day as a half day (0.5) as well as a full day and count it
as such against the entitlement. A "Half day" mode joins the calendar
toolbar on desktop and mobile; half days render as a diagonal half-fill in
the day cell, and the used/remaining totals carry the fraction.
Closes#552
* refactor(vacay): half-day as a per-person toggle with a corner badge
The diagonal half-fill collided with the two-person split — one person's
half day looked like two people off. Drop the diagonal: the cell fill keeps
showing who is off, a half day gets a small corner ½ badge, and a hover
tooltip lists who is off on that date and how much (only when a half day is
present). Logging moves from a third mode to a Half-day toggle that applies
to the selected person, on desktop and mobile.
* feat(vacay): mark half days with an orange corner dot
Swap the ½ badge for a small orange dot in the bottom-right corner, mirroring
the blue trip dot — quieter in the grid, and the hover tooltip still spells
out who is on a half day.
* fix(vacay): make the fraction migration idempotent
The crosswalk test replays the migration tail (rewind + re-run), which expects
every later migration to be idempotent. Guard the vacay_entries ADD COLUMN
behind a pragma_table_info check so the second run is a no-op. Also only set
dayVisual's `half` flag when true, so full days keep their exact
{ background, numColor } shape for the mobile day-model unit test.
* style(vacay): nudge the mobile day dots slightly inward
The blue trip dot and orange half-day dot sat right in the cell corner on
mobile; move both a touch toward the centre so they read cleaner.
* 3.3.0 (#1472)
* feat(plugins): grow the frame bridge — fill pages, confirm, openExternal, live context
- page/trip-page hosts pass fill: the frame pins to 100% height and ignores
trek:resize, so a kit plugin's auto height report no longer collapses a full
page into a floating island with dead space below (widgets keep self-sizing)
- trek:confirm renders the native ConfirmDialog host-side (the sandbox has no
allow-modals) and answers trek:confirm:result; one at a time
- trek:openExternal opens validated http(s) URLs in a noopener tab — the
sandbox has no allow-popups, so plugins simply couldn't link out before
- trek:notify accepts an optional duration, clamped to 1.5-15s
- context gains dir (rtl/ltr) and is re-pushed when locale or format settings
change, not just on appearance mutations
- core events for the trip in view are forwarded as trek:event — names only,
never payloads, mirroring the server-side events surface
* fix(plugins): move trip warnings out of the content area
The warning pills overlaid the top of every planner tab at full width, sitting
on the map and its toolbar. Now a warning from a plugin that owns a trip-page
tab renders as a compact chip in the navbar centre (click jumps to the tab; the
navbar centre is free on trip pages), and everything else floats above the
content at the bottom instead. Mobile keeps all warnings in the bottom overlay
since the desktop navbar isn't there. The trip tab's frame also opts into the
new fill mode.
* feat(plugin-sdk): 1.4.0 — motion library + new bridge helpers in the kit
Mirrors the host's animation vocabulary 1:1 into TREK_UI_CSS (menu/popover/
modal/backdrop/toast enters, drawer variant under 640px, page-enter, stagger,
skeleton shimmer, chart reveals) including the reduced-motion degrade to a
gentle fade. window.trek grows confirm(), openExternal(), onEvent() and a
notify duration, and applyContext now stamps lang/dir on the document so RTL
hosts get RTL plugin UIs.
* feat(plugins): surface registry download counts in browse
The registry now aggregates GitHub release download counts per plugin
(TREK-Plugins#18) as an entry-level downloadCount. Project it through
browse/detail and show it as a compact stat on the browse cards and in the
detail meta grid. Counts are raw asset downloads (updates and CI included),
so the UI says downloads, not installs.
* docs(plugins): document the grown bridge surface and motion classes
* fix(plugins): harden the new bridge paths
Review pass over the bridge additions:
- keep the unstable useToast() object out of the effect deps (ref instead) —
it re-created the effect on every parent render, and with the new live
repost that meant a trek:context flood into the frame
- reset loads/height/confirm state and key the iframe when a host swaps
pluginId in place (tab bar, /plugins/:id) — the new plugin's document was
refused as a 'navigated' frame and every kit promise hung
- confirm dialogs always lead with the host-controlled plugin name so a
plugin can't dress its dialog up as a TREK system prompt; answer/refuse
moved out of setState updaters (StrictMode ran them twice)
- Number.isFinite on the notify duration (NaN parked a sticky toast)
- don't forward other plugins' namespaced broadcasts as trek:event; a
plugin's own plugin:{id}:* broadcasts now reach its frame though
- teach the SDK dev preview the confirm/openExternal contract so
trek.confirm() resolves in /preview
- 999,950 downloads formats as 1M, not 1000k
* fix(planner): let plugin warning chips grow wider before truncating
The nav-centre chip capped at 340px, so a longer warning (e.g. the TREK x
Japan weather prompt) was ellipsised almost immediately. Scale it with the
viewport up to 520px so most messages read in full while still yielding on
narrow desktops.
* feat(plugins): sort the plugin browser by download count
Discover now honours the sort dropdown (it was always alphabetical) and adds
a 'Most downloads' option that ranks the registry by downloadCount. The sort
keys are scoped per tab — updates-first stays with Installed, most-downloads
with Discover — and snap back to name when the tab can't offer them.
* feat(plugin-sdk): auto-upgrade native <select> to a host-styled dropdown
A sandboxed plugin can't reach the host's components, and a native <select>
draws its popup from the OS — so plugin dropdowns never matched TREK. The design
kit now enhances every <select> into a keyboard-accessible listbox that uses the
kit tokens, keeping the real element as the value/form source (it still fires
change). Authors write a plain <select> and get the host look for free; opt a
field out with data-trek-native. validate warns when a plugin ships a <select>
without inlining the kit.
* feat(plugins): add issue url link
* feat(plugins): reservations write + cross-trip reads
- db:write:reservations -> reservations.create/update/delete, gated exactly like
the REST/MCP path (reservation_edit + trip membership, acting user host-bound,
no impersonation) and delegating to ReservationsService so the accommodation,
budget-sync, booking-notification and reservation:* broadcasts match the web
app 1:1 — a booking/flight/import plugin can finally write a reservation
- trips.listMine / reservations.listMine: enumerate every trip and booking the
acting user can access (membership baked into listTrips, never a raw
cross-tenant SELECT) — dashboards/aggregates were impossible before
- audit: derive auditability from METHOD_PERMISSION so a new capability method
can't be added un-audited by omission
- typed ctx.reservations.* / ctx.trips.listMine, perm label (en/de), wiki
* feat(plugins): read scopes for journal, atlas, vacay and day notes
- db:read:journal / db:read:atlas / db:read:vacay expose the acting user's OWN
journals / visited countries+regions / vacation plan across all their trips
(user-scoped like costs.listMine, each gated on its addon being enabled),
reusing the addon's existing readers
- db:read:daynotes -> daynotes.list(tripId, dayId), trip-scoped and
membership-checked like the other trip reads
- typed ctx.journal / atlas / vacay / daynotes, perm labels (en/de), wiki,
audit resource labels, tests
* feat(plugins): day notes write scope
- db:write:daynotes -> daynotes.create/update/delete, gated under the app's
'day_edit' permission (like days) with the day verified to belong to the trip;
reuses dayNoteService and broadcasts the same dayNote:* events so open
sessions update live
- typed ctx.daynotes.create/update/delete, perm label (en/de), wiki, tests
* feat(plugins): run declared background jobs on a schedule
- plugins already declared jobs {id, schedule} but the cron was never wired. The
host now schedules them: host-entry reports each job's schedule, the supervisor
starts the jobs (node-cron) when the plugin goes active and stops them on
kill/deactivate so nothing leaks
- opt-in via a new jobs:run permission — scheduled work runs with NO acting user
(its trip reads stay refused; it can only use ctx.db and declared egress), so
background execution is a distinct, admin-granted capability. Invalid crons are
skipped and a throwing job can't break the host
- extracted a small, unit-tested scheduler (plugin-jobs.ts); perm label (en/de),
wiki, tests
* feat(plugins): read scope for saved-place collections
- db:read:collections -> collections.listMine() / collections.get(id): the acting
user's own collections (user-scoped, gated on the Collections addon), reusing
collectionsService
- typed ctx.collections, perm label (en/de), wiki, audit resource labels, tests
* fix(plugins): translate new permission labels to all locales + cover the new wiring
- add the 8 new admin.plugins.perm.* labels (reservations/day-notes writes, the
journal/atlas/vacay/day-notes/collections reads and jobs:run) to the remaining
20 locales so the strict i18n key-parity test passes again
- cover the create-rpc-host reservation / day-note / cross-trip / addon-read deps
(the real side-effect wiring the mocked rpc-host tests don't exercise) so the
src/nest 80% branch-coverage gate holds
* feat(plugins): dev-link — hot-reload a local plugin against real data
Answers a plugin developer's ask: today you either get `trek-plugin-sdk dev`
(fast hot-reload but MOCK/fixture data) or the full build->pack->upload->activate
cycle (real data, no watcher). Neither gives "local dir + hot-reload + real data".
- POST /admin/plugins/link registers a plugin from a LOCAL built directory by
symlinking it into the plugins volume — the loader already forks the resolved
real path, so ZERO loader change — and registering it INACTIVE as `local:link`.
Validates the manifest + refuses native binaries exactly like a sideload.
- POST /admin/plugins/:id/reload re-forks a linked plugin via the existing
deactivate->activate primitive (same grants, no re-consent unless the manifest
widened perms). A best-effort fs.watch auto-reloads on rebuild.
- It runs through the UNCHANGED capability RPC host: real, membership-gated data,
acting user host-bound, no impersonation — code origin never touches the gate.
- Gated behind TREK_PLUGINS_DEV_LINK on top of admin + kill-switch, because a
linked plugin bypasses the install-time signature model and, under `npm run
dev`, the OS jail is off. Off by default; never reachable in production.
- discovery follows a symlinked <root>/<id>; uninstall/link never delete the
author's source (link-safe removal for POSIX symlinks and Windows junctions).
* docs(plugins): document the dev-link real-data hot-reload workflow
Adds a "Test against a real instance's data (dev-link)" subsection next to the
mock-data SDK preview: TREK_PLUGINS_DEV_LINK, POST /link with a local built dir,
activate + consent, hot-reload via the file-watch / POST /:id/reload / Restart,
and the dev-only security caveats.
* feat(plugins): dev-link admin UI
- surface devLink (TREK_PLUGINS_DEV_LINK) in GET /admin/plugins so the panel shows
the link form only where dev-link is enabled
- AdminPluginsPanel: a "Link a local plugin" form (path -> POST /link), a Dev-Link
badge for source_repo=local:link, and adminApi.pluginLink/pluginReload
- fix: the plugin menu treated any non-local:upload source_repo as a GitHub repo,
so a dev-linked plugin rendered github.com/local:link links — exclude local:link
- labels for the 6 new dev-link UI strings across all 22 locales
* docs(plugins): document the dev-link admin UI
The dev-link section showed only the curl call — surface the Admin → Plugins
"Link a local plugin" field (the primary path) and the Dev-link badge, with curl
kept as the scripting alternative.
* feat(plugins): enrich core events with { entity, entityId }
Subscribed plugins now learn WHICH entity changed, not just the event name — a
reservation/place/day/... id derived host-side from an explicit per-family
whitelist. Threaded through the six event hops WITHOUT touching actingUserId: the
handler still runs with no user, so the id is not dereferenceable (a trip read is
still refused; the id says what to react to, not what it contains). A non-entity id
can never surface — budget:member-paid-updated yields the itemId, never the userId
— and bulk/reorder/sub-entity payloads carry no id. The mapper is pure, synchronous
and never throws into the core broadcast. No new permission (reuses
events:subscribe); backend-only.
* feat(plugins): packing write scope with #858 privacy-scoped broadcasts
- db:write:packing -> packing.create/update/delete, gated under the app's
'packing_edit' permission (like the REST path) with the host-bound acting user
as owner; reuses packingService
- replicates the packing privacy model 1:1 (the controller/service helpers aren't
exported): create/delete fan out to the item's viewers only (owner + recipients,
or the whole room for a Common item); update runs the four public<->private
transitions, dropping a freshly-privatized item from the room BEFORE re-adding it
owner-only so it never leaks. A stale write is BAD_PARAMS with no broadcast
- typed ctx.packing.create/update/delete, perm label (22 locales), wiki, tests
(rpc-host gating + the four transitions + owner-scoped delete)
* feat(plugins): tableContributor hook — host-rendered view columns/actions (backend)
The registry backend for plugin-contributed columns/actions in the native planner
views (the tabular-reservations use case), mirroring placeDetailProvider:
- hook:table-contributor + the tableContributor hook (getContributions(view,
tripId, ctx) -> TableContribution[]), double-gated (implement + grant) like the
other provider hooks
- GET /api/view-contributions/:view/:tripId — view whitelist + membership gate +
per-provider timeout/fail-safe, plus the hardening the older provider hooks lack:
every field is String-coerced + length-capped, kind/tone/target enum-whitelisted,
per-provider counts capped (<=20 columns / <=10 actions), and a column url must be
http/https/mailto (a javascript:/data: url is click-XSS into the native DOM)
- typed pluginsApi.viewContributions + the ViewContribution union, perm label
(22 locales), wiki, hardening tests
* feat(plugins): render tableContributor columns/actions in the reservations view
The frontend for the tableContributor hook: a reusable PluginContributions layer
(usePluginViewContributions + PluginColumns/PluginActions) that renders the
host-normalized column/action leaves NATIVELY — a column is text/badge/link, an
action is a button that calls the plugin route or opens its sandboxed frame in a
modal (plugin markup only ever runs inside the opaque-origin iframe). Wired into the
reservations cards (both ReservationCard and TransitJourneyCard) as a strictly-
additive footer keyed by reservation id: zero change to a card when no plugin
contributes. Fetched once per view, fail-safe.
* docs(plugins): bring the permissions wikis current with this cycle
Plugin-Permissions.md was missing every permission added this cycle — add rows for
the read scopes (journal/atlas/vacay/daynotes/collections), the write scopes
(reservations/daynotes/packing, packing noting the #858 owner-scoping), jobs:run
and hook:table-contributor, and correct the events:subscribe row for the new
{ entity, entityId } hint. Add the jobs:run row to Plugin-Development.md too.
* fix(maps): stop quick one-finger pans zooming the map on mobile (#1440)
The global drag-drop-touch polyfill installs document-level touch listeners
on phones. On every single-finger touchend it records a timestamp, and if the
next touch starts within 500ms it synthesises a dblclick on the target, which
the map's default double-click-zoom turns into a zoom-in. Two quick one-finger
pans therefore zoomed instead of panning.
The polyfill only bridges HTML5 drag-and-drop to touch for planner reordering,
which is already disabled on mobile (#1432), so gate its import to viewports
>=1024px (the lg breakpoint useIsMobile uses). Removes the phantom-dblclick
source on phones while keeping touch DnD on large viewports; fixes both the
Leaflet and GL renderers.
* fix(feeds): emit TZID + VTIMEZONE so subscribed calendars respect time zones (#1453)
exportICS emitted timed DTSTART/DTEND as bare floating times (no Z, no TZID),
which iOS/Google Calendar render in the subscriber's local zone instead of the
zone TREK shows. Resolve an IANA zone per timed event — transport endpoints use
their stored timezone (departure drives DTSTART, arrival drives DTEND), while
assignments and hotel/restaurant reservations derive it from place coordinates
via tz-lookup — and attach TZID backed by a VTIMEZONE component. The all-trips
feed now carries deduped VTIMEZONE blocks so TZID references still resolve.
* feat(plugins): render tableContributor contributions in the places + day views
Extends the tableContributor frontend to all three planner views: hoist the shared
PluginCardFooter into PluginContributions, wire the places sidebar (keyed by place
id, rendered as a sibling after each row so the drag/scroll row stays untouched)
and the day panel (keyed by day id, guarded for a null day). Strictly additive +
fail-safe like the reservations view — nothing renders when no plugin contributes.
* fix(vacay): source holiday subdivisions from ISO 3166-2 so all states show
The state/region picker for public-holiday calendars was built from the
union of each holiday's counties for the current year, so a subdivision
only appeared if some holiday that year was tagged with it. States with
no state-specific holiday (e.g. US-WA, and AR/FL/NV/WY in 2026) silently
vanished, blocking calendar creation (#1456).
Source the full, correctly-named subdivision list per country from
ISO 3166-2 instead, merged with any nager county code ISO lacks. Only
region-partitioned countries get a picker, so nationwide-only countries
keep allowing a country-level calendar. No server change needed —
selecting a state already yields federal holidays via applyHolidayCalendars.
* fix(costs): settlements honor custom per-member splits (#1458)
calculateSettlement read each member's custom split amount but its query
never selected budget_item_members.amount, so hasCustomSplit was always
false and every settlement fell back to the equal split. Select bm.amount
so custom amounts drive the balances.
Also blank the Overview 'Per Person' / 'Per Person·Day' columns and CSV
for custom-split items, where a single averaged figure is meaningless.
* feat(plugins): read-convenience + todos + packing bags + tags + roster
A wave of small, high-value capabilities:
- weather:read (ctx.weather.get) — the host's cached forecast, tenant-free
- db:read:categories (ctx.categories.list) — the global place-category list
- db:read:tags / db:write:tags (ctx.tags) — the acting user's own tags, ownership
re-checked before each write
- trips.members (ctx.trips.members) — the trip roster (id + display fields),
membership-checked
- db:read:todos / db:write:todos (ctx.todos) — a trip's to-dos, gated by the app's
packing_edit like the REST path, broadcasts todo:*
- packing bags on ctx.packing (listBags/createBag/updateBag/deleteBag/setBagMembers)
under db:write:packing — no privacy, plain room broadcasts
perm labels (22 locales), both wikis, rpc-host gating + create-rpc-host wiring tests
* fix(dashboard): render next-trip boarding pass stats on Safari (#1459)
The boarding-pass bar carved its ticket-stub notches with a two-layer
radial-gradient mask composited via mask-composite: intersect (and legacy
-webkit-mask-composite: source-in). Safari mis-composites that multi-layer
path to fully transparent, hiding the entire stats bar while Chrome renders
it fine.
Split .hero-pass into an outer wrapper (left notch) and a .hero-pass-inner
glass panel (right notch), each carrying a single-layer mask so the
mask-composite path is never exercised. Renders identically across engines
and degrades safely where mask-image is unsupported.
* feat(plugins): write scopes for atlas, vacay, journal and collections
The write half of the user-scoped addon reads:
- db:write:atlas -> ctx.atlas.markCountry/unmarkCountry/markRegion/unmarkRegion +
bucket-list create/delete. Every row is the acting user's own (visited_countries/
visited_regions/bucket) — no trip scoping, no cross-tenant surface. Unblocks
AirTrail-style two-way sync (#214)
- db:write:vacay -> ctx.vacay.toggleEntry/toggleCompanyHoliday. The plan is
resolved HOST-SIDE from the acting user's active plan — a plugin can never name
another plan, and toggleEntry only toggles the acting user's own PTO day
- db:write:journal -> ctx.journal.createEntry/updateEntry/deleteEntry, self-gated
by journeyService.canEdit (owner/contributor) against the acting user
- db:write:collections -> ctx.collections.create/update/savePlace/copyToTrip/
deletePlace, schema-validated; the service's per-collection role checks
(assertAccess 404 / assertCanEdit 403) map onto RESOURCE_FORBIDDEN
All addon-gated, userless contexts refused, audited. Perm labels (22 locales),
both wikis, gating + wiring tests.
* fix(admin): name the Costs add-on consistently in the catalog
The budget add-on catalog entry still resolved to 'Budget' while the
feature is labeled 'Costs' everywhere else (trip tab, navbar). Align
admin.addons.catalog.budget.name with each locale's trip.tabs.budget
label. Closes#1464
* feat(plugins): file attach, collab content and gated member-add
- db:write:files -> ctx.files.create/createLink/update/softDelete under the app's
separate file_upload/file_edit/file_delete rights. Content arrives as bounded
base64 (10MB decoded cap, well under the app's 50MB), the extension is validated
against the central blocklist BEFORE anything touches disk, and link targets
must live on the same trip (findForeignLinkTarget). Broadcasts file:*
- db:write:collab -> ctx.collab.createNote/createPoll/votePoll/createMessage
under collab_edit + the Collab addon, emitting the same collab:* events as the
app; service-reported errors surface as BAD_PARAMS
- db:write:members -> ctx.trips.addMember. Adding a member GRANTS TRIP ACCESS, so
it is deliberately its own permission behind the app's member_manage right
(default: trip owner only) and never bundled with a lower-risk write; the acting
user is recorded as the inviter, target must exist, owner/duplicate adds no-op
Perm labels (22 locales), both wikis, gating + wiring tests.
* fix(maps): honor check-in/out times for hotel bookend legs (#1465)
The day route drew the accommodation as the day's start/end whenever the
edge stop was a place, ignoring the morningIsSleptHere/eveningIsOvernight
provenance already computed by getDayBookendHotels. On a check-in day an
airport placed before check-in got a spurious hotel -> airport leg, and on
a check-out day a later "home" stop still got a home -> hotel return leg.
Add time-aware shouldDrawMorningLeg/shouldDrawEveningLeg helpers: the
morning leg is the home-base default on a check-in day but is dropped when
the first place is timed before check-in; the evening return leg is off on
a check-out day unless the last place is timed at/before check-out. Wire
them into the map polyline, the sidebar hotel connectors, and the Google
Maps export so all three stay consistent.
* feat(plugins): host-mediated notifications and LLM access
Two host-owned integration primitives — the plugin supplies intent, the host
owns the sensitive part:
- notify:send -> ctx.notify.send({title, body, link?, scope, targetId}). Delegates
to notificationService.send with a new plugin_notification event (raw title/body
carried as passthrough params), so recipient resolution, channel fan-out
(bell inbox + email/ntfy/webhook) and per-user preferences all match core 1:1.
Recipients are FORCED to the acting user (scope 'user', targetId === uid) or a
trip they belong to (scope 'trip'); scope 'admin' refused; the in-app link must
be a relative /path (open-redirect-safe). No arbitrary recipient, no impersonation.
Users can mute plugin notifications like any other event.
- ai:invoke -> ctx.ai.complete(prompt) / ctx.ai.extract(text, jsonSchema). Runs the
admin/user-configured provider via resolveLlmConfig + the existing extraction
client under the acting user; the host holds the (encrypted) key, the plugin
never sees it. Refused when no provider is configured; 20k-char caps. Output is
DATA (complete -> {text}, extract -> {results}) and never auto-written, so
prompt-injection can't reach a write without the plugin's own gated call.
plugin_notification wired through the shared NotificationEventKey + all 22 locales
(inbox passthrough + external channels). Perm labels (22 locales), both wikis,
gating + wiring tests.
* fix(budget): offer every Frankfurter-supported currency (#1470)
The cost currency picker was gated by a hardcoded 47-code list, so
currencies the app can actually convert (OMR, CRC, UGX, MKD, ALL, and
~115 more) couldn't be selected. Replace CURRENCIES/SYMBOLS with the full
set the Frankfurter v2 FX API supports (archived BGN/HRK dropped), unify
the dashboard offline fallback onto it, and teach currencyDecimals about
the newly reachable zero- and three-decimal currencies. A currenciesWith
helper keeps a previously saved (now-archived) selection selectable so it
isn't silently wiped.
* feat(plugins): tableContributor into the costs, packing and files views
Extends the shipped tableContributor hook to three more native views — no new
permission, no new attack surface: the same host-normalized, length-capped,
url-allowlisted (http/https/mailto), enum-bounded, fail-safe pipeline, just more
render sites.
- server: add costs/packing/files to the view-contributions whitelist
- client: widen the ViewName union + the api view type; render PluginCardFooter
keyed by entityId in the budget category table (a colSpan footer row per item),
the packing category group (footer after each item row, drag untouched) and the
files list (footer after each row)
A currency plugin can now drop a converted-amount column onto a cost row, a
receipts plugin a 'view receipt' action onto a file, etc. Controller test asserts
the three new views are accepted; both wikis updated.
* fix(pdf): repeat day header on overflowing itinerary export pages (#1471)
* feat(plugins): map-marker provider hook — plugins can overlay trip-map markers
New declarative provider hook `mapMarkerProvider` (#587 "show bookings on map",
the single most-requested contribution class, with zero contribution point until
now):
- hook:map-marker-provider permission + MapMarkerProvider/MapMarkerContribution SDK
types + HOOK_PERMISSION wiring
- GET /api/map-markers/:tripId (MapMarkersController) mirrors the view-contributions
hardening: membership-gated, providers invoked host->plugin on a 5s timeout,
fail-safe. Every field normalized server-side — coordinates range-checked
(-90..90 / -180..180), strings String-coerced + length-capped, icon/tone enum-
whitelisted, popup url http/https/mailto only (a javascript:/data: url would be
click-XSS), marker count capped at 200 per plugin
- client: PluginMapMarkers layer renders the markers as plain Leaflet Marker+Popup
inside the trip map; plugin JS NEVER runs on the map canvas, every value is
host-vetted data. Threaded tripId through MapView; fail-safe fetch
Declarative-only by design, mirroring placeDetailProvider/tableContributor. Perm
label (22 locales), controller hardening test, both wikis.
* feat(plugins): show page plugins in the mobile bottom nav
Page plugins were reachable from the desktop nav pill (Navbar) but not the mobile
tab bar — you had to type /plugins/:id. BottomNav now reads page plugins from the
plugin store and appends them the same way global addons are, mirroring Navbar.
One-file client nav wiring; no new capability surface.
* feat(plugins): per-user plugin settings form + ctx.settings runtime read
Users can now enter their own per-plugin config (an API key, a preference) —
the prerequisite for almost every real integration, previously unreachable
(scope:'user' settings were only listed read-only in the admin panel).
- migration: plugin_user_config (plugin_id, user_id, config JSON) — each user's
own values, separate from the admin-owned instance plugins.config
- PluginsService.getUserConfig / updateUserConfig / getUserConfigDecrypted +
readUserSettingDecrypted: secrets encrypted at rest (apiKeyCrypto), masked to
the client, an unchanged secret (the mask) keeps its stored ciphertext, and only
DECLARED scope:'user' keys are ever stored
- GET/POST /api/plugin-settings/:id (PluginUserSettingsController) — its own
user-gated path (not the admin surface, not the /:id/* proxy), JwtAuthGuard only,
scoped to the acting user
- runtime: ctx.settings.get(key) -> the acting user's decrypted value (unconditional
RPC, not sensitive cross-tenant; userless job/onLoad gets undefined)
- client: a Plugins tab in Settings host-renders each active plugin's scope:'user'
fields as an editable form (secrets write-only), reusing the declarative field
shape — no plugin markup executes
i18n (22 locales), wiki, rpc-host + service + masking/encryption tests.
* fix(journey): keep skeleton suggestions in sync with linked trip places (#1473)
Journey skeleton suggestions mirror a linked trip's day-assigned places, but
sync relied on scattered per-event hooks that several assignment mutation paths
never called: unassign, move and time-change fired nothing, no remove-on-unassign
capability existed, and every MCP assignment tool synced nothing. Skeletons drifted
from the trip.
Add an idempotent reconcileTripSkeletons(tripId) that re-mirrors the trip's
day-assigned places onto every linked journey (add missing skeletons, refresh
date/time/location on move, remove skeletons for unassigned places; filled entries
are detached + noted, never destroyed). Call it from every REST assignment handler
and MCP assignment tool, and fire onPlaceDeleted on single MCP delete_place for
parity. Extract a shared insertSkeletonEntry helper.
* fix(memories): drop hidden Immich assets so Live Photo motion parts don't show a broken thumbnail (#1474)
* fix(transit): anchor arrive-by search time to the destination timezone (#1479)
* feat(plugins): host-brokered OAuth client + trustworthy inbound webhooks
Two integration primitives where the host owns the sensitive part.
Trustworthy webhooks:
- auth:false routes now receive req.headers, but ONLY an explicit, credential-free
allowlist (the common provider signature/event headers — stripe-signature,
x-hub-signature-256, svix-*, x-gitlab-event, …). Cookie/Authorization/X-Socket-Id
and every session/forwarded-auth header are stripped; authenticated routes get {}.
A plugin can finally verify a provider signature without any way to leak a session.
Host-brokered outbound OAuth (oauth:client):
- the HOST runs the whole flow — authorize -> callback -> token exchange -> refresh —
with PKCE + single-use, user-bound, TTL'd state, and HOLDS the tokens. The client
secret + refresh token never leave the host; the plugin only triggers connect and
reads a short-lived access token via ctx.oauth.getAccessToken() for the acting user.
- provider config (authorize/token url + scopes + client id/secret) is the plugin's
admin-owned instance settings; endpoints must be https (SSRF backstop, private/local
hosts refused). Tokens per-user + encrypted at rest (apiKeyCrypto).
- GET/POST /api/plugin-oauth/:id/{status,connect,callback,disconnect} — JwtAuth-gated,
the callback always redirects to an in-app /settings path (never leaks an error).
- Settings -> Plugins gains a Connect/Disconnect control per configured plugin.
migration: plugin_oauth_tokens + plugin_oauth_state. Perm labels + form strings
(22 locales), both wikis, service (PKCE/state/exchange/refresh/encrypt) + controller
+ proxy header-allowlist + rpc-host gating + create-rpc-host wiring tests.
* fix(navbar): re-measure sliding tab pill after font load and resize (#1481)
The active tab pill was measured once in a layout effect keyed only on activeTab, so on a hard reload it captured the active (bold) label's width against fallback-font metrics and never re-ran when the web font swapped in, leaving the pill slightly offset.
Re-measure after document.fonts.ready resolves and on ResizeObserver changes (container + active button), with an idempotent state update to avoid redundant renders.
* fix(collections): keep the Add-place button reachable after the first save
On a wide/desktop layout the collection toolbar (which hosts the Add
button) was gated on !mapOverlay, so it unmounted as soon as the list
gained its first place with coordinates — leaving only an easy-to-miss
"+" in the map overlay. Keep the toolbar rendered whenever the user can
add a place, and drop the now-redundant map-overlay Add button so there
is a single, predictable Add affordance in every state.
Fixes#1485
* feat(plugins): days + accommodations reads/writes, endpoints on the reservation write path
Community feedback on the 3.2.1 plugin surface: a plugin could write days but
never list them (no way to learn day ids), day_accommodations had no surface at
all, and trips.getReservations was the one reservation read that dropped the
endpoints/day_positions hydration.
- trips.getDays / trips.getAccommodations under db:read:trips (tripRead gate),
wired to the same dayService lists the REST GETs use
- trips.getReservations now returns the hydrated REST-parity list (endpoints,
day_positions, joins, normalized accommodation_id) - strict superset
- new db:write:accommodations scope: ctx.accommodations create/update/delete
gated by day_edit like the accommodations REST path, with the partner-hotel
reservation + delete cascade and broadcasts intact
- reservation create/update pin the endpoints shape up front (BadParams instead
of a mid-transaction NOT-NULL or a silently dropped row)
- perm label in all 22 locales, consent PERM_KEYS, wiki tables
* feat(plugins): day-detail widget slot in the day panel
Widgets can now mount inside the trip planner's day panel
(capabilities.widget.slot: 'day-detail'), scoped to the open day via a dayId in
trek:context - the same pattern as the place-detail slot. Covers the requested
per-day plugin content (logistics, outfit planning, live flight status) without
a new plugin type. Day-detail widgets stay off the dashboard, the consent panel
labels the slot in all 22 locales.
* feat(plugins): let the frame CSP serve a plugin's own static assets
The sandboxed frame runs at an opaque origin, so script-src 'self' never
matched and a plugin's own <script src>/<link> files were blocked - authors had
to inline entire React builds into index.html. Add a scheme-less host-source
pinned to the plugin's own /plugin-frame/<id>/ path (charset-checked Host +
plugin id so a stray token can't widen the policy; malformed Host falls back to
inline-only). Multi-file client builds now load as-is; remote hosts stay
blocked, so script URLs remain useless as an egress channel.
* fix(plugin-sdk): catch the package up to the server capability surface
The npm SDK's validator still knew only the 3.2.1 permission set, so
'trek-plugin-sdk validate' (and pack/publish, which run it) hard-rejected any
manifest using the newer scopes - db:write:reservations, notify:send,
hook:map-marker-provider and 25 more. Sync KNOWN_PERMISSIONS with the server
envelope (48 entries), mirror the full PluginContext (reservations,
accommodations, notify/ai/oauth/settings, packing writes + bags, file writes,
collab, tags/todos/daynotes/collections/atlas/vacay/journal, weather,
categories), type the tableContributor/mapMarkerProvider hooks + the
entity/entityId event hint, accept the day-detail widget slot, and extend
createMockHost so plugin unit tests can exercise all of it.
* feat(plugins): grant-scoped entity snapshots on core events
An events:subscribe handler so far learned only WHICH entity changed - useful
for cache busting, useless for reacting to content, and the userless handler
can't refetch. Now the broadcast tap derives a whitelisted field snapshot of
the changed entity and the supervisor attaches it per plugin, only where the
granted set holds the family's matching db:read:* permission (trips family ->
db:read:trips, budget -> db:read:costs, packing -> db:read:packing, dayNote ->
db:read:daynotes, file -> db:read:files). No acting user is ever synthesized.
The whitelists are explicit per family, so user ids (owner/paid_by/uploaded_by/
participants/members), trips.feed_token and future migration columns never
travel; a private packing item (#858) yields no snapshot at all because its
core broadcast is owner-scoped; deletes/reorders/bulk ops carry none.
* feat(plugins): pdf-section, atlas-layer and journal-entry provider hooks
Three more declarative provider surfaces in the map-marker mould - plugins
return data specs, the host normalizes, caps and renders; a slow or failing
provider contributes nothing:
- hook:pdf-section-provider: sections (title + paragraphs + a simple table)
appended to the trip PDF export, escaped into the same HTML/print pipeline
as the core content
- hook:atlas-layer-provider: per-user country tint layers on the Atlas map
(ISO 3166-1 alpha-2 codes only, tone-whitelisted, non-interactive pane so
mark/unmark clicks keep working)
- hook:journal-entry-provider: extra rows on a journal entry card, gated by
the same journey access check as the journal routes + the Journey addon
Permission labels in all 22 locales, consent PERM_KEYS, SDK types + manifest
validator in both SDK copies, wiki tables, per-controller hardening tests.
* feat(plugins): trip-page plugins can replace core planner tabs and pick their spot
A trip-page plugin that takes over a core surface (a transit planner
superseding Transports, a costs plugin superseding the budget tab) had to sit
awkwardly next to the tab it replaces. capabilities.tripPage now names the
core tabs to hide while the plugin is active - whitelisted (transports,
buchungen, listen, finanzplan, dateien, collab), 'plan' deliberately not
replaceable, and the tabs return the moment the plugin is deactivated - plus
an optional 0-based position for the plugin's own tab. The feed re-validates
the values out of the DB blob so a hand-edited row can't hide anything else,
the admin list chips a replacing plugin (all 22 locales), and a saved session
tab that got replaced falls back to the plan view.
Also fixes the plugins feed dropping the day-detail widget slot to 'sidebar',
which would have mounted a day-panel widget on the dashboard.
* fix(plugins): audit follow-ups — normalization, secret cleanup, cron leak, slot filter
Adversarial audit of the whole plugin PR surfaced 12 confirmed issues; this
addresses them:
- place-details provider was the ONE hook controller with no normalization: a
plugin's href/label/value went to the client raw and unbounded. Now normalized
like journal-entry-rows (safeUrl http/https/mailto, length + count caps).
- trip-warnings capped message length + per-provider count (was unbounded).
- uninstall(deleteData) now also purges plugin_user_config, plugin_oauth_tokens,
plugin_oauth_state, plugin_meta_migrations and the capability audit — encrypted
per-user API keys + OAuth refresh tokens no longer survive a 'delete all data'
and get silently re-adopted on a same-id reinstall.
- supervisor: a crash-restart cycle leaked the dead child's node-cron tasks and
re-scheduled fresh ones, so a job fired N+1 times per tick after N crashes.
onExit now stops them, mirroring kill().
- dashboard sidebar no longer mounts place-detail/day-detail widgets (they belong
in the planner panels).
- reservation endpoint validation relaxed to match the 3.2.1 service: a coord-less
endpoint is accepted and dropped downstream instead of BadParams (no breaking
change), while a bad role/non-string still rejects up front.
- a replaced core tab reached by programmatic nav now falls back to the plan view.
- trips.update caps title/description like the places path; plugin-db guard bans
load_extension as defense-in-depth.
- wiki: event snapshots, string-typed context ids, dayId in the payload, the live
provider hooks and the costs update/delete grant are now documented correctly.
* feat(plugins): phase-0 lifecycle hardening + per-plugin RPC rate limit
Operational-readiness fixes from the completeness audit:
- Re-activation after a failure worked again: a plugin left in 'error' state by
a load-failure or crash-auto-disable stayed in the running map, so the admin's
'enable' button was a silent no-op. activate() now replaces a dead entry.
- Per-plugin RPC rate limit at the dispatch boundary: every ctx.* call runs
synchronously on the host thread, so a plugin in a tight loop could freeze the
whole instance (and the reap sweep). A token bucket (generous burst) + an
in-flight cap now throttle a runaway plugin with a retryable HOST_ERROR; a
legitimate plugin never notices.
- plugin_error_log retention (500 rows/plugin) so a crash-looper can't grow
trek.db without bound; the crash-timestamp array is trimmed to its window too.
- TREK_PLUGIN_PERMISSIONS=off now logs a loud one-time warning that the OS
permission jail is disabled.
* feat(plugins): read symmetry + broker — collab/journal/atlas reads, file content, trip create, rates
The plugin API leaned write-heavy: collab and journal could be written but not
read, files listed but not read, and there was no way to create a trip or see
exchange rates. This closes those gaps in the established RPC+gate pattern (zero
architecture risk), and it's what unlocks the importer + finance plugin classes:
- collab reads: ctx.collab.listNotes/listPolls/listMessages under a new
db:read:collab (membership + Collab addon, like the REST GETs)
- ctx.journal.getEntries(journeyId): a journey's entries, journey-access-checked,
under the existing db:read:journal
- ctx.atlas.bucketList(): the acting user's bucket list, under db:read:atlas
- ctx.files.getContent(tripId, fileId): a file's bytes as base64 under a NEW
db:read:files:content grant (reading a passport scan is more sensitive than its
filename), size-capped at 10MB before it crosses the IPC pipe, trashed files
refused
- ctx.trips.create(input): a new trip owned by the acting user, gated by the app's
trip_create right + a bound user — the capability importers need
- ctx.rates.get(base): cached currency exchange rates, tenant-free like weather
Also caps trips.update title/description like the places path, and the plugin-db
guard now bans load_extension (defense-in-depth). SDK, mock-host, i18n (22
locales), consent labels and the wikis are all in lockstep.
* feat(plugins): deeper integration + user-facing activity transparency
Wave 2 of the completeness work — richer extension points, deeper metadata, and
the transparency that makes the broad read grants accountable:
- db:meta now attaches to reservations + accommodations too (not just
trip/place/day), gated by reservation_edit / day_edit respectively — the
natural home for an external-id mapping (AirTrail/calendar/booking-import sync)
without forking the core schema.
- reservation-detail widget slot: a widget can mount on a booking card, scoped to
the open reservation via reservationId in trek:context (the place-detail /
day-detail pattern, third instance).
- tableContributor gains the transports + todos views, so a plugin can add
host-rendered columns/actions there too.
- User activity log: GET /api/plugin-activity + a Settings → Plugins panel showing
every host-mediated action a plugin took bound to the signed-in user, across all
plugins, newest first — the user-facing half of the hash-chained audit. This is
what legitimizes the deliberately broad read grants: not just the admin, the
person whose data is read can see what was done in their name.
- DX: the local dev server now binds a default acting user, so the canonical
ctx.trips.getPlaces(tripId) call works locally instead of failing RESOURCE_
FORBIDDEN; the create scaffold drops the dead manifest routes[] / capabilities.nav
fields the host ignores.
SDK, i18n (22 locales), consent labels and the wikis are all in lockstep.
* fix(memories): load Immich album photos on Immich v3
Immich v3 removed the `assets` property from AlbumResponseDto, so
`GET /api/albums/:id` no longer carries album contents. TREK read album
photos from that property, which now parses as undefined and degrades to
an empty array — hence "No photos yet" in the Journey gallery picker even
though the album header shows the right count (that count comes from
`GET /api/albums` -> assetCount, which v3 still returns).
Two call sites read the removed property. Besides getAlbumPhotos (the
reported bug), syncAlbumAssets failed silently on v3: it reported
`success: true, added: 0` while syncing nothing.
Fetch album contents via an `albumIds`-filtered `POST /api/search/metadata`
when `assets` is absent, and feature-detect rather than probe a version.
The two paths are not interchangeable: on v2, searchMetadata
unconditionally scopes results to `[self, ...partners]`
(`asset.ownerId = ANY(userIds)`), so an albumIds search against an album
shared by a non-partner returns nothing. v3 added an albumIds branch that
checks AlbumRead and skips that owner filter. v2 also hard-defaults
`visibility` to `timeline`, dropping archived assets. So v2 must keep
reading the album detail body, which this preserves exactly.
`withExif: true` is required on the search path: it has no default and
gates an inner join, so without it Immich omits `exifInfo` entirely and
every photo's city/country goes null.
The existing test mock returned an album detail body *with* `assets` — it
encoded the v2 assumption, which is why this shipped green. It now models
v3 by default, with explicit v2 coverage asserting no search call is made.
Fixes#1492
* feat(plugins): daily AI/notify budgets, runtime scheduler & reliable event redelivery
Per-plugin daily caps on ai.complete/ai.extract and notify.send (defaults
200 / 100, overridable via TREK_PLUGIN_AI_PER_DAY / TREK_PLUGIN_NOTIFY_PER_DAY),
seeded from the capability audit so a mid-day restart resumes the count instead
of resetting it. Surfaced at GET /plugins/:id/budget.
ctx.scheduler (at / in / every / cancel): persistent, userless timers that
survive restarts and fire a scheduled() handler, riding the existing jobs:run
grant so no new consent or admin setup is needed. Backed by
plugin_scheduled_tasks, swept every 30s, capped at 100 tasks/plugin with an 8 KB
payload and a 60s recurring floor; rows are removed on uninstall.
Core events that fire while a subscriber is mid-restart are now held in a
bounded in-memory buffer (200/plugin, 15 min TTL) and replayed once it goes
active again, with the events:subscribe grant and snapshot gating re-evaluated
at replay time so nothing leaks if a grant was revoked while the plugin was down.
* feat(plugins): GDPR data-subject rights — durable per-plugin erasure + export
New hook:user-data grant with two userless lifecycle handlers a plugin can put
on its definition: deleteUserData and exportUserData. Neither carries an acting
user — the plugin only learns the userId and touches its own db — so the grant
reads nothing from core data; it exists purely so a plugin can honour a GDPR
erasure or data-access request.
When a TREK account is deleted (admin or self-service), every installed plugin
holding the grant gets a row in a new durable erasure queue and its
deleteUserData runs on the next sweep, retried until it ACKs — so erasure
survives the plugin being offline or the server restarting. The core deletion
path notifies the runtime through a dependency-free relay (like the event sink),
keeping the auth/admin services decoupled from the plugins layer, and a plugin
bookkeeping error can never fail the account deletion.
Portability is served by GET /api/admin/plugins/user-data/:userId/export, which
fans exportUserData out to the active granted plugins and aggregates what each
holds about the user. Queue rows are purged on uninstall; the grant is labelled
in all 22 locales.
* feat(plugins): atomic ctx.db.tx for consistent multi-write on a plugin's own db
Plugins could already query/exec/migrate their own SQLite file, but a multi-step
write (move an item between tables, decrement one row and increment another) had
no way to be atomic. db.tx([{sql, args?}, …]) runs up to 100 statements in a
single transaction — all commit or all roll back — and reads within the batch see
its own earlier writes, so read-modify-write is safe. Each op is one statement:
a read returns { rows }, a write { changes }. The same guard (no ATTACH/PRAGMA/
RECURSIVE, size + row caps) applies to every statement in the batch.
* fix(memories): filter hidden Immich assets at the source, not just the picker
#1474 has the same root cause as #1492: the Immich v3 migration. On v2,
searchAssetBuilder hard-defaulted metadata search to `timeline` visibility
(`visibility = options.visibility ?? Timeline`), so hidden Live Photo
motion parts could never come back from a search. v3 defaults to any
visibility except `locked`, so they do — which is why the reporter is on
Immich 3.0.1 and why the bug never appeared before.
Ask for `visibility: 'timeline'` explicitly on the search path. That
restores v2 semantics on both versions and stops hidden assets crossing
the wire, which also fixes a pagination wart: a full page half-made of
motion parts previously rendered as a half-empty page, because hasMore
counts the raw page length while the filter shrinks the rendered set.
The client-side filter was display-only, applied in searchPhotos and
getAlbumPhotos — both picker-listing paths. Nothing guarded persistence
or rendering: getOrCreateTrekPhoto stores any id it is handed, pipeAsset
forwards Immich's 400/404 verbatim, and the photo grid is a plain <img>
with no onError. So syncAlbumAssets, which filtered `type === 'IMAGE'`
only, could persist a hidden IMAGE as a permanently broken tile. It now
applies the same guard, extracted as isVisibleAsset().
Albums keep their filter rather than requesting `timeline` visibility:
albums legitimately contain archived assets, and both the v2 album body
and the v3 album search return them.
Does not address tiles already persisted before this — those still render
broken and need a separate fix.
Refs #1474
* docs(memories): correct Immich version boundaries in the hidden-asset comments
Verified against the v1.120.0 → v3.0.0 OpenAPI specs and server source. The
previous comments said "Immich v2 hard-defaulted metadata search to timeline
visibility". That is true only for 1.133–1.144.
- `visibility` was added in 1.133.0. Before that, searchAssetBuilder applied
`.$if(options.isVisible !== undefined, ...)` with no default, so pre-1.133
servers returned hidden assets too. #1474 was therefore not purely a v3
regression.
- Those servers strip the `visibility: 'timeline'` filter rather than
rejecting it: Immich validates with `whitelist: true` and no
`forbidNonWhitelisted`. So the request stays valid, the filter is a no-op,
and isVisibleAsset() is the ONLY guard there. Say so, so it does not get
removed later as redundant.
- `albumIds` only exists from 1.135.0. Because unknown properties are stripped,
an albumIds search against an older server would silently drop the album
filter and return the entire library as the album's contents. Feature
detection on `assets` (present through 1.144.1, absent on v3) makes that
unreachable; a version probe with a wrong boundary would not.
Also cite Immich's own enum, which documents AssetVisibility.Hidden as
"Video part of the LivePhotos and MotionPhotos".
Comments only — no behavior change.
* feat(plugins): dashboard trip-card badges + a mock-host driver for plugin tests
Two additions that round out the plugin platform's breadth and its authoring DX.
tripCardProvider hook (hook:trip-card-provider): a plugin returns small declarative
badges for the dashboard trip cards. The dashboard fetches all visible cards in one
call; the host access-checks every tripId for the acting user, bounds each field
(label/value length, enum tone, http/https/mailto-only url), caps the count and drops
any badge for a card that wasn't requested — plugin JS never runs on the dashboard.
Rendered as text chips under the card meta; labelled + gated in all 22 locales.
createMockHost now exposes run(def) — the other half of a plugin unit test. Where the
ctx recorders capture what a plugin read, run() fires its own entry points (route, job,
scheduled, event, plugin-event, deleteUserData, exportUserData, provider hooks) against
the same mock ctx, and host.scheduled surfaces the timers it armed. A handler the plugin
didn't declare throws a clear error instead of a silent no-op.
* feat(plugins): include plugin data + code in backups, applied on restart
A TREK backup archived travel.db + uploads + the encryption key, but each plugin's
own SQLite file — the ONLY copy of the user data it holds — and its installed code
lived in separate trees that were never captured, so a restore left the plugins rows
with no data or code behind them.
createBackup now adds plugins-data/ (each plugin's db + WAL sidecars, so SQLite
recovers a consistent snapshot) and plugins-code/ (skipping dev-links by realpath, so
an author's linked source is never bundled). Restore can't swap those live — the
runtime holds each plugin db open — so it STAGES the extracted trees beside the live
ones and the runtime swaps them in at the next boot, before it opens anything. Same
"applies on restart" model the bundled encryption key already uses: no plugin quiesce,
no swap under open handles, no new admin setup. Older archives without the trees restore
exactly as before.
* fix(plugins): audit — runtime robustness, security & data-lifecycle fixes
Fixes from an adversarial audit of the plugin system, host/runtime side:
Robustness:
- getPluginDataDb recreated a handle a terminal-failure dispose had closed but
left cached, so a re-enabled plugin's db:own threw on every call — recreate
when the cached handle is shut.
- ctx.ws.broadcast* now carry _inv, so the host can bind the acting user (the
capability was silently refused, i.e. dead, without it).
- ctx.events.emit swallows a rejected emit instead of crashing the child into a
terminal 'error'; an uncaught throw AFTER activation is treated as a crash
(restart with backoff), not a load failure.
- A crash-respawned child gets the same activation deadline as a first activation
and the buffered-event queue is cleared on the timeout path, so a hung onLoad
after a crash can't peg a core and orphan events forever.
- Expired buffered events are pruned by the reaper, not only at flush; the
scheduler + erasure sweeps scope their LIMIT window to ACTIVE plugins so a
backlog for inactive plugins can't starve deliverable work.
Security / integrity:
- Unix-domain-socket / named-pipe connects are refused by default in the egress
guard (a host-local pivot to docker.sock / DB sockets), under the same policy
as private IPs.
- db.tx refuses transaction-control statements (a raw COMMIT would break its
atomicity) and caps rows across the WHOLE batch, not per statement.
- plugin_capability_audit is retention-capped per plugin (chain-safe: retained
rows stay self-verifying), so it can't grow unbounded in the shared db.
- A cap of 0 in TREK_PLUGIN_AI_PER_DAY / _NOTIFY_PER_DAY now disables the broker
instead of falling back to the default.
GDPR data lifecycle:
- Account deletion now erases host-side per-user plugin tables (config, OAuth
tokens/state) and enqueues the own-db erasure from the CORE path, so it works
even when the runtime is disabled or pre-boot; guest deletion does the same.
- uninstall keeps a pending erasure when data is retained (deleteData=false);
erasure delivery is no longer grant-re-checked (a queued erasure is a duty);
export flags installed-but-inactive plugins as pending instead of omitting them.
Backup/restore:
- Plugin DBs are WAL-checkpointed before archiving (no torn/stale snapshots).
- Restore applies the staged trees immediately by quiescing the plugins (no
unbounded gap where a later unrelated restart would revert diverged data);
the swap is content-level (safe on a volume-mounted root) and preserves
dev-links; the decompressed-size cap is operator-raisable.
* fix(plugins): audit — hook-output hardening, dashboard slot & mock-host parity
- Map-marker and atlas-layer tones were validated on String(tone) but emitted
raw, so a non-string tone (an object with a matching toString) slipped through
and crashed the client that renders it — check the raw value against the enum.
- View-contribution column/action caps are now PER ENTITY, not per view, so a
plugin's columns no longer vanish from every table row past the first 20; the
dashboard trip-card badge cap is per card (≥ one on every visible card).
- A reservation-detail widget no longer also renders as a context-free dashboard
sidebar card (the inline filter was missing that slot).
- mock-host matches the real host: it ignores asUserId on trip reads (bind the
acting user), throws on a wrong user-scope notify target instead of coercing,
enforces the scheduler caps, and detects RETURNING as a read in db.tx — so a
passing author test can't hide a production RESOURCE_FORBIDDEN.
* feat(plugins): full ctx parity in the dev server + fire jobs/events/hooks locally
The trek-plugin dev server injected only ~6 of the ~35 ctx areas, so any plugin
touching ctx.costs/packing/files/notify/ai/settings/scheduler/meta/oauth/db.tx/…
hit a TypeError in local dev while the same code passed mock-host tests and worked
installed. It also could only exercise routes.
Delegate every non-db-own capability to a grant-enforcing mock host (the same one
unit tests use) while keeping the real node:sqlite for db:own and dev-native ws
capture + logging — so the whole surface works in dev with the exact production
permission rules. dev-fixtures.json now takes the createMockHost options shape, so
you can seed the full surface. New GET /__dev/fire/<kind>[/<name>][/<fn>] fires a
job, scheduled timer, event subscription, GDPR handler or provider hook against the
dev ctx, closing the "can't test non-routes locally" gap.
* feat(plugins): wire the photoProvider + calendarSource hooks to real core consumers
Both hooks were declared, typed and documented but NO core code ever invoked them,
so an author could build, mock-test and install a photo or calendar plugin that
silently did nothing. Give each a real consumer that fans out to it, exactly like
the other eight provider hooks:
- GET /api/plugin-photos/search (+ /sources, /item) aggregates photoProvider results
for the picker — {id, title?, thumbnailUrl, fullUrl, takenAt?}, thumbnail/full URLs
http/https-only (they become <img src>), per-source count capped, failing source
skipped.
- GET /api/plugin-calendar?start=&end= aggregates calendarSource events for the
signed-in user — {id, title, start, end, allDay} ISO, count capped, failing source
skipped, sensible default window.
Both run with the acting user bound. The SDK interfaces now pass ctx as the last arg
(so a source can reach ctx.settings/oauth/http for its backend), and the wiki marks
them live instead of "reserved — no core consumer".
* feat(plugins): close the create-heavy API asymmetries importers/sync hit
Core services implemented these but plugins had no path to them, so the flagship
importer/sync integrations hit real walls. Added, each reusing the EXISTING grant
(no new consent):
- ctx.trips.removeMember(tripId, userId) — reconcile DEPARTURES, not just additions
(db:write:members + member_manage). Never removes the owner (that would orphan the
trip); ownership transfer stays a separate deliberate action.
- ctx.journal.createJourney({title, subtitle?, trip_ids?}) / deleteJourney(journeyId)
— an importer can now bootstrap the journal it fills with entries and clean it up
(db:write:journal), instead of only appending to journals a human created first.
Wired end-to-end (envelope → rpc-host → create-rpc-host reusing tripService/
journeyService → both SDK copies → mock-host) and documented. (trips.delete needs its
own destructive permission + consent copy and collab edit/delete + collections.delete
remain — tracked as small follow-ups.)
* feat(plugins): strip emojis from plugin-rendered text so it matches TREK's lucide UI
Plugin authors (especially AI-generated ones) sprinkle emojis into the declarative
text TREK renders in its OWN chrome — hook contributions (badges, columns, warnings,
PDF sections, map-marker/atlas labels, journal rows, place details, trip-card badges,
calendar + photo titles) and notifications — which clashes with TREK's lucide-only icon
language.
A shared stripEmoji() removes emojis (incl. flag/ZWJ/variation-selector sequences) and
tidies the leftover whitespace, applied at the render boundary in every hook-contribution
normalizer and in notify.send — so no matter what a plugin returns, the text TREK draws
stays emoji-free. It does NOT touch a plugin's own sandboxed /ui frame (the author's to
design), and it leaves photo ids verbatim (they round-trip to getById). The validate CLI
warns when a manifest name/description contains emojis, nudging authors to the declarative
`icon` field (a lucide name) instead.
* fix(plugins): harden the restore-apply path — regressions from the backup/dev fix pass
A final audit of the fix pass caught three regressions clustered in the two newest
surfaces; the restore path could both crash the server and destroy data.
- CRITICAL: a restore quiesces plugins via supervisor.shutdownAll() AFTER closeDb(), but
shutdownAll killed children without first marking them stopped, so each child 'exit'
took the CRASH path and wrote crash-accounting rows into the now-closed core DB — the
throw escaped an EventEmitter listener as an uncaughtException and killed the whole
process mid-restore. shutdownAll now marks every entry stopped and drops it from
`running` BEFORE the kills (so onExit early-returns), and the onStatus/onLog DB hooks
are wrapped in try/catch (also covers the stderr→onLog path). This also stops a normal
shutdown from logging phantom "crashed" rows.
- HIGH: swapContents cleared live entries then MOVED staged ones in, so a crash mid-move
permanently deleted a plugin's only data copy (staging was already emptied, so a retry
couldn't restore it). It now COPIES each staged entry over the live one and only deletes
staging at the very end — `staged` stays the complete source of truth, making the whole
operation crash-idempotent.
- HIGH: the dev server lost the actingUserId=1 default in the mock-host refactor, so a
fresh scaffold refused every user-bound capability. Restored.
* fix(plugins): final-audit medium/low findings
- GDPR export flags an active plugin whose export errored/timed out as `pending`
instead of silently omitting it (collectUserExport now returns a discriminated
result), so a data-access export never reads complete while missing data.
- Account deletion also enqueues an erasure for plugins UNINSTALLED with retained
data (an orphan data dir) — a same-id reinstall now honours the deletion instead
of re-adopting the user's data forever.
- oauth.getToken returns null in a userless context (matching the SDK/mock contract)
instead of throwing RESOURCE_FORBIDDEN a background caller can't handle.
- Crash-backoff restart is identity-guarded (+ the timer is tracked and cleared like
the activation timer), so a disable + re-enable during the backoff window can no
longer respawn a ghost child from the replaced entry.
- db.tx transaction-control guard strips leading comments first, so `/* */COMMIT`
can't slip past the start-anchored check and break batch atomicity.
- createJournal inherits its cover only from a trip that was actually LINKED
(access-checked), closing a cross-tenant cover-image read on plugin + REST paths.
- trip-warnings drops a null array element instead of losing ALL of that provider's
warnings; plugin-activity floors a non-integer ?limit so it can't 500.
- The trek-plugin dev server binds loopback only and refuses cross-site requests to
its side-effectful /__dev/fire endpoints (it serves real routes + no-auth dev
actions).
* fix(plugins): clear no-misleading-character-class in the emoji stripper
The character class listed the ZWJ, variation selectors and combining keycap
marks as members, which eslint reads as an accidental combined grapheme and
rejected on CI. Pull the emoji glyphs out into Extended_Pictographic /
Regional_Indicator alternatives so only the joiner/selector code points stay in
the class, with a scoped disable where the rule still can't tell them apart.
While here, reset lastIndex before the /g regex is reused in hasEmoji() so a
second call can't resume mid-string and miss a leading emoji.
* fix(security): trip-scope note-file deletion and guard the LLM base URL
Two reported issues:
- deleteNoteFile only matched on the note id and file id, so a member of trip A
could delete a file attached to a note in trip B by guessing its id. Thread the
trip id through the service and controller and scope the delete to it, the way
every other collab operation already does.
- The LLM extraction clients fetched the user-configured base URL directly, so a
user could point it at the cloud-metadata endpoint (169.254.169.254) and read
the echoed error body. Route both clients through a new safeFetchLlm() that
blocks the link-local/metadata range while still allowing a local or LAN Ollama
(loopback and private ranges stay reachable), pinned to the resolved IP so a
hostname can't rebind to the metadata address after the check.
* fix(security): route every LLM client through the SSRF guard
The base-URL SSRF fix covered the openai-compatible and anthropic clients but
missed the native Ollama /api/chat client and the /api/tags + /api/pull model-
management calls, whic…
* fix(plugins): repair plain-HTTP egress and forward the private-egress opt-out
Two pre-existing bugs in the plugin egress guard, found by running a plugin
against a real service end to end.
1. Every plain-HTTP request a plugin made was refused, whatever host it had
declared. Node pre-normalises `net.connect()` args into an [options, cb]
array and passes THAT array as the single argument; undici's plain-HTTP
connector takes this path, its TLS connector does not. classifyConnect read
`host` off the array, got undefined, and fell back to 'localhost' — so a
fetch to a declared, public host was rejected with the nonsense message
"localhost is not in the plugin's declared hosts". It failed closed, so it
was never a security hole, and it went unnoticed because the only shipped
egress plugin uses HTTPS. unwrapConnectArgs() unwraps the normalised form
before anything reads host/path.
2. TREK_PLUGIN_ALLOW_PRIVATE_EGRESS could never have any effect. The guard that
reads it runs INSIDE the child, whose env is scrubbed to a four-entry
whitelist that never included it — so a documented setting (wiki/
Environment-Variables.md) was wired to nothing, and no plugin could reach a
self-hoster's LAN service no matter what the operator set. Forwarded only
when set, so the default stays the secure block-private policy.
Regression tests cover the normalised form in both directions: the real host is
now resolved, and an undeclared host, a private IP and a unix socket are all
still refused when passed that way.
* feat(notifications): let a plugin register a notification channel
TREK's four channels (in-app, email, webhook, ntfy) were a closed set:
notificationService.send() dispatched with four copy-pasted `if` blocks and no
provider abstraction, so a fifth channel meant editing eight files by hand. A
plugin could produce a notification via ctx.notify.send(), but never deliver
one.
A plugin now registers a channel with `hooks.notificationChannel` +
`hook:notification-channel` on a plain `type: 'integration'` — not a new manifest
type, so the TREK-Plugins registry schema and both its CI gates are untouched.
Core refactor
- New channel registry (services/notifications/): email/webhook/ntfy become
ExternalChannel providers wrapping the EXISTING send functions — no delivery
logic is rewritten, only relocated. In-app deliberately stays out: it writes
typed rows with scope/target/callbacks, not a rendered title+body, the same
line shared/ already draws with i18n/externalNotifications.
- The event text is now rendered once per recipient instead of once per channel.
- The channel set is open: NotifChannel becomes a string, the matrix is
registry-derived, and the UI columns are server-driven. The DB column was
already bare TEXT and the Zod contract already a string record — only the
TypeScript and the two UIs were ever closed.
The hook runs USERLESS. Every other hook is user-initiated, so actingUserId falls
out of the request; a notification is host-initiated for an ARBITRARY recipient,
so ctx.settings.get() would return undefined. The host resolves the recipient's
decrypted scope:'user' settings itself and passes them as an argument. That is
what lets a channel plugin be handed someone's push token WITHOUT being handed
the right to read their trips as them.
Enabling the plugin is the opt-in: a plugin channel is not gated on the admin's
`notification_channels` list. A built-in always exists in code and needs an
explicit switch; a plugin channel only exists because an admin enabled that
plugin. (Nothing could write a `plugin:` id into that CSV anyway, and the admin
toggle rebuilt it from three booleans, silently dropping anything else — so
requiring a second opt-in meant the channel could never be turned on at all.)
Also fixed, found while building this:
- Plugin settings keys were unvalidated, so a field named `__proto__` or
`constructor` resolved off Object.prototype: a REQUIRED field with such a name
reported as configured for every user who had configured nothing — enough, for
a channel, to be dispatched to everyone with no credentials. Keys are now
constrained at install and the config blob is parsed null-prototype, so it is
impossible even for an already-installed plugin.
- A `select` field's options were cast straight through, so the obvious
`["1","5"]` form rendered every dropdown entry BLANK (the client reads
value/label). Now coerced, and malformed options are rejected.
Also adds: operator-supplied egress hosts (a plugin talking to a self-hosted
service can't name the operator's host at publish time, so an admin adds it
post-install and the runtime re-spawns the child with the widened allow-list —
only for a plugin that DECLARED operatorEgress, and only an admin, never a user);
settings-page actions (a "Test connection" button, user-initiated so
ctx.settings.get() returns the clicking user's own value); and a Gotify-shaped
notification-channel template in the SDK.
Verified end to end against a real Gotify container, not just in tests.
* docs(wiki): document the plugin notification-channel surface
Covers the pieces added in the previous commits, in the pages a reader would
actually reach for:
- Plugins.md (the admin-facing page) had none of it: notification channels,
settings actions, and a full "Allowed hosts" section — including what
operator-supplied egress deliberately does NOT let anyone do.
- Plugin-Development.md: the notificationChannel hook (and why it is the one hook
with no acting user), settings-page actions, operatorEgress, and the manifest
reference rows.
- Plugin-Cookbook.md: a "become a notification channel" recipe and a
"Test connection button" recipe.
- Plugin-Permissions.md: hook:notification-channel, operatorEgress under the
outbound section, and settings actions under "not a permission".
- Notifications.md: plugin channels alongside the four built-ins.
* fix(sdk): allow empty egress if and only if operatorEgress is true
* ci: don't run repo-specific workflows on forks
Guard release, publish, wiki-deploy and issue/PR-triage workflows with a
`github.repository` check so they no-op in forks instead of failing or
acting on the fork's own issues, PRs, tags and registries.
Also skip the Docker Scout scan for pull requests from forks: Docker Hub
secrets are never exposed there, so the login step could not succeed.
Tests and lint stay ungated — they need no secrets and are the gate for
incoming fork PRs.
* feat(sdk): add missing methods in mock-host
* fix(airports): rebuild the json file
* fix(airports.json): add small airports too
* fix(public transit): only show public transit option when a trip has actual dates
* fix(plugins): reap a queued erasure only once the plugin's data is gone
The orphan reap deleted every queue row whose plugin had left the registry, but
uninstall(deleteData=false) removes the plugins row while deliberately keeping the
data dir AND the queued erasure so a same-id reinstall can still honour it. The reap
now deletes a row only when the plugin's data dir is actually gone; a deleteData=true
uninstall already clears the rows itself.
* fix(backup): snapshot the core DB and swap restores atomically
createBackup archived travel.db via the archiver's lazy live-file read, so a WAL
auto-checkpoint firing mid-stream could write a torn database into the zip. It now
VACUUM INTOs a point-in-time snapshot and archives that, the same guarantee plugin
DBs already get. restoreFromZip swapped the DB by unlink-then-copy, which on an
interrupted restore could leave no valid travel.db; it now copies to a temp file and
renames it into place (atomic), dropping the stale -wal/-shm sidecars first.
* fix(deploy): Recreate strategy for the SQLite volume, pin the root compose image
The Helm Deployment had no strategy, so the default RollingUpdate would start a second
pod holding the same ReadWriteOnce PVC before the old one exits — a Multi-Attach
deadlock or two writers on one SQLite file. Default to Recreate (overridable for
ReadWriteMany). The root docker-compose.yml pinned trek:dev, a tag no workflow builds,
so a clone-and-up at the release tag ran a stale image; pin it to :latest like the README.
* fix(security): re-validate LLM endpoint fetch redirects per hop (GHSA-fmq9)
safeFetchLlm left undici's default redirect:'follow', so a configured LLM
endpoint could 302 to http://169.254.169.254/ and reach cloud-metadata
credentials — the DNS pin does not cover an IP-literal redirect hop, since
net.connect skips the pinned lookup for a literal IP. Follow redirects
manually now, re-resolving/re-checking/re-pinning each hop (allowing LAN/
localhost as before). Also block the Alibaba metadata IPs directly.
* fix(plugins): throttle the plugin log channel to prevent host-thread DoS
The per-plugin RpcRateLimiter only guarded the ctx.* (req) channel; ctx.log.*,
stdout/stderr and unknown evt topics reached a synchronous INSERT+prune on the
host thread unthrottled, so a while(true) ctx.log.error(...) loop could freeze
the instance. Route every plugin-driven log path through a per-plugin log token
bucket; excess lines are dropped with a summary line on resume.
* fix(plugin-sdk): serve dev /ui frame at /ui/index.html so relative assets resolve (#1526)
The dev server embedded the plugin UI as <iframe src="/ui"> (no trailing
slash), so a multi-file build's relative asset URLs (./assets/x.js from Vite
base:'./') resolved against the origin root -> /assets/x.js -> 404, even though
the files are served at /ui/assets/*. The real host loads the frame at
/plugin-frame/<id>/index.html where the same relative URLs resolve correctly, so
dev now matches it by loading /ui/index.html (and also serves /ui/ as index.html).
* i18n: improve Russian translations (#1539)
* v3.4.0 (#1527)
* fix(plugins): unknown column
* fix(mcp): reuse MCP sessions instead of creating one per tool call
The /mcp CORS layer never set exposedHeaders, so Access-Control-Expose-Headers
was absent and browser-context MCP clients (Claude Desktop connectors,
Claude.ai, MCP Inspector) could not read Mcp-Session-Id off the initialize
response. Unable to echo it back, every request looked like a fresh initialize:
one McpServer and one session per tool call, until the per-user cap returned a
429 and the integration died until the container was restarted.
The idle sweep was not at fault — it expires on lastActivity with a 1h default,
so sessions born seconds apart are nowhere near expiry, hence 'cleaned 0'.
- expose Mcp-Session-Id, MCP-Protocol-Version and WWW-Authenticate
- evict a user's least-recently-active session at the cap rather than
refusing the request, so a client that cannot persist its session id (or a
proxy that strips the header) can never wedge the server
- close the McpServer/transport orphaned by every session-less non-initialize
POST, which was leaked: never mapped, never swept, never closed
- return the cap error as JSON-RPC so clients surface the real reason
- warn on session-less POSTs to make a header-stripping proxy diagnosable
* chore(deps): declare @modelcontextprotocol/sdk ^1.29.0
Matches the version already resolved in the lockfile; no dependency-tree change.
* docs(mcp): add the reverse proxy specs for MCP
* fix(plugins): stop the row ⋯ menu being clipped by the sidebar
The menu was an in-flow `absolute` div, and its ancestor (PageSidebar) is
`overflow-hidden` — which clips absolutely-positioned descendants regardless
of z-index. With enough plugins installed a row sits low enough that its menu
runs past the sidebar's bottom edge and gets chopped, taking Delete with it,
so the plugin could no longer be uninstalled from the UI.
Portal the menu to <body> and position it `fixed` against the ⋯ button,
flipping upward when the bottom is tight and re-anchoring on scroll/resize.
A fixed child of <body> has no overflow ancestor, so nothing can clip it.
Closes#1523
* feat(plugins): surface author-signature status and add a scoped re-trust override
TREK has always verified an author's Ed25519 signature and TOFU-pinned the key on
first install, but none of it was ever shown: a successfully-installed UNSIGNED
plugin looked identical to a signed one, and a signature-refused update left the
plugin quietly pinned at its old version with the reason dying in a toast.
Give the four refusal conditions machine-readable codes (SIGNATURE_MISSING /
_INCOMPLETE / _KEY_CHANGED / _INVALID), persist a refusal on the plugin row so the
admin list keeps showing it, and badge Signed/Unsigned in the list and in Discover.
Only SIGNATURE_KEY_CHANGED is overridable — an author can legitimately rotate a key;
a signature that does not verify means the bytes are not what the author signed, and
there is no story where waving that through is right. The override re-pins and
updates in ONE call (POST :id/retrust): a re-pin that waited for a follow-up /update
would leave the plugin pinned to a key no install had ever been verified against if
that second call never came. The artifact must still verify under the new key, so a
re-trust moves the pin from one verified key to another.
assertRetrustable re-derives the condition server-side, so the UI hiding the button
is a convenience, not the control, and it echoes back the full key the admin was
shown so a re-key since the dialog rendered is refused. The rotation is written to
the admin audit log with both fingerprints — after an incident, "which key did we
move from, and to what?" is the question a single key cannot answer.
* fix(plugins): let a plugin's frame reach the hosts an admin added for it
The frame's connect-src was built from the manifest's http:outbound grants alone, but
the child's egress guard is the UNION of those and the hosts an admin added after
install for an operatorEgress plugin (a self-hosted Gotify, an ntfy — hosts the author
cannot know in advance). So such a plugin WITH A UI could call the operator's host from
its server and was CSP-blocked in its own iframe.
Match the frame to the child. The admin consented to these hosts at install and the
child already reaches them, so this widens no trust boundary that isn't already crossed.
Both sources stay validated on the way in, and the interpolation filter is unchanged.
* fix(plugins): report a rejected events.emit on the plugin's own log stream
The host can reject an emit (an undeclared event name, a rate limit). The rejection must
not escape — a detached rejection crashes the child and terminally disables the plugin
over one bad emit — but swallowing it silently left an author with no way to discover
that `emits` was missing from their manifest. Surface it as a warning instead.
* feat(plugin-sdk): expose the raw request body for webhook signature checks
A webhook author must run their HMAC over the exact bytes the sender signed. `body` is
the PARSED value, and re-serializing it will not reproduce those bytes — key order,
whitespace and unicode escaping all differ — so the signature never matches. Document
`rawBodyBase64`, which the host already sets on auth:false routes.
* fix(plugin-sdk): refuse to overwrite a released artifact, and keep the packed zip
A released artifact is IMMUTABLE: the registry pins its sha256, so overwriting the bytes
of a release already in the registry breaks the checksum for everyone who installed that
version — they can no longer install or update it. The old code blanket-caught every
`gh release create` failure (auth, network, a bad repo) and turned it into a --clobber
upload. Probe for the release explicitly and refuse unless --force.
Also stop deleting plugin.zip on the way out. It is the exact bytes the release and the
entry's sha256 pin were computed from; a re-pack on another machine or SDK version can
differ (CRLF, walk order), so anyone re-running `entry`/`sign` afterwards must hash THAT
file, not a rebuild.
* fix(plugin-sdk): don't fail submit when the fork already has an upstream remote
`gh repo clone` of a fork may already have wired `upstream`, in which case a bare
`remote add` exits non-zero and took the whole submit down. Set it either way.
* fix(plugin-sdk): make the dev server behave like the host
Three ways `dev` lied to an author about how their plugin would run in production:
- ctx.db: one try/catch wrapped both the node:sqlite probe AND opening the database, so
an mkdir/permission failure silently degraded to the in-memory stub — which swallows
every write while reporting success. A db:own plugin "worked" in dev and persisted
nothing. Probe separately: fail loudly on a real error, degrade only on old Node, and
say plainly that the stub discards writes.
- notificationChannel: the host fires it with no acting user and hands the recipient's
decrypted settings in as a separate `config` argument — send(msg, config, ctx) /
test(config, ctx). Firing it like an ordinary hook passed `ctx` where `config` belongs,
so a channel plugin read its settings off ctx and was broken in production.
- /preview pinned tripId 42 while the scaffold seeds trip 1, so the widget's first
trek:invoke hit assertMember(42) and 500'd. Preview against a trip that exists.
* fix(plugin-sdk): reject unknown flags, wire up create's new flags, and add --help
`parse()` accepts any --x, so a flag a command does not read was silently dropped:
`create --template notification-channel` cheerfully scaffolded a blank plugin. Silently
ignoring an author's explicit instruction is worse than refusing it, so unknown flags are
now an error, and create actually forwards --template/--egress/--required-addons.
A bare --permissions used to split the string "true" into a permission literally named
`true`; listFlag() now treats a valueless flag as absent.
Since an unknown flag is now fatal, `--help` has to exist: it is intercepted before the
flag check and prints usage on stdout with exit 0.
* chore(plugins): drop three unused eslint-disable directives
The no-console rule is not enabled for these files, so the directives were dead and
eslint reports them as unused.
* fix(plugin-sdk): bring preflight back in step with the registry's gates
preflight exists to tell an author what TREK-Plugins' CI will say before they open the
PR. The registry now verifies author signatures, and preflight didn't — so it drifted
into the one failure mode it must never have: a false green. An author trusts a green.
- Verify the signature against the artifact bytes. preflight already downloads them for
the sha256 check, so this costs one call. Without it, signing with the wrong key (or
re-packing after signing) sails through and is caught at review.
- Check the signature SHAPE (checkSignatureShape): a key with no signed version, a
signature with no key, a malformed key or signature. TREK refuses to install a
half-signed entry, so such an entry is dead on arrival.
- Default apiVersion to 1 before comparing. It is OPTIONAL in the manifest — install/
manifest.ts and `entry` both default it — so a manifest that legally omits it was
failing preflight with "manifest apiVersion undefined != entry 1" while the registry
passed it. A false RED, which teaches authors to ignore the tool.
- Check requiredAddons/pluginDependencies parity, and operatorEgress without an
http:outbound grant.
The verifier is a port of the host's install/verify-signature.ts (the registry has its
own port); sign.ts's verifyArtifact only understands the bare key/signature pair the SDK
itself emits and cannot judge a minisign key. A test pins all three to the same verdicts.
* feat(plugins): enforce compatibility range
* chore: bump sdk version
* test(e2e): repair the trip-creation specs
create-trip and trip-planner have been failing for a while — long enough for
three separate UI changes to drift past them, which nothing caught because CI
runs vitest only and never invokes Playwright.
Each failure was masking the next:
- The release-notice modal greets a freshly seeded user and its backdrop
swallowed the click on .add-trip-card. Added a shared dismissSystemNotices()
helper (the X only shows on the notice's last page, so it has to page through
first).
- .modal-backdrop no longer exists — the class was namespaced to
.trek-modal-backdrop so content blockers stop hiding it.
- input[type=text].first() is no longer the Title field: the cover-image search
inputs now sit above it, so the specs were typing the trip name into the photo
search box and creating an untitled trip that never matched getByText(title).
* fix(planner): gate drag & drop on pointer type, not viewport width (#1432)
The 3.2.1 fix disabled drag on "mobile", but nothing in the client has ever
detected touch — "mobile" was inferred from viewport width, at four independent
breakpoints. A tablet is a coarse-pointer device at a *desktop* width, so an
iPad (820-1366px) fell on the wrong side of all four, which is why iPhone was
fixed and iPad was not:
- useTripPlanner's isMobile (<768px) is what disarmed `draggable`, so on iPad
rows stayed draggable and a scroll swipe became an HTML5 drag.
- TripPlannerPage hardcoded isMobile={false} on the desktop PlacesSidebar, so
its drop handlers and the drop-to-import overlay could never be disabled —
that overlay is the reported symptom.
- The arrow-button reorder fallback was revealed only below 767px, leaving iPad
with no drag *and* no fallback.
- touchDragPolyfill loaded drag-drop-touch at >=1024px, synthesising drags from
touchmove — on a landscape iPad that re-armed the very gesture hijack 3.2.1
removed.
Adds useIsTouch() ((pointer: coarse)) as a signal separate from isMobile: layout
stays width-driven, so the iPad keeps the desktop two-pane planner, while every
drag affordance is gated on isMobile || isTouch. The reorder arrows now show on
coarse pointers, and the polyfill loads only on hybrid laptops
((pointer: fine) and (any-pointer: coarse)), which also removes the #1440
phantom-dblclick map zoom on tablets.
Guarded by unit tests plus an e2e spec on real WebKit in an iPad Pro 11 context
— the engine matters, since every browser on iPadOS is WebKit underneath, which
is why the reporter hit this in all three they tried.
* test(planner): guard the day-plan reorder arrows against phantom clicks
The hover rule that reveals the reorder arrows had been dead since the
TypeScript migration: it targeted `.place-row:hover .reorder-btns`, and neither
class exists — the component renders `.reorder-buttons` inside an unclassed row.
The buttons sat at opacity:0 on desktop with nothing to reveal them.
That was not merely invisible. opacity:0 still hit-tests, so every itinerary row
and day note carried an invisible, fully clickable target that silently
reordered the trip:
opacity: "0", visibility: "visible", pointerEvents: "auto"
elementFromPoint(centre) -> BUTTON, inside .reorder-buttons
The repair — pointer-events: none while hidden, plus a working
.dp-row:hover/:focus-within reveal (focus-within so the buttons are also
reachable by keyboard, which the file's JS-hover pattern cannot do) — lives in
index.css and DayPlanSidebar.tsx. Both files also carry the #1432 drag gating,
so they went with that commit rather than being split mid-file; this commit is
the regression guard for them.
E2E rather than unit, because jsdom does not evaluate :hover. Against the old
CSS it fails on exactly the right assertion: "hidden arrows must not swallow
clicks" — expected false, received true.
* fix(pdf): keep the header gap on day-header overflow pages (#1531)
The repeated <thead> day header (#1471) carried no gap before the first
card on overflow pages: the 12px sat in .day-body's block-start padding,
which a fragmented box only paints on its first fragment. Move it to the
thead cell so it repeats with the header; the day's first page renders
pixel-identically.
* fix(budget): keep "no one paid yet" when editing an expense (#1533)
The ExpenseModal payer initializer fell back to the current user whenever the
edited item had no payer, so reopening an expense saved with "no one paid yet"
silently reselected "You" — and re-saving then recorded the current user as the
payer, corrupting the balances. An absent payer on an existing expense is a
deliberate value, so only a brand-new expense defaults to me.
* feat(sdk): support for plugin icons
* fix(planner): keep the places filter applied and visible across tab switches
The category and all/unplanned/tracks filters lived twice: a local copy in
the sidebar driving the checkboxes and the list, and a page-level copy
driving the map markers, synced only when a checkbox was clicked. Switching
planner tabs unmounts the sidebar, so the local copy reset to "All
Categories" while the markers stayed filtered — and the only way out was
toggling any category on and off again (#1541).
The filter now lives once in the trip store, next to selectedDayId: it
survives the Plan tab unmounting (and the mobile places sheet closing),
keeps both sidebar instances in agreement, and resets when another trip
loads.
* feat(airtrail): import connecting flights as one multi-leg booking
AirTrail flights always imported as separate single-leg reservations, so a
layover could not be expressed and the connection country ended up counted
as visited in Atlas (#1535).
The import picker now detects connection chains among the listed flights —
each leg departing from the airport the previous one landed at, onward
within 24 hours, and never back to the origin (an out-and-back is a return,
not a connection) — and offers to import each chain as one flight with
layover stops, on by default. The joined booking keeps per-leg airline,
flight number, times and seat in metadata.legs, files every leg on its own
trip day, and mirrors the first/last leg flat, exactly like the manual
multi-leg form. With the connection stored as a stop endpoint, the existing
Atlas role filter excludes the layover country on its own.
AirTrail has no multi-leg flight a joined booking could round-trip to, so
it imports detached from live sync, with every source flight id recorded in
metadata.airtrail_ids — the picker and the server-side dedupe both treat
those legs as imported, per leg, even across trip members. The server
re-validates each requested chain and falls back to individual imports when
it does not actually connect.
* fix(airtrail): stop syncing a booking once it grows extra stops
A linked flight that becomes multi-leg locally no longer matches the single
AirTrail flight it was imported from: pushing would rewrite that flight to
span the whole route, and the next pull would flatten the layover chain
back to a plain from/to. Both sync directions now detach the link instead —
the same state a joined import starts in, surfaced by the existing "Not
synced" badge.
TransportModal also carries metadata.airtrail_ids through re-saves (like it
already does for transit itineraries and day positions), so editing a
joined booking cannot cost it its import dedupe and get its legs re-offered
in the picker.
* fix(planner): default a new accommodation to checking out the next day
The hotel picker pre-filled "Apply to days" with the same day for check-in
and check-out — a stay that ends the day it begins. New accommodations now
default to the following day for check-out; the last trip day keeps the
same-day range, and editing still seeds from the stored range.
* docs(wiki): document the AirTrail import and connection joining
* fix(plugins): fold resolvePluginIcon into PluginIcon
pluginIcon.ts and PluginIcon.tsx resolve to the same file on the
case-insensitive filesystems dev checkouts commonly sit on (Windows,
macOS) — tsc sees both casings of one module and fails with TS1149 in
every importer. Keep the resolver and the component in one module.
* feat(sdk): add update verification
* chore: remove test files
* fix(sdk) rework the sdk helpers and DX/UX
* fix(sdk) harden dev environment
* fix(budget): add back the multi payer selection
* fix(map): stop MapLibre mouse rotation from reversing near mid-screen
Since MapLibre 5's camera rewrite, the right-button rotate handler flips
its sign whenever the cursor sits above a mid-screen line it derives by
re-projecting the map center. That line drifts with the bearing by a
fraction of a pixel, so inside the 100px band around the screen center a
steady horizontal drag lands alternately above and below it — every
processed movement reverses the previous one and the camera ping-pongs in
place instead of rotating (#1545). A real hand crossing the line mid-drag
flips the rotation direction outright. maplibre-gl 4.x rotated from plain
horizontal movement and had none of this.
Passing aroundCenter: false opts the handler out of the around-center
mode and restores the 4.x/mapbox-gl behaviour: horizontal drag rotates,
vertical drag pitches, in one continuous motion. Applied to all three GL
map builds (planner, journey, settings preview); mapbox-gl keeps its
options untouched.
* fix(budget): settle in the trip's real currency, not always EUR
The settlement route read `currency` off the row returned by canAccessTrip,
whose SELECT never included the column. A cast hid the mistake, so trip.currency
was always undefined and the settlement was told every trip is in EUR.
Balances are netted in the trip currency and converted to the display currency
once. With the trip mislabelled EUR, expenses in the trip's own currency still
cancelled out, but an expense booked in a foreign currency was divided by its
frozen rate into trip-currency units, then converted again as if those were
euros — inflating balances by the EUR/trip rate (~27x for a RUB trip with a USD
expense, #1543). MCP was unaffected: it reads the currency with its own SELECT.
Select the currency in canAccessTrip so the read is real. The settlement maths
was correct all along; it was simply being lied to about the base currency.
Fixes#1543
* feat(trips): let users set the trip currency, and rebase the budget when it changes
The trip currency is the base every expense and settle-up is netted against, but
the only picker for it lived in the legacy Budget addon panel — so on the Costs
panel a trip was stuck with whatever it was created as. Add the field to the trip
form, on create and on edit, gated on trip_edit. The REST and MCP write paths
already accepted `currency`; only the form was missing.
Changing it is not a rename, though: an expense's frozen `exchange_rate` is
"units of its currency per 1 trip currency", and `currency = NULL` means "the
trip's own", so both are relative to the outgoing base. Swapping it out from
under them redenominates the implicit rows (9 000 RUB becoming 9 000 EUR) and
leaves the frozen rates pointing at a currency the trip no longer uses — the same
mismatch that inflated #1543.
So rebaseTripCurrency() runs first, while the old currency is still on the row:
it pins the implicit rows to the outgoing currency and re-freezes every rate
against the incoming one, for expenses and settle-up transfers alike. No stored
amount is rewritten — each keeps the figure the user typed, in the currency they
typed it in, and its real-world value survives the switch.
Also covers the #1543 data as a settlement regression test.
* feat(budget): give settle-up payments their own currency
A transfer settling a shared bill can be made in any currency — paying a rouble
debt in euros is normal — and the server has stored `currency` + a frozen
`exchange_rate` on every transfer since #1445, re-freezing it on edit. The UI
just never let anyone choose one, so a payment silently inherited whatever the
viewer's display currency happened to be.
Add the picker to the payment modal, mirroring the expense modal, and reopen an
existing payment in the currency it was actually recorded in. The ledger row now
shows a foreign payment as `$30.00 -> 27,00 EUR` like a foreign expense does,
instead of stamping the display currency's symbol onto the raw stored number.
The Settle buttons on the suggested flows keep sending the display currency:
those amounts are computed in it.
* feat(settings): make the display currency optional, falling back to the trip's
Costs already resolved `default_currency || trip.currency || 'EUR'`, but the
setting could never actually be empty: the store seeded 'USD', so a user who had
never touched it silently had every trip converted into dollars, and the picker
offered no way to unset it.
Seed it empty and lead the picker with a "Trip currency" option, so an unset
preference means "show each trip in its own currency" instead of forcing them all
through one code. An explicit empty value persists and beats the admin-set
instance default — it is a deliberate choice, not an absence.
This also brings the public share's fallback to life: the share payload has
resolved sharer's currency -> trip currency since #1361, but the trip-currency
branch was unreachable while every owner had a currency forced on them.
Plugins are handed `formats.currency` as a concrete code, so PluginFrame now
resolves the same chain rather than passing an empty string through the bridge.
* docs(wiki): explain the three currencies and how they relate
Trip currency, expense currency and display currency answer three different
questions, and nothing said so: the trip currency wasn't documented at all (it
had no picker until now), and Budget-Tracking conflated the other two while
still claiming 47 currencies and a display currency that always came from
Settings.
Add a Currencies page as the one place they're defined together — the trip
currency as the accounting base, the expense currency as the receipt with its
rate frozen at entry, the display currency as presentation-only — plus what
happens when a trip's currency changes, which currency a public share renders
in, and what belongs to the Costs addon versus the trip itself.
Rewrite Budget-Tracking's currency section against it, document the currency
field in Creating-a-Trip and the display currency in Display-Settings (which
never mentioned it), and note the sharer-or-trip fallback in Public-Share-Links.
* feat(help): serve the in-app wiki from disk instead of fetching GitHub
The /help pages fetched their markdown from raw.githubusercontent.com at
runtime, so a self-hosted install was served docs from main rather than the
version it was actually running, and help was unusable without network access.
The wiki/ directory was in the repo the whole time; the bundled-snapshot
fallback the code reached for was gitignored and never populated by any build
step, so it was dead code.
Read wiki/ straight from disk instead. server/{src,dist}/services both sit
three levels under the repo root, so a single __dirname anchor resolves in dev,
a built source install, vitest and Docker with no copy or build step. GitHub is
kept strictly as a fallback for when the directory cannot be resolved, decided
once at load by probing for _Sidebar.md: a page missing from a present wiki is
a genuine 404, since falling back per-file would reintroduce the version skew
this removes.
Ship wiki/ in the image: .dockerignore excluded it outright, so the COPY alone
would have produced an image with no wiki, and the GitHub fallback would have
masked that at runtime. Add a real path-containment check on assets now that
the path becomes a filesystem read rather than a URL, and document
TREK_WIKI_DIR as an off-by-default escape hatch across the deployment configs.
* fix(plugins): serve frame assets root-relative and cross-origin loadable
res.sendFile(absolutePath) resolves against the rewritten req.url under
the Nest ExpressAdapter and 404s spuriously (files-download already
works around the same trap), which broke every plugin frame document.
And helmet's CORP: same-origin made the browser drop the opaque-origin
frame's own script/style subresources, so a multi-file plugin client
could never boot. Serve root-relative and mark frame responses
cross-origin — sandbox + per-plugin CSP stay the isolation boundary.
* feat(plugins): let a plugin ship its own settings page
A plugin that declares capabilities.settingsUi: true gets its
client/settings.html framed as a card under Settings -> Plugins — same
opaque-origin sandbox and postMessage bridge as its widget, sized via
trek:resize. Hosts that predate the flag strip it at install, so old
instances keep working and simply don't show the card.
* feat(map): open maps framed on their places (builds on #1393) (#1556)
* fix(map): fit MapLibre routes reliably
* fix(map): default planner map to world view
* fix(map): only await route geometry when a route is actually pending
The fit armed a pending route-refit slot on every fitKey change, even with no
route drawn, and only ever cleared it when a route arrived. So a route toggled
on much later — after the user had panned elsewhere — was mistaken for the
awaited geometry and yanked the camera back, and only on the first toggle.
Arm the slot only when a route is already on screen: updateRouteForDay lays down
straight lines in the same batch as the fit and upgrades them to real geometry a
moment later, so an empty route at fit time means none is coming.
* fix(collections): open the empty collection map on the world view
A collection with no mappable places centred on Paris, the same hardcoded
default this branch removes everywhere else.
* feat(map): open the map framed on its places
A trip in Japan opened on the world view at 0,0 and only then animated a fitBounds
flight across the planet — the hardcoded default was the map's answer to a question
its own places already answer.
Each renderer now derives its opening camera from the places it receives, at
construction: MapView for Leaflet, MapViewGL for MapLibre and Mapbox (whose zoom
runs one level below Leaflet's, measuring against a 512px world tile rather than
256px). Doing it at construction is what confines it to load — the map is built
once, and by then the trip's places are in hand. Nothing recomputes it afterwards,
so the camera stays where the user leaves it, and the opening fit stands down
rather than overruling the gentler zoom a lone place opens at.
A trip with no coordinates still falls back to the world view. Collections and the
public shared-trip page frame themselves the same way.
* refactor(settings): drop the default map centre and zoom
Nothing reads them now that every map frames itself on its own places, and a
home-city default was the wrong answer for the next trip on the other side of the
world. The style preview keeps a fixed location of its own: it needs a city to show
label density and 3D buildings, which open ocean cannot.
* fix(map): frame the map the way each renderer can actually draw
Two defects the unit tests missed and running the app exposed.
MapLibre and Mapbox opened on Null Island at zoom 2 regardless of the places: the
effect that mirrors an external centre prop onto the camera also ran on mount, so
it jumped straight to the default nobody passed and threw away the camera the map
had just been built with. It now only responds to actual changes, which is what it
was for. Leaflet was unaffected — its controller already guarded on the centre
changing.
A trip spanning Sydney, Reykjavik and Santiago lost Sydney's marker entirely. The
narrowest arc containing all three crosses the antimeridian, and framing there is
only sound on a renderer that repeats the world: MapLibre and Mapbox draw a marker
on whichever copy is nearest the camera, Leaflet draws one world and puts the
marker at its absolute position — off-screen. Leaflet now spans the long way round,
as L.latLngBounds would. The test that should have caught this wrapped the x-offset
in its own projection helper, quietly assuming behaviour Leaflet does not have; it
now models each renderer's real wrapping.
---------
Co-authored-by: Azalea <noreply@aza.moe>
* feat(mcp): expose public transit planning tools (#1558)
* feat(mcp): add public transit planning tools
* refactor(transit): reuse local time conversion
* fix(mcp): harden transit journey validation
* refactor(transit): centralize itinerary processing
* fix(mcp): return the transit itineraries the provider actually offers
search_transit_routes validated each itinerary leg's mode against
SCHEDULED_TRANSIT_MODES, but that constant is the request-side filter
whitelist — the modes a caller may ask for — not the modes MOTIS can return.
Its default TRANSIT mode expands to TRAM,FERRY,AIRPLANE,BUS,COACH,RAIL,ODM,
RIDE_SHARING,FUNICULAR,AERIAL_LIFT,OTHER, and street legs can be BIKE/CAR/
RENTAL. Any itinerary carrying one of those failed the parse and was dropped
by the flatMap, so the tool reported fewer routes than exist — or none at all.
Against the live provider, Trondheim → Ålesund returns 5 itineraries and 3 of
them contain an AIRPLANE leg, so the tool silently discarded them; the web app
shows all 5, because it treats mode as a free string and renders anything
non-WALK as a transit leg.
Accept any mode token on a leg and keep the existing "at least one non-WALK
leg" rule as the real gate, which restores parity with the web app. Everything
downstream keeps its mode !== 'WALK' semantics, so a journey created over MCP
is identical to one created in the app.
Dropping an itinerary is still possible when provider data fails the remaining
consistency rules, but it is indistinguishable from "no routes exist" — so
search_transit_routes now reports a `dropped` count alongside the results.
---------
Co-authored-by: Uzini <43294422+Uziniii@users.noreply.github.com>
* fix: show map poi search controls on mobile (#1555)
* fix(pdf): use the trip's actual currency instead of hardcoded EUR (#1519)
* fix(pdf): use the trip's actual currency instead of hardcoded EUR
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(i18n): drop hardcoded currency text from pdf.costLabel across locales
Remove redundant EUR/currency references from pdf.costLabel translations
across 21 locales now that the PDF export correctly renders amounts
with their actual trip currency via formatMoney().
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(pdf): drop hardcoded euro-shaped icon from place price chip
svgEuro rendered a fixed € glyph next to the price chip regardless of
the trip's actual currency, undermining the currency fix. Swap for a
currency-neutral coin icon.
---------
Co-authored-by: Nguyen Trong Binh <nguytb15@VN1N07HO1CD1015.local>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
* fix(trips): pin place prices when the trip currency changes
A place price with no currency of its own means "the trip's own currency" —
that is how the PDF export and the place chips read it since #1519. But
rebaseTripCurrency() only pinned budget_items and budget_settlements, so
switching a trip's currency silently redenominated every implicit place
price: a €15 museum on a trip moved to JPY started reading as ¥15. Same
class of mismatch as #1543, on the places surface.
Pin priced, currency-less places to the outgoing currency inside the same
transaction. No amount is rewritten — the figure the user typed keeps its
real-world value, it just stops being ambiguous about which unit it is in.
Bump updated_at as well: it doubles as the optimistic-concurrency token
(#1135), so a client holding the pre-switch row can no longer write the
pin away.
Document place prices in the wiki's currency model, and drop the now-stale
"in EUR" from the PDF export's estimated-cost stat.
* fix(budget): re-split expenses when a member is removed
* fix(plugins): derive plugin req.user.isAdmin from role, not is_admin
* chore(test): make jsdom Storage work under Node's native Web Storage globals
* fix: recompute vacay calendar when week start changes (#1554)
* feat(mobile): comprehensive mobile UI rewrite
Add a dedicated phone UI (<768px) under client/src/mobile/, mounted by
MobileShell while the desktop/tablet code paths stay untouched. Each page
delegates to an M-screen that reuses the existing hooks, stores and services.
Covers dashboard, trip planner (plan timeline, map, places, tabs and sheets),
vacay, atlas, journey, collections, settings, admin and notifications, with a
shared component kit (MSheet, MBottomNav, MGlassBar, MSegmented, MToggle, …)
and design tokens in mobile.css.
Empty states use an animated TREK mascot (MDancingTrek) with per-scene props.
Adds the mobile* i18n namespaces (mobileTrip/mobileJourney/mobileVacay/
mobileAtlas/mobileNav/mobileAdmin/mobileSettings/mobileCollections) across all
22 locales.
* fix(websocket): handle socket 'error' events to prevent crash on malformed frames (#1584)
* fix(packing): apply templates into the active list tab (#1581)
* Fix packing list readability on mobile
* Localize packing quantity label in overflow menu
* feat(mobile): accommodation & save-to-collection sheets, trip loading splash
Add a dedicated mobile accommodation sheet and a save-to-collection sheet,
plus an animated trip loading splash with stepped status copy. Make the vacay
day/company-holiday toggles optimistic so cells react instantly and roll back
on error. Wire up the "subscribe to all trips" calendar feed and cover-image
add on the new-trip sheet. Sync the new dashboard/trip i18n keys across all
22 locales.
* fix(atlas): stream boundary GeoJSON instead of caching parsed bundles (#1585)
* fix(atlas): stream boundary GeoJSON instead of caching parsed bundles
* fix(atlas): scan admin1 features in one pass to avoid O(n^2) build
createFeatureSplitter re-scanned each partial Feature from the start on
every gunzip chunk, so a large feature (Canada ~5.5MB, ~340 chunks) made
the one-time admin1 store build O(n^2). That pushed the streaming build
(~4.2s) past the 15s test timeout under CI coverage/fork contention and
added the same latency to the first live region request.
Carry scan state (position, brace depth, string state) across chunks so
each character is examined once. Output is byte-identical (3228 features,
197 countries) and peak build memory is unchanged (~192MB RSS under a hard
512MB cap); build time drops to ~1.7s.
* fix(i18n): add packing.quantity to all locales
The packing.quantity key existed only in en (Qty) and de (Menge), so
i18n:parity:strict and the client parity test failed for the other 20
locales. Add the key everywhere; en/de are unchanged.
---------
Co-authored-by: jubnl <jgunther021@gmail.com>
* fix(trips): keep accommodations on their dates when the trip range shifts (#1288)
The v3.1.3 fix re-anchored dated bookings after a trip date change but
explicitly excluded hotels: day_accommodations has no absolute date columns,
so stays remained glued to positionally re-dated day rows and shifted with
the range. updateTrip now snapshots day dates before generateDays and a new
resyncAccommodationDays re-anchors each stay (and its linked hotel
reservation, restamping its stale reservation_time) to the days holding its
pre-change dates; out-of-range stays stay glued so whole-trip moves still
shift together. Unlinked dated hotels resync like any booking, and the
date-change block is wrapped in a transaction.
Changing the start date now also asks how plans should follow via a new
date_shift_mode field ('keep_bookings' default / 'shift_all', which reuses
the reorder/insert restamp path to glue everything), exposed in the trip
edit modal (all 22 locales), the shared contract, and the MCP update_trip
tool. Clients no longer show stale state: the initiator reloads
reservations + accommodations after saving, collaborators refetch on a
date-changing trip:updated, and reconnect hydration nudges the planner's
accommodations too.
* fix(extract): retry with json_object and surface AI import failures (#1546)
OpenAI-compatible providers that only support json_object (DeepSeek,
Mistral, some vLLM/llama.cpp) reject the json_schema response_format
with a 400, and the resulting error was swallowed silently: not logged
server-side and never rendered by the background-task widget, leaving
only a generic "no reservations" message.
- Retry the chat/completions request once with response_format
json_object when the json_schema attempt returns 400 (non-NuExtract
only); the system prompt already dictates the output shape
- Log swallowed llm-parse errors with an [llm-parse] tag so failures
show up in server logs
- Render task warnings under the empty-preview note in the background
tasks widget so the actual provider error reaches the user
* fix(costs): keep ticket item amounts visible on narrow screens (#1568)
* chore: update repo url
* chore: update repo url
* chore: update repo url
* chore: update repo url
* fix(security): block IPv6 transition addresses (NAT64/6to4/Teredo) in SSRF guard
An attacker-controlled DNS record pointing at a NAT64 (64:ff9b::/96),
6to4 (2002::/16), or Teredo (2001:0000::/32) address that embeds a
private IPv4 (e.g. 64:ff9b::a9fe:a9fe = 169.254.169.254) bypassed the
SSRF guard: none of the guard functions recognised these ranges, so on a
host that routes the transition prefix the connection reached the
embedded private target (cloud metadata / internal SSRF).
Add a shared embeddedTransitionIpv4() detector and re-apply each guard's
own blocklist to the extracted IPv4 in isAlwaysBlocked, isPrivateNetwork,
isLinkLocal (ssrfGuard.ts) and isBlockedIp (egress-policy.ts). A
transition address to a public IPv4 stays allowed so legitimate
IPv6-only egress is unaffected.
egress-policy.ts keeps the detector inline to preserve its dependency-free
contract for the isolated plugin subprocess.
* chore: Add star history
* Revert "chore: Add star history"
This reverts commit f9d5f75837.
* fix(map): draw transit routes even without other places on the day
The reservation overlays hide any transport whose from/to endpoints
project closer than a per-type pixel threshold (200px for transit) to
declutter tiny no-op straight connectors. A transit journey, though,
draws its real rail/bus alignment rather than a straight endpoint line,
so on a zoomed-out day — one with no other places to tighten the map
onto — its stations fall under the threshold and the whole route
vanishes.
Exempt a transit booking that carries real per-leg geometry from the
proximity gate in both renderers (Leaflet + MapLibre); a geometry-less
transit keeps the straight-arc declutter.
* fix(atlas): resolve region AND country by coordinates against the bundled polygons
Rebased onto dev's streaming atlas index (#1576): region resolution now
resolves a place's lat/lng directly against the same bundled admin1
polygons the client renders — offline, deterministic, and guaranteed to
match a bundle feature — rather than trusting Nominatim's address level,
which can name a subdivision the bundle doesn't carry (Barcelona's ES-B
province vs the bundle's ES-CT autonomous community) and never highlight.
Country resolution moves to the same coordinates-first order, so a place
stored "..., San Francisco, CA" no longer resolves to Canada.
admin1 is held as per-country GeoJSON text (never parsed whole, #1576),
so a country's regions are flattened to the compact Float64Array form and
cached on first use — only visited countries pay the parse. The stale GB
constituent-country rescue is removed, and a one-time migration clears the
re-derivable place_regions cache so every place re-resolves under the new
logic.
Closes#1547
* fix(client): convert mixed-currency day/trip cost totals instead of mislabeling raw sums (#1561)
Day headers, the plan sidebar footer and the PDF day/cover totals summed
raw place prices across currencies and labeled the result with a single
currency — a $2,730.27 hotel on a NOK trip read as "2730 NOK".
Totals now convert every amount into a base currency via the existing
frankfurter rates (sidebar: the user's display currency, falling back to
the trip's; PDF: the trip currency, resolved once before rendering so the
document is consistent), marked with "≈" when a conversion happened. When
a rate is unavailable (offline, blocked egress, unknown code) they fall
back to an honest per-currency breakdown ("2 500 kr + $2,730.27") instead
of folding foreign amounts into a mislabeled number. All-same-currency
trips make no FX request, so offline PDF export keeps working.
Also swaps the place inspector's hardcoded € chip icon for a neutral one
and formats the price via formatMoney in the place's own currency.
Note: day-header totals move from "50 EUR" to Intl formatting ("50 €").
* feat(atlas): let a visited region be hidden, cascading to the country when none remain (server)
Countries already have a hide/tombstone mechanism (hidden_countries, #1490):
a zero-count derived country can be dismissed and stays gone across reloads.
Regions had no equivalent — unmarkRegionVisited only ever deleted a
manually-marked visited_regions row, a no-op for the common case of a region
derived fresh from place_regions on every request, so there was no way to
dismiss one at all (place-derived or otherwise).
Adds the region-level counterpart:
- New hidden_regions table (user_id, region_code, country_code), mirroring
hidden_countries.
- getVisitedRegions() filters its result through it, the same way getStats()
already filters through getHiddenCountries().
- unmarkRegionVisited() now tombstones unconditionally (not just for a
manually-marked region — a region with a real place attached, e.g. one
misassigned by a border-simplification gap, is exactly the case this
exists for) and derives the country code from the region code's
"<country>-<rest>" prefix when there's no visited_regions row to read it
from.
- Cascade: after hiding a region, if the country has no other visible region
left (checked against place_regions + visited_regions, minus hidden_regions),
the country is hidden too via the existing unmarkCountryVisited.
- markRegionVisited() clears both tombstones on re-mark, so a region (and its
cascade-hidden country) can come back.
Note the cascade only has a visible effect on a country with no real place
attached to it — getStats' places-derived country entries are never
suppressed by hidden_countries (#1490's deliberate "reappears with a real
place" rule), so hiding every region of a country that DOES have real places
leaves the country visible, by the same existing design. Test coverage
reflects this.
Server-side only — client wiring (a way to trigger this from the map) is a
separate commit.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(atlas): let a visited region be hidden, cascading to the country when none remain (client)
Client wiring for the server-side hide-region feature (previous commit).
Clicking a visited region on the map used to open the hide/unmark
confirmation only when it was manually marked (visitedRegions[...].manuallyMarked);
a region derived from real place data instead opened the country-detail
view, with no way to dismiss it at all. Now any visited region offers the
same "Remove this region from your visited list?" confirmation regardless of
how it was derived — country details remain reachable via the country
search/sidebar, which was never gated on this in the first place.
The confirm handler's optimistic country-removal check dropped its
`&& r.manuallyMarked` filter on the remaining-regions count, matching the
server's unconditional cascade — but keeps the existing "only when the
country has zero real places/trips" guard, since a country backed by real
data is never actually hidden server-side (#1490) and removing it from the
UI early would just flash and reappear on the next reload.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(settings): make fresh-instance unit defaults internally consistent
A brand-new instance — no admin-set defaults, no saved value — rendered
temperatures in Fahrenheit and times on the 12-hour clock while distances
defaulted to metric, a mix that matches no locale. The store also seeded
'fahrenheit'/'12h' while DisplaySettingsTab's fallback said 'celsius', so the
two paths disagreed about the intended default in the first place.
Default to one system — celsius / metric / 24h — matching the already-metric
distance_unit. The unit defaults now live in a single exported DEFAULT_SETTINGS
that both the store and DisplaySettingsTab's fallback read, so they can't drift
apart again. Admin-set user defaults (getAdminUserDefaults) and any value a user
has already saved still take precedence; only the code-level default changed.
* feat(i18n): complete Catalan (ca) translation
Adds Catalan as a supported language — the full shared/src/i18n/ca locale
(all domain files plus the notification texts), registered in
SUPPORTED_LANGUAGES and the client locale loader. Rebuilt onto current dev so
the locale is at full key parity with en (i18n:parity:strict clean).
* fix(map): stop real road-route fetches from dying under StrictMode
useTransportRoutes cached its AbortController in a ref that was
created once and aborted on unmount. React StrictMode's dev-only
mount->cleanup->remount cycle ran that abort during the *simulated*
cleanup, permanently poisoning the controller before the real mount's
fetch ever started — every road-routed booking (car/bus/taxi/bicycle)
silently fell back to a straight line in local dev, while production
builds (no StrictMode double-invoke) worked fine.
Fix: create a fresh AbortController per effect run instead of a
ref-cached singleton, and synchronously un-mark a job as "attempted"
in that same run's cleanup if it didn't settle before the cleanup
fired — so a StrictMode remount (or any other pre-completion
cancellation) retries instead of being skipped forever. Verified
against the real OSRM endpoint in a running dev server: all four
road-routed legs on a live trip now resolve with real road geometry
instead of straight lines.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): add reservation route-visibility util
Extracts the route-visibility filter that was hand-copied, identically,
in both MapView.tsx and MapViewGL.tsx into one pure, unit-tested
function: a reservation's route shows on the map when it's a transit
booking with the day-route toggle on, or its id is in the caller's
visible-ids set. isRoutableReservation (>= 2 endpoints) is exported
separately since callers besides the map filter need the same check.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): add per-trip connections-visibility resolution util
Pure storage/resolution logic for a trip's booking-route visibility
preference, keyed trek:visible-connections:<tripId>. Two modes:
'only' (nothing shown except the listed ids — today's existing
behavior; a legacy bare-array localStorage value parses as this mode
for backward compatibility) and 'all-except' (everything routable
shown except the listed ids). A trip with no stored preference falls
back to the account-wide default (all or nothing) without writing
anything, so flipping the account setting later never silently
overrides a trip with an explicit per-trip choice already recorded.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(settings): add always-show-booking-routes account setting
New map_always_show_routes account setting, defaulting to off, i18n'd
across all 22 supported locales. Lives in Display > Travel & map,
directly under Booking route labels — its closest sibling — using
that section's immediate-save On/Off pattern rather than a separate
toggle+Save flow, since it's a booking-display preference, not a map
render-config option.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): per-trip booking-route visibility with account default and bulk toggle
Lets a user see a booking's route on the map without manually
toggling it per item, two ways:
- The account setting from the previous commit sets the default for
any trip that's never had its routes touched before.
- A new bulk "show all / hide all" button in the day-plan toolbar
flips a trip explicitly between the two connectionsVisibility modes,
independent of the per-item toggle (which still edits whichever
mode's id list is active, in both directions, including while the
account default is on).
useTripPlanner.ts resolves a trip's effective visible-connection ids
from connectionsVisibility.ts + the account setting + the trip's
routable reservations, and exposes it through the same
visibleConnections/toggleConnection contract MapView, MapViewGL and
DayPlanSidebar already had, plus allConnectionsShown/
toggleAllConnections for the new bulk control. MapView/MapViewGL
consume it via the shared reservationRoutes util instead of each
carrying their own copy of the filter.
The bulk toggle's tooltip gets its own map.showAllConnections/
hideAllConnections i18n keys (all 22 locales) distinct from the
per-item toggle's text, and matches the per-item toggle's active
(solid blue) styling rather than the toolbar's generic hover tint.
Manually verified end-to-end in a running dev server: the account
default seeding an untouched trip, the bulk toggle flipping a trip
between all-shown/all-hidden, and a single per-leg override while the
trip is in all-shown mode.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* Update documentation for booking visibility change
* fix(i18n): add Catalan keys for the booking-route toggle + fix the language-count test
#1483 added map.showAllConnections/hideAllConnections and
settings.alwaysShowRoutes/Hint to every locale that existed when it was
written; Catalan (#1418) landed just after and was missing them, which
broke i18n parity. Also bumps the SUPPORTED_LANGUAGES test to 23 entries
and asserts the Catalan one — the ca addition left that assertion stale.
* test(e2e): add a reproducible documentation screenshot run
The wiki screenshots had drifted three months and two releases behind the UI:
39 of 46 assets came from a single commit in April, and pages such as
Budget-Tracking carry a disclaimer that their own images are out of date.
Retaking them by hand is what created that drift in the first place.
This adds a `screenshots` Playwright project that captures them from a dev
build instead:
- seed.ts populates a demo trip over the REST API — deliberately in JPY, so the
3.4.0 currency work (per-trip currency, frozen FX rates, foreign-currency
settle-up) is actually visible rather than hidden behind EUR.
- shot.ts settles the page before capturing (fonts, images, transitions) so
captures don't catch skeleton loaders, and rewrites /auth/app-config to clear
dev_mode — the E2E backend runs NODE_ENV=development, which would otherwise
put a "Dev: Notifications" tab no real deployment has into every admin shot.
- promote.mjs downscales 2880px captures to 1600px on the way into
wiki/assets/, which is 70% smaller for detail the wiki never renders.
Tabs are located by their visible label, so a rename like Budget → Costs fails
the run loudly instead of quietly capturing the wrong panel.
npm run shots && npm run shots:promote
* docs(wiki): retake screenshots against 3.4.0
Regenerated with `npm run shots` from the dev build. Replaces 21 assets and
adds 13 surfaces that had no screenshot at all.
Notable corrections:
- Collections.md referenced assets/Collections.png, which was never committed —
the only broken image in the wiki. It now exists.
- The Budget panel became Costs in 3.3.0 (#1464); every budget image still
showed the old label.
- The trip-create dialog gained a Currency field in 3.4.0, absent from the old
TripCreate.png.
- Settings grew a Plugins tab and the admin sidebar a Plugins entry; neither
appeared in the old sidebar shots.
- Weather now renders in Celsius. A fresh instance defaults to Fahrenheit while
distance defaults to metric, so the seed pins both units — the mismatched
defaults are a separate bug, not fixed here.
Koffi is installed from the community registry rather than dev-linked, because
dev-link and sideload both stamp a badge on the plugin card that an ordinary
install never shows.
Total 3.6 MB for 34 images, against 26 MB for the 46 assets already in the
directory.
No wiki page text is touched here — pages still point at the old filenames
where those were kept, and the newly added images are not referenced yet.
* fix(help): make wiki links and anchors work in the in-app reader
The wiki pages are written in GitHub-wiki style, and 455 of their links across
81 pages use the bare relative form — `[Currencies](Currencies)` — against only
114 `[[..]]` links. processMarkdown only rewrote the latter, so in-app every one
of those 455 fell through to HelpPage's external-link branch, opened a new tab
and 404'd. Since 6c87bf2f serves the wiki from disk, that is the primary way
users read these docs.
Rewriting them in the renderer fixes every page at once and keeps the sources
GitHub-compatible, so contributors can keep writing either form and neither
target breaks. Rewriting the 81 files instead would have fixed today's pages and
left the trap open for tomorrow's.
Also:
- Code is protected from the rewrite. Plugin-Development.md documents
`actions[key](ctx)`, which reads as a markdown link and would otherwise be
corrupted into `actions[key](/help/ctx)` inside a verbatim snippet.
- `[[Page#anchor|Slug]]` no longer renders its anchor as visible link text.
- Headings carry GitHub-compatible ids, so the 22 in-page `](#anchor)` links
scroll instead of doing nothing.
* docs(wiki): correct settings, map and addon docs against 3.4.0
The wiki described settings that no longer exist and pointed at UI labels that
had been renamed. Each correction below was checked against the code.
- The Settings tab is labelled **General**, not "Display"
(shared/src/i18n/en/settings.ts:6). Every "Settings → Display" path was wrong.
The *display currency* setting keeps its name — only the tab was renamed.
- Colour mode is on the **Appearance** tab, not Display.
- The "Route calculation" setting does not exist: no hits for route_calc /
routeCalc / auto_route / calculateRoutes anywhere in client or server.
- Default map centre and zoom were removed in 3.4.0 (0f4766e1). Replaced with
what actually happens now, written from client/src/utils/mapViewport.ts:
every map frames itself on its own places, world view when a trip has no
coordinates.
- A third map provider, MapLibre GL / OpenFreeMap, was undocumented. It needs no
access token, which is the reason a reader would choose it over Mapbox.
- Budget-Tracking.md said the feature is called Costs everywhere and then told
the reader to open the "Budget" tab. The screenshot disclaimer is gone too —
the images now show Costs.
- The admin tab table was missing Plugins.
- The `packing` addon is seeded as "Lists", not "Packing list management".
- Install docs pinned `mauriceboe/TREK:3.0.15` as the exact-release example,
four minor versions stale.
* docs(wiki): document passkeys, calendar feeds, appearance, plugins and help
Five shipped features had no user documentation at all:
- **Passkeys** — WebAuthn enrolment and sign-in, admin policy, RP ID/origins.
Previously mentioned only in passing in Environment-Variables.
- **Calendar Feeds** — the subscribable per-trip and per-user ICS feeds. Note
Day-Plans-and-Notes documents only the one-off .ics *export*; the two are
cross-referenced so it is clear which is which.
- **Appearance Settings** — the whole tab, including the custom accent colour
and its contrast check.
- **Admin: Plugins** — installing from the registry, the pre-install permission
review, egress hosts, and what Reviewed/Signed/Unsigned actually guarantee.
- **In-App Help** — that the wiki ships in the image and is served from disk
since 3.4.0, with the GitHub fallback and TREK_WIKI_DIR.
One deliberate deviation from the brief: the Appearance settings are documented
as account-level, not per-device. They persist to /api/settings on the user
account with nothing in localStorage, so the dashboard widget picker's
desktop/mobile split still changes both from either device.
All five are listed in _Sidebar.md.
* docs(wiki): add screenshots for the remaining surfaces
Second capture pass, bringing the run to 42 screenshots. Adds the surfaces that
need more than a navigation to reach: collection and journey detail, MCP access,
two-factor setup, the settle-up payment dialog, and the trip file manager.
Two fixes to the harness itself, both of which had produced misleading images:
- The admin captures showed a "Dev: Notifications" tab that only exists when the
server runs NODE_ENV=development. The run now clears dev_mode in the
/auth/app-config response; switching the server to production instead would
have enabled HSTS and broken the run over http://localhost.
- The settle-up capture clicked "Settle up", which does not open a view — it
records the transfers. It zeroed every balance and photographed "Everyone's
square", and because the specs share one database it poisoned Costs.png in the
same run. Screenshot specs must not mutate state; it now captures the
"Add payment" dialog instead.
Koffi is installed from the community registry rather than dev-linked, since
dev-link and sideload both badge the plugin card in a way no ordinary install
does.
* test(e2e): capture the detail pages and dialogs
Adds the second wave of screenshot specs (collection/journey detail, MCP access,
2FA, settle-up dialog, files) and enables the mcp, documents and collab addons
in the seed so their surfaces render instead of 404ing.
* docs(wiki): point plugin authors at the agent skill and the registry
Plugin-Development jumped straight into scaffolding without saying that two
supporting resources exist. TREK-Plugins was referenced only in passing far down
the page, and Plugin-Skill — an agent skill that teaches Claude Code and other
SKILL.md-compatible agents to build and publish a plugin — was not mentioned
anywhere in the wiki.
Both are called out up front, with a note that neither is required: the registry
only matters once you want other instances to find your plugin.
* test(e2e): capture the four collab surfaces separately
One Collab.png illustrated chat, notes, polls and the What's Next widget, so at
most one of those four wiki pages showed the feature it described. Each now has
its own capture.
Two things the collab seed needed:
- The conversation is posted by three different people. Every collab write is
attributed to the acting user, and a single-voice chat log would misrepresent
the feature outright.
- Each member therefore gets its OWN request context, created with an explicit
`storageState: undefined`. Without that, newContext inherits the project's
storageState — the admin's trek_session cookie — and extractToken reads the
cookie BEFORE the Authorization header (server/src/middleware/auth.ts:9). The
posts still return 200; they are just all recorded as the admin. That is
exactly what happened on the first attempt, and the DB was the only place it
showed.
The collab view is not tabbed — CollabPanel renders all panels at once — so the
captures target cards by seeded content rather than clicking tabs or matching
headings, whose DOM text is 'Notes'/'Polls' while CSS renders them uppercase.
* docs(wiki): show every screenshot on the page it belongs to
Finishes the wiring the screenshot commits deliberately left out.
- Embeds the 14 images that were committed but displayed nowhere: the Costs
panel and settle-up dialog, the trip planner, transports, documents,
collection and journey detail, the notifications inbox, the Offline and
Account settings tabs, appearance, admin user defaults, registration and
password reset.
- Splits the four collab pages onto their own images. Chat, Notes, Polls and
What's Next each showed the same Collab.png until now; the overview shot moves
to Real-Time-Collaboration, which had no image at all.
- Day-Plans-and-Notes pointed at TripPlaner.png — one 'n'. It now uses the
correctly spelled file, and the misspelled one is deleted since nothing else
referenced it.
- Removes 45 dead '<!-- TODO: screenshot -->' markers whose screenshot had long
since been added. 9 remain, each on a page that genuinely still lacks the
image it asks for — so the marker means something again and the gap is
greppable, which is how this drifted unnoticed for three months.
Every asset in wiki/assets/ is now referenced by a page, and every image
reference resolves to a file.
* docs(wiki): add the four collab screenshots
Chat now shows a real three-person conversation rather than one voice talking to
itself, and the poll shows three separate votes across two options.
44 images, 4.6 MB total.
* test(help): point the asset test at the correctly spelled screenshot
The integration test hard-coded assets/TripPlaner.png — one 'n' — so deleting
the misspelled file broke it. It was the only thing keeping that filename
alive.
* docs(wiki): regenerate the screenshots on top of the rebased dev
ba3733da changed the fresh-instance defaults to celsius/metric/24h. Every
capture showing a clock — chat timestamps, bookings, day plans — and the General
settings tab itself were still on the 12-hour clock, so 39 of 44 images needed a
new run. Regenerating them is one command, which is the point.
The seed keeps pinning the units explicitly: it now matches the new defaults, but
stating them keeps the captures reproducible if a default moves again.
* docs(wiki): regenerate screenshots after rebasing onto dev
dev added a Catalan translation, an always-show-booking-routes account setting
and a bulk route toggle in the day-plan toolbar since the last run — all visible
on captures we ship. 15 of 44 images changed.
Also resolves the Map-Features conflict from 41d12e89: upstream's new bulk-options
section is kept, with the two 'Settings → Display' paths corrected to 'General'.
The tab is labelled General (shared/src/i18n/en/settings.ts:6), and upstream's own
i18n key for that section is settings.general.travelMap.
* chore: only allow manual trigger for the build&push
* feat(mobile): rebuild Settings & Admin panels natively
* feat(mobile): native booking, transport & expense form sheets
* feat(mobile): unify empty states, streamline nav, collab & collections
* feat(mobile): white atlas bucket button + custom month/year picker
* feat(mobile): show accommodation route legs in the day plan
* feat(mobile): auto-draw the day route in map mode
* feat(mobile): move the vacay invite button left of the year switcher
* fix(mobile): don't render the global dock inside the trip planner
* feat(mobile): reachable add-list, roomier packing rows & owner avatars
* fix(mobile): use the custom date picker for the to-do due date
* feat(mobile): open a running trip on today's day
* chore(client): serve the PWA manifest in dev for on-device testing
* fix(i18n): complete the Catalan (ca) locale for the mobile rewrite
* test(transport): match the shortened Manual/Automated switch labels
* test(mobile): mock the admin permissions panel MAdmin renders
* feat(vacay): rework the desktop vacation planner in the glass style
Move the calendar, sidebar cards and stats onto the --vg-* glass tokens,
soften the person-day fills, pad every month to six week-rows so the
cards line up, tighten the sidebar spacing and add You/Pending badges to
the people list.
* feat(vacay): redesign the vacation settings dialog
Widen the modal and split the options into two columns, and restyle the
holiday calendar rows, inputs and add button to match the rest of the
planner.
* feat(journey): rework the journeys dashboard in the glass style
Move the hero to a blur-mask magazine cover, put the journey cards on
glass with cover/title overlays, widen the grid to match the rest of the
app, drop the top toolbar, and add a subtitle field to the create flow.
* feat(journey): modernise the journey detail page
Centre and widen the feed+map layout, restyle the hero with a frosted
cover, glass stat pill and round controls, turn the view switch into a
glass segmented pill with a labelled add-entry button, and give the
timeline day headers, entry cards, suggestion cards and the entry
editor a rounder, glassier look.
* feat(journey): lay out the journey settings dialog in two columns
Widen the dialog and split the options into two glass cards, and round
the inputs and the save/cancel buttons.
* feat(journey): lay out the entry editor in two columns
Widen the new/edit entry dialog and split it into two columns — title,
photos and the story on the left (the story now grows to fill the
column), date/location, pros & cons, mood and weather on the right.
* feat(journey): polish the detail, settings and share UI
Move the gallery upload into the view switch as a matching pill and drop
the old button, badge the day-place and photo counts, tidy the entry
editor location field, round the settings inputs and buttons, restyle the
contributor role and share controls, and put delete next to copy.
* feat(journey): always feature a hero on the frontpage
The desktop frontpage only rendered the hero header when a journey trip was currently live, so with no live trip there was no header at all. The hero now always features one journey — the live one when there is one, otherwise the most recent — and its eyebrow reads "Latest Journey" when the featured journey is not live.
* feat(mobile): customizable bottom navbar layout
Add a Mobile section under Settings → Appearance where you arrange the bottom navbar: reorder items and choose which sit in the bar versus under the "More" popover, with a live preview. Dashboard stays pinned first. The layout persists per-user in the appearance config and drives MBottomNav; an un-customised account keeps the current Dashboard + Vacay/Atlas dock.
* fix(mobile): widen the nav customizer zone buttons to a single row
* refactor(mobile): show the nav customizer only in mobile settings
The bottom-nav layout only exists on phones, so configuring it from the desktop settings made little sense. Drop the desktop customizer (and its two now-unused labels) and keep the editor on the mobile appearance screen.
* fix(mobile): neutral tiles and no subtitle in the bottom-nav More menu
* feat(desktop): TREK mascot empty states + trip-open splash
Bring the mobile TREK mascot to desktop as a shared EmptyState (mascot + a single title, no subtitle, one uniform look) across the app empty states: reservations, transport, costs, packing, to-dos, files, collab notes/polls/chat, upcoming activities, the day plan, journey timeline and gallery, collections, atlas, dashboard and notifications. The trip planner now plays the same little mascot journey while a trip loads.
* feat(desktop): animated glassmorphism background for the trip loading splash
The trip-open splash now takes over the viewport with a calm, slowly drifting pastel gradient wash (soft pink/lavender/periwinkle, toned down), a gently breathing hue veil, a whisper of filmic grain and a soft vignette, with the mascot journey floating on a frosted liquid-glass card. Theme-aware for light and dark; the background freezes under prefers-reduced-motion.
* fix(planner): keep the empty-day itinerary as small text, no mascot
* fix(mobile): cap the bottom-nav dock at Dashboard + 2 so the More slot fits
* feat(mobile): reorderable dashboard arrangement
Add a mobile-only dashboard order control under Settings → Appearance → Mobile: reorder how the trip list and the inline widgets (currency, collections, world clocks, upcoming reservations) stack on the phone dashboard — e.g. move the currency converter above your trips. The featured trip always stays pinned on top. Stored per-user in the appearance blob (dashboard.mobileOrder); an un-arranged account keeps the current order.
* feat(plugins): map layer provider hook for the trip map
Plugins can now draw bounded vector overlays (polylines, polygons, metric
circles) on the trip map via hooks.mapLayerProvider — routes, reachable-range
corridors, zones. Same contract as map markers: declarative data only, tone
palette styling with clamped numerics, hard per-plugin budgets (4 layers /
150 features / 8000 vertices), oversized or partly-invalid shapes dropped
whole. Rendered beneath the core day route.
The GL renderer now also mounts the plugin markers it was missing — both
hooks work on Leaflet and Mapbox/MapLibre, desktop and mobile.
* feat(plugins): route provider hook with planner route profiles
A routeProvider plugin can now register routing profiles (capabilities.
routeProfiles, max 3) that show up in the planner's route toggle next to
Driving/Walking — e.g. EV routing with charging stops. Selecting one asks
exactly that plugin to route the day's stops (POST /api/plugin-routes,
20 s budget for external solvers); the result is validated whole (leg
count must match the waypoint pairs, vertex/via caps) and rendered like
an OSRM response: geometry on the map, via points as stops on the line,
per-leg times + an optional note ("25 min charge") on the connectors.
A failing provider degrades to straight lines, same as an OSRM outage.
The feed only offers profiles whose hook grant is actually recorded, and
the route endpoint re-checks provider + declared profile against the DB
row, so a hand-edited capabilities blob can't invent routable profiles.
* feat(plugins): day schedule contributions in the planner
A dayScheduleProvider plugin can attach time entries to the day plan —
"35 min charging at this stop", "45 min security before this flight".
Rows are host-rendered under their anchor (a place/booking row, or a
day's start/end) on both the desktop sidebar and the mobile timeline,
and the contributed minutes are folded into the day's route-footer
total as "+X min".
Because this output feeds displayed timing, the server clamps minutes
to a day, checks every dayId against the trip's own days and budgets
the item count per provider; labels go through the usual sanitize+cap
pipeline.
* feat(plugins): geolocation bridge permission
New geolocation:read permission — the first bridge-level grant: a plugin's
sandboxed frames can ask the HOST for the browser position over postMessage
(window.trek.geolocation.get()/watch()). The sandbox itself never gains the
geolocation API; the host page reads navigator.geolocation, so the browser's
own site permission prompt still applies on top of the admin consent, and
the position only travels parent -> frame, never to the server.
The feed flags granted plugins, an ungranted frame gets a 'forbidden'
answer without touching the browser API, one GPS watch per frame, and the
watch is force-stopped when the frame closes. The dev preview answers with
a fixed position so trek.geolocation works without a prompt.
* fix(mobile): render trip-page plugin tabs
The mobile trip shell listed plugin tabs in the Mehr sheet and let the user
open them, but MTripTabPanel had no plugin: branch — the tab came up as an
empty scroller. Mount the same sandboxed PluginFrame the desktop planner
uses, filling the panel above the dock.
* test(plugins): add granted_permissions to the feed fixture schema
The plugins feed now selects granted_permissions (route profiles + the
geolocation flag are gated on the recorded grant), so the hand-written
in-memory schema this suite shares needs the column.
* fix(map): don't crash the trip map when the pane API is absent
PluginMapLayers called map.getPane/createPane unconditionally, which threw
'map.getPane is not a function' on a minimal Leaflet map (breaking MapView
in tests and any renderer state without pane support). Pane placement is a
z-index optimization — guard it and fall back to drawing the plugin
features without the dedicated pane when it isn't available.
* harden(plugins): bound normalize work, scope route cache, enforce geo grant live
Adversarial audit of the four new hooks found no trust-boundary defect; these
close the low-severity gaps it confirmed:
- map-layers / day-schedule / plugin-routes: slice the raw provider array before
iterating, so an all-invalid oversized payload can't force an unbounded scan on
the host event loop (the output was already capped; this bounds the work too).
- RouteCalculator: a plugin route is trip-/day-specific, so its cache key now
includes tripId/dayId — two days with identical coordinates no longer share one
plugin route (its charging stops/notes differ per day).
- PluginFrame: a live geolocation watch re-checks the grant on every fix and stops
at once if an admin revokes geolocation:read while the frame stays open.
* test(client): update stale assertions to the mobile-rewrite UI
The mobile rewrite moved empty states to the shared EmptyState mascot
component (title only, no subtitle) and replaced the loading gif with
TripLoadingSplash, but several tests still asserted the old copy/markup:
- CollabChat / PackingListPanel / ReservationsPanel / AtlasPage /
JourneyDetailPage: assert the current empty-state (title + scene mascot)
instead of the removed hint paragraphs.
- TripPlannerPage: mock TripLoadingSplash (its mascot-cycling setInterval
never drains under fake timers, aborting 14 splash-gated tests) and
assert the current loading node.
Test-only — no component behavior changed.
---------
Co-authored-by: jubnl <jgunther021@gmail.com>
Co-authored-by: Dieter Blomme <dieterblomme@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: sld272 <zjrdmczh@outlook.com>
Co-authored-by: Nguyen Trong Binh <nguytb15@VN1N07HO1CD1015.local>
Co-authored-by: Pavel Zolotarevskiy <code@fxgn.dev>
Co-authored-by: jubnl <66769052+jubnl@users.noreply.github.com>
Co-authored-by: Azalea <noreply@aza.moe>
Co-authored-by: Uzini <43294422+Uziniii@users.noreply.github.com>
Co-authored-by: Daniel <drmoreno271@gmail.com>
Co-authored-by: trongbinhnguyen <43725147+trongbinh15@users.noreply.github.com>
Co-authored-by: Konstantinos Thermos <info@subdee.org>
Co-authored-by: Konstantinos Thermos <subdee@users.noreply.github.com>
Co-authored-by: Lucas Español <lucas.espanol@tutanota.com>
Co-authored-by: fbnlrz <frlrnzn@gmail.com>
* fix(transit): lead arrive-by results with the latest arrival (#1479)
MOTIS returns arrive-by itineraries ascending with the deadline-adjacent
connection last, so the visible top of the list arrived ~2h before the
requested time — misread as a timezone bug. Sort arrive-by results by
endTime descending so the connection closest to the requested arrival
leads, mirroring depart-by.
Claude-Session: https://claude.ai/code/session_01BCrZUoPzHZF6H4C6avMXoz
* fix(transit): lead arrive-by results with the latest arrival (#1479)
MOTIS returns arrive-by itineraries ascending with the deadline-adjacent
connection last, so the visible top of the list arrived ~2h before the
requested time — misread as a timezone bug. Sort arrive-by results by
endTime descending so the connection closest to the requested arrival
leads, mirroring depart-by.
* fix(reservations): run AirTrail modal hooks before the isOpen early return (#1602)
The sectionItems useMemo added by #1535 sat below 'if (!isOpen) return null'.
The modal mounts closed, so the first open render executed one more hook
than the previous render and React unmounted the whole tree (error #310),
blanking the page. Move the memo above the early return and cover the
closed-then-open transition with a regression test.
* fix(planner): only draw the check-in-day hotel morning leg when provably at/after check-in (#1597) (#1607)
On an accommodation's check-in day the hotel -> first-stop leg was drawn
by default for any PLACE first stop, suppressed only when that place was
explicitly timed before check-in. An un-timed first place (e.g. "Home"
on day 1 of a driving holiday) therefore always produced a phantom
hotel -> Home leg, regardless of the check-in time.
Flip the default to mirror shouldDrawEveningLeg: the morning leg is now
drawn only when the first place is provably timed at/after check-in
(you dropped your bags first). The drawn map route, the sidebar hotel
connectors, and the Google Maps export all share this helper and
inherit the fix. The optimizer anchors (#1321) are unchanged.
* fix(notifications): never fall back to the admin ntfy topic for per-user sends (#1608)
* chore: correct shields.io url
* chore: correct shields.io url
* chore: update helm repo link
* chore: document new helm chart url
* chore: document new helm chart url
* fix(atlas): make Kosovo selectable on the Atlas map (#1609)
Kosovo's user-assigned ISO code (XK/XKX) was missing from both code
tables in the atlas pipeline: the geo builder's A3_TO_A2 map (so the
shipped admin0 bundle carried ISO_A2: null) and the client's A2_TO_A3
map. With neither resolvable, onEachFeature attached no hover/click
handlers to Kosovo's polygon and the country search dropped it, making
Kosovo impossible to select.
- add XK<->XKX to the client A2_TO_A3 table and the builder A3_TO_A2 map
- stamp ISO_A2: "XK" on the Kosovo feature in the shipped admin0 bundle
- guard both with tests (atlasModel resolution + ATLAS-BUNDLE-003)
* chore: normalize docker image references to mauriceboe/trek (lowercase)
* fix(memories): honor Synology skip-SSL on photo streaming (#1611)
The synology_skip_ssl setting was forwarded on the JSON API path
(login/browse/test-connection) but not on the image-byte fetches, so a
NAS with a self-signed certificate passed Test Connection while
/api/photos/:id/thumbnail and /original failed with 500
"Failed to fetch asset" — and the TLS error was silently swallowed.
- pipeAsset: accept fetchOptions and forward to safeFetch; log the
underlying error on the 500 path (without the URL, which carries _sid)
- fetchSynologyThumbnailBytes / streamSynologyAsset: pass
rejectUnauthorized derived from synology_skip_ssl; log on failure
- i18n: replace hardcoded "Immich" with {provider_name} in 7 memories
keys across 14 locales (source of the reported "Failed to connect to
Immich" banner on the Synology test), fix translated placeholder
names in tr/ca/pl, and add a placeholder-parity spec to prevent
regressions
- client: gallery provider badge matched 'synology' instead of the real
id 'synologyphotos', showing the raw id; label is now
"Synology Photos"
* chore: make chart.liketrek.com the canonical helm chart url
* feat(plugins): prefer the registry's resolved store screenshot (#1613)
Browse/detail use the screenshotUrl the aggregate step resolves (cover at the latest commit, or the first resolving README image), falling back to the docs/screenshot.png guess when the field is absent. Fixes blank store cards for plugins without a committed docs/screenshot.png.
---------
Co-authored-by: Maurice <61554723+mauriceboe@users.noreply.github.com>
* 3.3.0 (#1472)
* feat(plugins): grow the frame bridge — fill pages, confirm, openExternal, live context
- page/trip-page hosts pass fill: the frame pins to 100% height and ignores
trek:resize, so a kit plugin's auto height report no longer collapses a full
page into a floating island with dead space below (widgets keep self-sizing)
- trek:confirm renders the native ConfirmDialog host-side (the sandbox has no
allow-modals) and answers trek:confirm:result; one at a time
- trek:openExternal opens validated http(s) URLs in a noopener tab — the
sandbox has no allow-popups, so plugins simply couldn't link out before
- trek:notify accepts an optional duration, clamped to 1.5-15s
- context gains dir (rtl/ltr) and is re-pushed when locale or format settings
change, not just on appearance mutations
- core events for the trip in view are forwarded as trek:event — names only,
never payloads, mirroring the server-side events surface
* fix(plugins): move trip warnings out of the content area
The warning pills overlaid the top of every planner tab at full width, sitting
on the map and its toolbar. Now a warning from a plugin that owns a trip-page
tab renders as a compact chip in the navbar centre (click jumps to the tab; the
navbar centre is free on trip pages), and everything else floats above the
content at the bottom instead. Mobile keeps all warnings in the bottom overlay
since the desktop navbar isn't there. The trip tab's frame also opts into the
new fill mode.
* feat(plugin-sdk): 1.4.0 — motion library + new bridge helpers in the kit
Mirrors the host's animation vocabulary 1:1 into TREK_UI_CSS (menu/popover/
modal/backdrop/toast enters, drawer variant under 640px, page-enter, stagger,
skeleton shimmer, chart reveals) including the reduced-motion degrade to a
gentle fade. window.trek grows confirm(), openExternal(), onEvent() and a
notify duration, and applyContext now stamps lang/dir on the document so RTL
hosts get RTL plugin UIs.
* feat(plugins): surface registry download counts in browse
The registry now aggregates GitHub release download counts per plugin
(TREK-Plugins#18) as an entry-level downloadCount. Project it through
browse/detail and show it as a compact stat on the browse cards and in the
detail meta grid. Counts are raw asset downloads (updates and CI included),
so the UI says downloads, not installs.
* docs(plugins): document the grown bridge surface and motion classes
* fix(plugins): harden the new bridge paths
Review pass over the bridge additions:
- keep the unstable useToast() object out of the effect deps (ref instead) —
it re-created the effect on every parent render, and with the new live
repost that meant a trek:context flood into the frame
- reset loads/height/confirm state and key the iframe when a host swaps
pluginId in place (tab bar, /plugins/:id) — the new plugin's document was
refused as a 'navigated' frame and every kit promise hung
- confirm dialogs always lead with the host-controlled plugin name so a
plugin can't dress its dialog up as a TREK system prompt; answer/refuse
moved out of setState updaters (StrictMode ran them twice)
- Number.isFinite on the notify duration (NaN parked a sticky toast)
- don't forward other plugins' namespaced broadcasts as trek:event; a
plugin's own plugin:{id}:* broadcasts now reach its frame though
- teach the SDK dev preview the confirm/openExternal contract so
trek.confirm() resolves in /preview
- 999,950 downloads formats as 1M, not 1000k
* fix(planner): let plugin warning chips grow wider before truncating
The nav-centre chip capped at 340px, so a longer warning (e.g. the TREK x
Japan weather prompt) was ellipsised almost immediately. Scale it with the
viewport up to 520px so most messages read in full while still yielding on
narrow desktops.
* feat(plugins): sort the plugin browser by download count
Discover now honours the sort dropdown (it was always alphabetical) and adds
a 'Most downloads' option that ranks the registry by downloadCount. The sort
keys are scoped per tab — updates-first stays with Installed, most-downloads
with Discover — and snap back to name when the tab can't offer them.
* feat(plugin-sdk): auto-upgrade native <select> to a host-styled dropdown
A sandboxed plugin can't reach the host's components, and a native <select>
draws its popup from the OS — so plugin dropdowns never matched TREK. The design
kit now enhances every <select> into a keyboard-accessible listbox that uses the
kit tokens, keeping the real element as the value/form source (it still fires
change). Authors write a plain <select> and get the host look for free; opt a
field out with data-trek-native. validate warns when a plugin ships a <select>
without inlining the kit.
* feat(plugins): add issue url link
* feat(plugins): reservations write + cross-trip reads
- db:write:reservations -> reservations.create/update/delete, gated exactly like
the REST/MCP path (reservation_edit + trip membership, acting user host-bound,
no impersonation) and delegating to ReservationsService so the accommodation,
budget-sync, booking-notification and reservation:* broadcasts match the web
app 1:1 — a booking/flight/import plugin can finally write a reservation
- trips.listMine / reservations.listMine: enumerate every trip and booking the
acting user can access (membership baked into listTrips, never a raw
cross-tenant SELECT) — dashboards/aggregates were impossible before
- audit: derive auditability from METHOD_PERMISSION so a new capability method
can't be added un-audited by omission
- typed ctx.reservations.* / ctx.trips.listMine, perm label (en/de), wiki
* feat(plugins): read scopes for journal, atlas, vacay and day notes
- db:read:journal / db:read:atlas / db:read:vacay expose the acting user's OWN
journals / visited countries+regions / vacation plan across all their trips
(user-scoped like costs.listMine, each gated on its addon being enabled),
reusing the addon's existing readers
- db:read:daynotes -> daynotes.list(tripId, dayId), trip-scoped and
membership-checked like the other trip reads
- typed ctx.journal / atlas / vacay / daynotes, perm labels (en/de), wiki,
audit resource labels, tests
* feat(plugins): day notes write scope
- db:write:daynotes -> daynotes.create/update/delete, gated under the app's
'day_edit' permission (like days) with the day verified to belong to the trip;
reuses dayNoteService and broadcasts the same dayNote:* events so open
sessions update live
- typed ctx.daynotes.create/update/delete, perm label (en/de), wiki, tests
* feat(plugins): run declared background jobs on a schedule
- plugins already declared jobs {id, schedule} but the cron was never wired. The
host now schedules them: host-entry reports each job's schedule, the supervisor
starts the jobs (node-cron) when the plugin goes active and stops them on
kill/deactivate so nothing leaks
- opt-in via a new jobs:run permission — scheduled work runs with NO acting user
(its trip reads stay refused; it can only use ctx.db and declared egress), so
background execution is a distinct, admin-granted capability. Invalid crons are
skipped and a throwing job can't break the host
- extracted a small, unit-tested scheduler (plugin-jobs.ts); perm label (en/de),
wiki, tests
* feat(plugins): read scope for saved-place collections
- db:read:collections -> collections.listMine() / collections.get(id): the acting
user's own collections (user-scoped, gated on the Collections addon), reusing
collectionsService
- typed ctx.collections, perm label (en/de), wiki, audit resource labels, tests
* fix(plugins): translate new permission labels to all locales + cover the new wiring
- add the 8 new admin.plugins.perm.* labels (reservations/day-notes writes, the
journal/atlas/vacay/day-notes/collections reads and jobs:run) to the remaining
20 locales so the strict i18n key-parity test passes again
- cover the create-rpc-host reservation / day-note / cross-trip / addon-read deps
(the real side-effect wiring the mocked rpc-host tests don't exercise) so the
src/nest 80% branch-coverage gate holds
* feat(plugins): dev-link — hot-reload a local plugin against real data
Answers a plugin developer's ask: today you either get `trek-plugin-sdk dev`
(fast hot-reload but MOCK/fixture data) or the full build->pack->upload->activate
cycle (real data, no watcher). Neither gives "local dir + hot-reload + real data".
- POST /admin/plugins/link registers a plugin from a LOCAL built directory by
symlinking it into the plugins volume — the loader already forks the resolved
real path, so ZERO loader change — and registering it INACTIVE as `local:link`.
Validates the manifest + refuses native binaries exactly like a sideload.
- POST /admin/plugins/:id/reload re-forks a linked plugin via the existing
deactivate->activate primitive (same grants, no re-consent unless the manifest
widened perms). A best-effort fs.watch auto-reloads on rebuild.
- It runs through the UNCHANGED capability RPC host: real, membership-gated data,
acting user host-bound, no impersonation — code origin never touches the gate.
- Gated behind TREK_PLUGINS_DEV_LINK on top of admin + kill-switch, because a
linked plugin bypasses the install-time signature model and, under `npm run
dev`, the OS jail is off. Off by default; never reachable in production.
- discovery follows a symlinked <root>/<id>; uninstall/link never delete the
author's source (link-safe removal for POSIX symlinks and Windows junctions).
* docs(plugins): document the dev-link real-data hot-reload workflow
Adds a "Test against a real instance's data (dev-link)" subsection next to the
mock-data SDK preview: TREK_PLUGINS_DEV_LINK, POST /link with a local built dir,
activate + consent, hot-reload via the file-watch / POST /:id/reload / Restart,
and the dev-only security caveats.
* feat(plugins): dev-link admin UI
- surface devLink (TREK_PLUGINS_DEV_LINK) in GET /admin/plugins so the panel shows
the link form only where dev-link is enabled
- AdminPluginsPanel: a "Link a local plugin" form (path -> POST /link), a Dev-Link
badge for source_repo=local:link, and adminApi.pluginLink/pluginReload
- fix: the plugin menu treated any non-local:upload source_repo as a GitHub repo,
so a dev-linked plugin rendered github.com/local:link links — exclude local:link
- labels for the 6 new dev-link UI strings across all 22 locales
* docs(plugins): document the dev-link admin UI
The dev-link section showed only the curl call — surface the Admin → Plugins
"Link a local plugin" field (the primary path) and the Dev-link badge, with curl
kept as the scripting alternative.
* feat(plugins): enrich core events with { entity, entityId }
Subscribed plugins now learn WHICH entity changed, not just the event name — a
reservation/place/day/... id derived host-side from an explicit per-family
whitelist. Threaded through the six event hops WITHOUT touching actingUserId: the
handler still runs with no user, so the id is not dereferenceable (a trip read is
still refused; the id says what to react to, not what it contains). A non-entity id
can never surface — budget:member-paid-updated yields the itemId, never the userId
— and bulk/reorder/sub-entity payloads carry no id. The mapper is pure, synchronous
and never throws into the core broadcast. No new permission (reuses
events:subscribe); backend-only.
* feat(plugins): packing write scope with #858 privacy-scoped broadcasts
- db:write:packing -> packing.create/update/delete, gated under the app's
'packing_edit' permission (like the REST path) with the host-bound acting user
as owner; reuses packingService
- replicates the packing privacy model 1:1 (the controller/service helpers aren't
exported): create/delete fan out to the item's viewers only (owner + recipients,
or the whole room for a Common item); update runs the four public<->private
transitions, dropping a freshly-privatized item from the room BEFORE re-adding it
owner-only so it never leaks. A stale write is BAD_PARAMS with no broadcast
- typed ctx.packing.create/update/delete, perm label (22 locales), wiki, tests
(rpc-host gating + the four transitions + owner-scoped delete)
* feat(plugins): tableContributor hook — host-rendered view columns/actions (backend)
The registry backend for plugin-contributed columns/actions in the native planner
views (the tabular-reservations use case), mirroring placeDetailProvider:
- hook:table-contributor + the tableContributor hook (getContributions(view,
tripId, ctx) -> TableContribution[]), double-gated (implement + grant) like the
other provider hooks
- GET /api/view-contributions/:view/:tripId — view whitelist + membership gate +
per-provider timeout/fail-safe, plus the hardening the older provider hooks lack:
every field is String-coerced + length-capped, kind/tone/target enum-whitelisted,
per-provider counts capped (<=20 columns / <=10 actions), and a column url must be
http/https/mailto (a javascript:/data: url is click-XSS into the native DOM)
- typed pluginsApi.viewContributions + the ViewContribution union, perm label
(22 locales), wiki, hardening tests
* feat(plugins): render tableContributor columns/actions in the reservations view
The frontend for the tableContributor hook: a reusable PluginContributions layer
(usePluginViewContributions + PluginColumns/PluginActions) that renders the
host-normalized column/action leaves NATIVELY — a column is text/badge/link, an
action is a button that calls the plugin route or opens its sandboxed frame in a
modal (plugin markup only ever runs inside the opaque-origin iframe). Wired into the
reservations cards (both ReservationCard and TransitJourneyCard) as a strictly-
additive footer keyed by reservation id: zero change to a card when no plugin
contributes. Fetched once per view, fail-safe.
* docs(plugins): bring the permissions wikis current with this cycle
Plugin-Permissions.md was missing every permission added this cycle — add rows for
the read scopes (journal/atlas/vacay/daynotes/collections), the write scopes
(reservations/daynotes/packing, packing noting the #858 owner-scoping), jobs:run
and hook:table-contributor, and correct the events:subscribe row for the new
{ entity, entityId } hint. Add the jobs:run row to Plugin-Development.md too.
* fix(maps): stop quick one-finger pans zooming the map on mobile (#1440)
The global drag-drop-touch polyfill installs document-level touch listeners
on phones. On every single-finger touchend it records a timestamp, and if the
next touch starts within 500ms it synthesises a dblclick on the target, which
the map's default double-click-zoom turns into a zoom-in. Two quick one-finger
pans therefore zoomed instead of panning.
The polyfill only bridges HTML5 drag-and-drop to touch for planner reordering,
which is already disabled on mobile (#1432), so gate its import to viewports
>=1024px (the lg breakpoint useIsMobile uses). Removes the phantom-dblclick
source on phones while keeping touch DnD on large viewports; fixes both the
Leaflet and GL renderers.
* fix(feeds): emit TZID + VTIMEZONE so subscribed calendars respect time zones (#1453)
exportICS emitted timed DTSTART/DTEND as bare floating times (no Z, no TZID),
which iOS/Google Calendar render in the subscriber's local zone instead of the
zone TREK shows. Resolve an IANA zone per timed event — transport endpoints use
their stored timezone (departure drives DTSTART, arrival drives DTEND), while
assignments and hotel/restaurant reservations derive it from place coordinates
via tz-lookup — and attach TZID backed by a VTIMEZONE component. The all-trips
feed now carries deduped VTIMEZONE blocks so TZID references still resolve.
* feat(plugins): render tableContributor contributions in the places + day views
Extends the tableContributor frontend to all three planner views: hoist the shared
PluginCardFooter into PluginContributions, wire the places sidebar (keyed by place
id, rendered as a sibling after each row so the drag/scroll row stays untouched)
and the day panel (keyed by day id, guarded for a null day). Strictly additive +
fail-safe like the reservations view — nothing renders when no plugin contributes.
* fix(vacay): source holiday subdivisions from ISO 3166-2 so all states show
The state/region picker for public-holiday calendars was built from the
union of each holiday's counties for the current year, so a subdivision
only appeared if some holiday that year was tagged with it. States with
no state-specific holiday (e.g. US-WA, and AR/FL/NV/WY in 2026) silently
vanished, blocking calendar creation (#1456).
Source the full, correctly-named subdivision list per country from
ISO 3166-2 instead, merged with any nager county code ISO lacks. Only
region-partitioned countries get a picker, so nationwide-only countries
keep allowing a country-level calendar. No server change needed —
selecting a state already yields federal holidays via applyHolidayCalendars.
* fix(costs): settlements honor custom per-member splits (#1458)
calculateSettlement read each member's custom split amount but its query
never selected budget_item_members.amount, so hasCustomSplit was always
false and every settlement fell back to the equal split. Select bm.amount
so custom amounts drive the balances.
Also blank the Overview 'Per Person' / 'Per Person·Day' columns and CSV
for custom-split items, where a single averaged figure is meaningless.
* feat(plugins): read-convenience + todos + packing bags + tags + roster
A wave of small, high-value capabilities:
- weather:read (ctx.weather.get) — the host's cached forecast, tenant-free
- db:read:categories (ctx.categories.list) — the global place-category list
- db:read:tags / db:write:tags (ctx.tags) — the acting user's own tags, ownership
re-checked before each write
- trips.members (ctx.trips.members) — the trip roster (id + display fields),
membership-checked
- db:read:todos / db:write:todos (ctx.todos) — a trip's to-dos, gated by the app's
packing_edit like the REST path, broadcasts todo:*
- packing bags on ctx.packing (listBags/createBag/updateBag/deleteBag/setBagMembers)
under db:write:packing — no privacy, plain room broadcasts
perm labels (22 locales), both wikis, rpc-host gating + create-rpc-host wiring tests
* fix(dashboard): render next-trip boarding pass stats on Safari (#1459)
The boarding-pass bar carved its ticket-stub notches with a two-layer
radial-gradient mask composited via mask-composite: intersect (and legacy
-webkit-mask-composite: source-in). Safari mis-composites that multi-layer
path to fully transparent, hiding the entire stats bar while Chrome renders
it fine.
Split .hero-pass into an outer wrapper (left notch) and a .hero-pass-inner
glass panel (right notch), each carrying a single-layer mask so the
mask-composite path is never exercised. Renders identically across engines
and degrades safely where mask-image is unsupported.
* feat(plugins): write scopes for atlas, vacay, journal and collections
The write half of the user-scoped addon reads:
- db:write:atlas -> ctx.atlas.markCountry/unmarkCountry/markRegion/unmarkRegion +
bucket-list create/delete. Every row is the acting user's own (visited_countries/
visited_regions/bucket) — no trip scoping, no cross-tenant surface. Unblocks
AirTrail-style two-way sync (#214)
- db:write:vacay -> ctx.vacay.toggleEntry/toggleCompanyHoliday. The plan is
resolved HOST-SIDE from the acting user's active plan — a plugin can never name
another plan, and toggleEntry only toggles the acting user's own PTO day
- db:write:journal -> ctx.journal.createEntry/updateEntry/deleteEntry, self-gated
by journeyService.canEdit (owner/contributor) against the acting user
- db:write:collections -> ctx.collections.create/update/savePlace/copyToTrip/
deletePlace, schema-validated; the service's per-collection role checks
(assertAccess 404 / assertCanEdit 403) map onto RESOURCE_FORBIDDEN
All addon-gated, userless contexts refused, audited. Perm labels (22 locales),
both wikis, gating + wiring tests.
* fix(admin): name the Costs add-on consistently in the catalog
The budget add-on catalog entry still resolved to 'Budget' while the
feature is labeled 'Costs' everywhere else (trip tab, navbar). Align
admin.addons.catalog.budget.name with each locale's trip.tabs.budget
label. Closes#1464
* feat(plugins): file attach, collab content and gated member-add
- db:write:files -> ctx.files.create/createLink/update/softDelete under the app's
separate file_upload/file_edit/file_delete rights. Content arrives as bounded
base64 (10MB decoded cap, well under the app's 50MB), the extension is validated
against the central blocklist BEFORE anything touches disk, and link targets
must live on the same trip (findForeignLinkTarget). Broadcasts file:*
- db:write:collab -> ctx.collab.createNote/createPoll/votePoll/createMessage
under collab_edit + the Collab addon, emitting the same collab:* events as the
app; service-reported errors surface as BAD_PARAMS
- db:write:members -> ctx.trips.addMember. Adding a member GRANTS TRIP ACCESS, so
it is deliberately its own permission behind the app's member_manage right
(default: trip owner only) and never bundled with a lower-risk write; the acting
user is recorded as the inviter, target must exist, owner/duplicate adds no-op
Perm labels (22 locales), both wikis, gating + wiring tests.
* fix(maps): honor check-in/out times for hotel bookend legs (#1465)
The day route drew the accommodation as the day's start/end whenever the
edge stop was a place, ignoring the morningIsSleptHere/eveningIsOvernight
provenance already computed by getDayBookendHotels. On a check-in day an
airport placed before check-in got a spurious hotel -> airport leg, and on
a check-out day a later "home" stop still got a home -> hotel return leg.
Add time-aware shouldDrawMorningLeg/shouldDrawEveningLeg helpers: the
morning leg is the home-base default on a check-in day but is dropped when
the first place is timed before check-in; the evening return leg is off on
a check-out day unless the last place is timed at/before check-out. Wire
them into the map polyline, the sidebar hotel connectors, and the Google
Maps export so all three stay consistent.
* feat(plugins): host-mediated notifications and LLM access
Two host-owned integration primitives — the plugin supplies intent, the host
owns the sensitive part:
- notify:send -> ctx.notify.send({title, body, link?, scope, targetId}). Delegates
to notificationService.send with a new plugin_notification event (raw title/body
carried as passthrough params), so recipient resolution, channel fan-out
(bell inbox + email/ntfy/webhook) and per-user preferences all match core 1:1.
Recipients are FORCED to the acting user (scope 'user', targetId === uid) or a
trip they belong to (scope 'trip'); scope 'admin' refused; the in-app link must
be a relative /path (open-redirect-safe). No arbitrary recipient, no impersonation.
Users can mute plugin notifications like any other event.
- ai:invoke -> ctx.ai.complete(prompt) / ctx.ai.extract(text, jsonSchema). Runs the
admin/user-configured provider via resolveLlmConfig + the existing extraction
client under the acting user; the host holds the (encrypted) key, the plugin
never sees it. Refused when no provider is configured; 20k-char caps. Output is
DATA (complete -> {text}, extract -> {results}) and never auto-written, so
prompt-injection can't reach a write without the plugin's own gated call.
plugin_notification wired through the shared NotificationEventKey + all 22 locales
(inbox passthrough + external channels). Perm labels (22 locales), both wikis,
gating + wiring tests.
* fix(budget): offer every Frankfurter-supported currency (#1470)
The cost currency picker was gated by a hardcoded 47-code list, so
currencies the app can actually convert (OMR, CRC, UGX, MKD, ALL, and
~115 more) couldn't be selected. Replace CURRENCIES/SYMBOLS with the full
set the Frankfurter v2 FX API supports (archived BGN/HRK dropped), unify
the dashboard offline fallback onto it, and teach currencyDecimals about
the newly reachable zero- and three-decimal currencies. A currenciesWith
helper keeps a previously saved (now-archived) selection selectable so it
isn't silently wiped.
* feat(plugins): tableContributor into the costs, packing and files views
Extends the shipped tableContributor hook to three more native views — no new
permission, no new attack surface: the same host-normalized, length-capped,
url-allowlisted (http/https/mailto), enum-bounded, fail-safe pipeline, just more
render sites.
- server: add costs/packing/files to the view-contributions whitelist
- client: widen the ViewName union + the api view type; render PluginCardFooter
keyed by entityId in the budget category table (a colSpan footer row per item),
the packing category group (footer after each item row, drag untouched) and the
files list (footer after each row)
A currency plugin can now drop a converted-amount column onto a cost row, a
receipts plugin a 'view receipt' action onto a file, etc. Controller test asserts
the three new views are accepted; both wikis updated.
* fix(pdf): repeat day header on overflowing itinerary export pages (#1471)
* feat(plugins): map-marker provider hook — plugins can overlay trip-map markers
New declarative provider hook `mapMarkerProvider` (#587 "show bookings on map",
the single most-requested contribution class, with zero contribution point until
now):
- hook:map-marker-provider permission + MapMarkerProvider/MapMarkerContribution SDK
types + HOOK_PERMISSION wiring
- GET /api/map-markers/:tripId (MapMarkersController) mirrors the view-contributions
hardening: membership-gated, providers invoked host->plugin on a 5s timeout,
fail-safe. Every field normalized server-side — coordinates range-checked
(-90..90 / -180..180), strings String-coerced + length-capped, icon/tone enum-
whitelisted, popup url http/https/mailto only (a javascript:/data: url would be
click-XSS), marker count capped at 200 per plugin
- client: PluginMapMarkers layer renders the markers as plain Leaflet Marker+Popup
inside the trip map; plugin JS NEVER runs on the map canvas, every value is
host-vetted data. Threaded tripId through MapView; fail-safe fetch
Declarative-only by design, mirroring placeDetailProvider/tableContributor. Perm
label (22 locales), controller hardening test, both wikis.
* feat(plugins): show page plugins in the mobile bottom nav
Page plugins were reachable from the desktop nav pill (Navbar) but not the mobile
tab bar — you had to type /plugins/:id. BottomNav now reads page plugins from the
plugin store and appends them the same way global addons are, mirroring Navbar.
One-file client nav wiring; no new capability surface.
* feat(plugins): per-user plugin settings form + ctx.settings runtime read
Users can now enter their own per-plugin config (an API key, a preference) —
the prerequisite for almost every real integration, previously unreachable
(scope:'user' settings were only listed read-only in the admin panel).
- migration: plugin_user_config (plugin_id, user_id, config JSON) — each user's
own values, separate from the admin-owned instance plugins.config
- PluginsService.getUserConfig / updateUserConfig / getUserConfigDecrypted +
readUserSettingDecrypted: secrets encrypted at rest (apiKeyCrypto), masked to
the client, an unchanged secret (the mask) keeps its stored ciphertext, and only
DECLARED scope:'user' keys are ever stored
- GET/POST /api/plugin-settings/:id (PluginUserSettingsController) — its own
user-gated path (not the admin surface, not the /:id/* proxy), JwtAuthGuard only,
scoped to the acting user
- runtime: ctx.settings.get(key) -> the acting user's decrypted value (unconditional
RPC, not sensitive cross-tenant; userless job/onLoad gets undefined)
- client: a Plugins tab in Settings host-renders each active plugin's scope:'user'
fields as an editable form (secrets write-only), reusing the declarative field
shape — no plugin markup executes
i18n (22 locales), wiki, rpc-host + service + masking/encryption tests.
* fix(journey): keep skeleton suggestions in sync with linked trip places (#1473)
Journey skeleton suggestions mirror a linked trip's day-assigned places, but
sync relied on scattered per-event hooks that several assignment mutation paths
never called: unassign, move and time-change fired nothing, no remove-on-unassign
capability existed, and every MCP assignment tool synced nothing. Skeletons drifted
from the trip.
Add an idempotent reconcileTripSkeletons(tripId) that re-mirrors the trip's
day-assigned places onto every linked journey (add missing skeletons, refresh
date/time/location on move, remove skeletons for unassigned places; filled entries
are detached + noted, never destroyed). Call it from every REST assignment handler
and MCP assignment tool, and fire onPlaceDeleted on single MCP delete_place for
parity. Extract a shared insertSkeletonEntry helper.
* fix(memories): drop hidden Immich assets so Live Photo motion parts don't show a broken thumbnail (#1474)
* fix(transit): anchor arrive-by search time to the destination timezone (#1479)
* feat(plugins): host-brokered OAuth client + trustworthy inbound webhooks
Two integration primitives where the host owns the sensitive part.
Trustworthy webhooks:
- auth:false routes now receive req.headers, but ONLY an explicit, credential-free
allowlist (the common provider signature/event headers — stripe-signature,
x-hub-signature-256, svix-*, x-gitlab-event, …). Cookie/Authorization/X-Socket-Id
and every session/forwarded-auth header are stripped; authenticated routes get {}.
A plugin can finally verify a provider signature without any way to leak a session.
Host-brokered outbound OAuth (oauth:client):
- the HOST runs the whole flow — authorize -> callback -> token exchange -> refresh —
with PKCE + single-use, user-bound, TTL'd state, and HOLDS the tokens. The client
secret + refresh token never leave the host; the plugin only triggers connect and
reads a short-lived access token via ctx.oauth.getAccessToken() for the acting user.
- provider config (authorize/token url + scopes + client id/secret) is the plugin's
admin-owned instance settings; endpoints must be https (SSRF backstop, private/local
hosts refused). Tokens per-user + encrypted at rest (apiKeyCrypto).
- GET/POST /api/plugin-oauth/:id/{status,connect,callback,disconnect} — JwtAuth-gated,
the callback always redirects to an in-app /settings path (never leaks an error).
- Settings -> Plugins gains a Connect/Disconnect control per configured plugin.
migration: plugin_oauth_tokens + plugin_oauth_state. Perm labels + form strings
(22 locales), both wikis, service (PKCE/state/exchange/refresh/encrypt) + controller
+ proxy header-allowlist + rpc-host gating + create-rpc-host wiring tests.
* fix(navbar): re-measure sliding tab pill after font load and resize (#1481)
The active tab pill was measured once in a layout effect keyed only on activeTab, so on a hard reload it captured the active (bold) label's width against fallback-font metrics and never re-ran when the web font swapped in, leaving the pill slightly offset.
Re-measure after document.fonts.ready resolves and on ResizeObserver changes (container + active button), with an idempotent state update to avoid redundant renders.
* fix(collections): keep the Add-place button reachable after the first save
On a wide/desktop layout the collection toolbar (which hosts the Add
button) was gated on !mapOverlay, so it unmounted as soon as the list
gained its first place with coordinates — leaving only an easy-to-miss
"+" in the map overlay. Keep the toolbar rendered whenever the user can
add a place, and drop the now-redundant map-overlay Add button so there
is a single, predictable Add affordance in every state.
Fixes#1485
* feat(plugins): days + accommodations reads/writes, endpoints on the reservation write path
Community feedback on the 3.2.1 plugin surface: a plugin could write days but
never list them (no way to learn day ids), day_accommodations had no surface at
all, and trips.getReservations was the one reservation read that dropped the
endpoints/day_positions hydration.
- trips.getDays / trips.getAccommodations under db:read:trips (tripRead gate),
wired to the same dayService lists the REST GETs use
- trips.getReservations now returns the hydrated REST-parity list (endpoints,
day_positions, joins, normalized accommodation_id) - strict superset
- new db:write:accommodations scope: ctx.accommodations create/update/delete
gated by day_edit like the accommodations REST path, with the partner-hotel
reservation + delete cascade and broadcasts intact
- reservation create/update pin the endpoints shape up front (BadParams instead
of a mid-transaction NOT-NULL or a silently dropped row)
- perm label in all 22 locales, consent PERM_KEYS, wiki tables
* feat(plugins): day-detail widget slot in the day panel
Widgets can now mount inside the trip planner's day panel
(capabilities.widget.slot: 'day-detail'), scoped to the open day via a dayId in
trek:context - the same pattern as the place-detail slot. Covers the requested
per-day plugin content (logistics, outfit planning, live flight status) without
a new plugin type. Day-detail widgets stay off the dashboard, the consent panel
labels the slot in all 22 locales.
* feat(plugins): let the frame CSP serve a plugin's own static assets
The sandboxed frame runs at an opaque origin, so script-src 'self' never
matched and a plugin's own <script src>/<link> files were blocked - authors had
to inline entire React builds into index.html. Add a scheme-less host-source
pinned to the plugin's own /plugin-frame/<id>/ path (charset-checked Host +
plugin id so a stray token can't widen the policy; malformed Host falls back to
inline-only). Multi-file client builds now load as-is; remote hosts stay
blocked, so script URLs remain useless as an egress channel.
* fix(plugin-sdk): catch the package up to the server capability surface
The npm SDK's validator still knew only the 3.2.1 permission set, so
'trek-plugin-sdk validate' (and pack/publish, which run it) hard-rejected any
manifest using the newer scopes - db:write:reservations, notify:send,
hook:map-marker-provider and 25 more. Sync KNOWN_PERMISSIONS with the server
envelope (48 entries), mirror the full PluginContext (reservations,
accommodations, notify/ai/oauth/settings, packing writes + bags, file writes,
collab, tags/todos/daynotes/collections/atlas/vacay/journal, weather,
categories), type the tableContributor/mapMarkerProvider hooks + the
entity/entityId event hint, accept the day-detail widget slot, and extend
createMockHost so plugin unit tests can exercise all of it.
* feat(plugins): grant-scoped entity snapshots on core events
An events:subscribe handler so far learned only WHICH entity changed - useful
for cache busting, useless for reacting to content, and the userless handler
can't refetch. Now the broadcast tap derives a whitelisted field snapshot of
the changed entity and the supervisor attaches it per plugin, only where the
granted set holds the family's matching db:read:* permission (trips family ->
db:read:trips, budget -> db:read:costs, packing -> db:read:packing, dayNote ->
db:read:daynotes, file -> db:read:files). No acting user is ever synthesized.
The whitelists are explicit per family, so user ids (owner/paid_by/uploaded_by/
participants/members), trips.feed_token and future migration columns never
travel; a private packing item (#858) yields no snapshot at all because its
core broadcast is owner-scoped; deletes/reorders/bulk ops carry none.
* feat(plugins): pdf-section, atlas-layer and journal-entry provider hooks
Three more declarative provider surfaces in the map-marker mould - plugins
return data specs, the host normalizes, caps and renders; a slow or failing
provider contributes nothing:
- hook:pdf-section-provider: sections (title + paragraphs + a simple table)
appended to the trip PDF export, escaped into the same HTML/print pipeline
as the core content
- hook:atlas-layer-provider: per-user country tint layers on the Atlas map
(ISO 3166-1 alpha-2 codes only, tone-whitelisted, non-interactive pane so
mark/unmark clicks keep working)
- hook:journal-entry-provider: extra rows on a journal entry card, gated by
the same journey access check as the journal routes + the Journey addon
Permission labels in all 22 locales, consent PERM_KEYS, SDK types + manifest
validator in both SDK copies, wiki tables, per-controller hardening tests.
* feat(plugins): trip-page plugins can replace core planner tabs and pick their spot
A trip-page plugin that takes over a core surface (a transit planner
superseding Transports, a costs plugin superseding the budget tab) had to sit
awkwardly next to the tab it replaces. capabilities.tripPage now names the
core tabs to hide while the plugin is active - whitelisted (transports,
buchungen, listen, finanzplan, dateien, collab), 'plan' deliberately not
replaceable, and the tabs return the moment the plugin is deactivated - plus
an optional 0-based position for the plugin's own tab. The feed re-validates
the values out of the DB blob so a hand-edited row can't hide anything else,
the admin list chips a replacing plugin (all 22 locales), and a saved session
tab that got replaced falls back to the plan view.
Also fixes the plugins feed dropping the day-detail widget slot to 'sidebar',
which would have mounted a day-panel widget on the dashboard.
* fix(plugins): audit follow-ups — normalization, secret cleanup, cron leak, slot filter
Adversarial audit of the whole plugin PR surfaced 12 confirmed issues; this
addresses them:
- place-details provider was the ONE hook controller with no normalization: a
plugin's href/label/value went to the client raw and unbounded. Now normalized
like journal-entry-rows (safeUrl http/https/mailto, length + count caps).
- trip-warnings capped message length + per-provider count (was unbounded).
- uninstall(deleteData) now also purges plugin_user_config, plugin_oauth_tokens,
plugin_oauth_state, plugin_meta_migrations and the capability audit — encrypted
per-user API keys + OAuth refresh tokens no longer survive a 'delete all data'
and get silently re-adopted on a same-id reinstall.
- supervisor: a crash-restart cycle leaked the dead child's node-cron tasks and
re-scheduled fresh ones, so a job fired N+1 times per tick after N crashes.
onExit now stops them, mirroring kill().
- dashboard sidebar no longer mounts place-detail/day-detail widgets (they belong
in the planner panels).
- reservation endpoint validation relaxed to match the 3.2.1 service: a coord-less
endpoint is accepted and dropped downstream instead of BadParams (no breaking
change), while a bad role/non-string still rejects up front.
- a replaced core tab reached by programmatic nav now falls back to the plan view.
- trips.update caps title/description like the places path; plugin-db guard bans
load_extension as defense-in-depth.
- wiki: event snapshots, string-typed context ids, dayId in the payload, the live
provider hooks and the costs update/delete grant are now documented correctly.
* feat(plugins): phase-0 lifecycle hardening + per-plugin RPC rate limit
Operational-readiness fixes from the completeness audit:
- Re-activation after a failure worked again: a plugin left in 'error' state by
a load-failure or crash-auto-disable stayed in the running map, so the admin's
'enable' button was a silent no-op. activate() now replaces a dead entry.
- Per-plugin RPC rate limit at the dispatch boundary: every ctx.* call runs
synchronously on the host thread, so a plugin in a tight loop could freeze the
whole instance (and the reap sweep). A token bucket (generous burst) + an
in-flight cap now throttle a runaway plugin with a retryable HOST_ERROR; a
legitimate plugin never notices.
- plugin_error_log retention (500 rows/plugin) so a crash-looper can't grow
trek.db without bound; the crash-timestamp array is trimmed to its window too.
- TREK_PLUGIN_PERMISSIONS=off now logs a loud one-time warning that the OS
permission jail is disabled.
* feat(plugins): read symmetry + broker — collab/journal/atlas reads, file content, trip create, rates
The plugin API leaned write-heavy: collab and journal could be written but not
read, files listed but not read, and there was no way to create a trip or see
exchange rates. This closes those gaps in the established RPC+gate pattern (zero
architecture risk), and it's what unlocks the importer + finance plugin classes:
- collab reads: ctx.collab.listNotes/listPolls/listMessages under a new
db:read:collab (membership + Collab addon, like the REST GETs)
- ctx.journal.getEntries(journeyId): a journey's entries, journey-access-checked,
under the existing db:read:journal
- ctx.atlas.bucketList(): the acting user's bucket list, under db:read:atlas
- ctx.files.getContent(tripId, fileId): a file's bytes as base64 under a NEW
db:read:files:content grant (reading a passport scan is more sensitive than its
filename), size-capped at 10MB before it crosses the IPC pipe, trashed files
refused
- ctx.trips.create(input): a new trip owned by the acting user, gated by the app's
trip_create right + a bound user — the capability importers need
- ctx.rates.get(base): cached currency exchange rates, tenant-free like weather
Also caps trips.update title/description like the places path, and the plugin-db
guard now bans load_extension (defense-in-depth). SDK, mock-host, i18n (22
locales), consent labels and the wikis are all in lockstep.
* feat(plugins): deeper integration + user-facing activity transparency
Wave 2 of the completeness work — richer extension points, deeper metadata, and
the transparency that makes the broad read grants accountable:
- db:meta now attaches to reservations + accommodations too (not just
trip/place/day), gated by reservation_edit / day_edit respectively — the
natural home for an external-id mapping (AirTrail/calendar/booking-import sync)
without forking the core schema.
- reservation-detail widget slot: a widget can mount on a booking card, scoped to
the open reservation via reservationId in trek:context (the place-detail /
day-detail pattern, third instance).
- tableContributor gains the transports + todos views, so a plugin can add
host-rendered columns/actions there too.
- User activity log: GET /api/plugin-activity + a Settings → Plugins panel showing
every host-mediated action a plugin took bound to the signed-in user, across all
plugins, newest first — the user-facing half of the hash-chained audit. This is
what legitimizes the deliberately broad read grants: not just the admin, the
person whose data is read can see what was done in their name.
- DX: the local dev server now binds a default acting user, so the canonical
ctx.trips.getPlaces(tripId) call works locally instead of failing RESOURCE_
FORBIDDEN; the create scaffold drops the dead manifest routes[] / capabilities.nav
fields the host ignores.
SDK, i18n (22 locales), consent labels and the wikis are all in lockstep.
* fix(memories): load Immich album photos on Immich v3
Immich v3 removed the `assets` property from AlbumResponseDto, so
`GET /api/albums/:id` no longer carries album contents. TREK read album
photos from that property, which now parses as undefined and degrades to
an empty array — hence "No photos yet" in the Journey gallery picker even
though the album header shows the right count (that count comes from
`GET /api/albums` -> assetCount, which v3 still returns).
Two call sites read the removed property. Besides getAlbumPhotos (the
reported bug), syncAlbumAssets failed silently on v3: it reported
`success: true, added: 0` while syncing nothing.
Fetch album contents via an `albumIds`-filtered `POST /api/search/metadata`
when `assets` is absent, and feature-detect rather than probe a version.
The two paths are not interchangeable: on v2, searchMetadata
unconditionally scopes results to `[self, ...partners]`
(`asset.ownerId = ANY(userIds)`), so an albumIds search against an album
shared by a non-partner returns nothing. v3 added an albumIds branch that
checks AlbumRead and skips that owner filter. v2 also hard-defaults
`visibility` to `timeline`, dropping archived assets. So v2 must keep
reading the album detail body, which this preserves exactly.
`withExif: true` is required on the search path: it has no default and
gates an inner join, so without it Immich omits `exifInfo` entirely and
every photo's city/country goes null.
The existing test mock returned an album detail body *with* `assets` — it
encoded the v2 assumption, which is why this shipped green. It now models
v3 by default, with explicit v2 coverage asserting no search call is made.
Fixes#1492
* feat(plugins): daily AI/notify budgets, runtime scheduler & reliable event redelivery
Per-plugin daily caps on ai.complete/ai.extract and notify.send (defaults
200 / 100, overridable via TREK_PLUGIN_AI_PER_DAY / TREK_PLUGIN_NOTIFY_PER_DAY),
seeded from the capability audit so a mid-day restart resumes the count instead
of resetting it. Surfaced at GET /plugins/:id/budget.
ctx.scheduler (at / in / every / cancel): persistent, userless timers that
survive restarts and fire a scheduled() handler, riding the existing jobs:run
grant so no new consent or admin setup is needed. Backed by
plugin_scheduled_tasks, swept every 30s, capped at 100 tasks/plugin with an 8 KB
payload and a 60s recurring floor; rows are removed on uninstall.
Core events that fire while a subscriber is mid-restart are now held in a
bounded in-memory buffer (200/plugin, 15 min TTL) and replayed once it goes
active again, with the events:subscribe grant and snapshot gating re-evaluated
at replay time so nothing leaks if a grant was revoked while the plugin was down.
* feat(plugins): GDPR data-subject rights — durable per-plugin erasure + export
New hook:user-data grant with two userless lifecycle handlers a plugin can put
on its definition: deleteUserData and exportUserData. Neither carries an acting
user — the plugin only learns the userId and touches its own db — so the grant
reads nothing from core data; it exists purely so a plugin can honour a GDPR
erasure or data-access request.
When a TREK account is deleted (admin or self-service), every installed plugin
holding the grant gets a row in a new durable erasure queue and its
deleteUserData runs on the next sweep, retried until it ACKs — so erasure
survives the plugin being offline or the server restarting. The core deletion
path notifies the runtime through a dependency-free relay (like the event sink),
keeping the auth/admin services decoupled from the plugins layer, and a plugin
bookkeeping error can never fail the account deletion.
Portability is served by GET /api/admin/plugins/user-data/:userId/export, which
fans exportUserData out to the active granted plugins and aggregates what each
holds about the user. Queue rows are purged on uninstall; the grant is labelled
in all 22 locales.
* feat(plugins): atomic ctx.db.tx for consistent multi-write on a plugin's own db
Plugins could already query/exec/migrate their own SQLite file, but a multi-step
write (move an item between tables, decrement one row and increment another) had
no way to be atomic. db.tx([{sql, args?}, …]) runs up to 100 statements in a
single transaction — all commit or all roll back — and reads within the batch see
its own earlier writes, so read-modify-write is safe. Each op is one statement:
a read returns { rows }, a write { changes }. The same guard (no ATTACH/PRAGMA/
RECURSIVE, size + row caps) applies to every statement in the batch.
* fix(memories): filter hidden Immich assets at the source, not just the picker
#1474 has the same root cause as #1492: the Immich v3 migration. On v2,
searchAssetBuilder hard-defaulted metadata search to `timeline` visibility
(`visibility = options.visibility ?? Timeline`), so hidden Live Photo
motion parts could never come back from a search. v3 defaults to any
visibility except `locked`, so they do — which is why the reporter is on
Immich 3.0.1 and why the bug never appeared before.
Ask for `visibility: 'timeline'` explicitly on the search path. That
restores v2 semantics on both versions and stops hidden assets crossing
the wire, which also fixes a pagination wart: a full page half-made of
motion parts previously rendered as a half-empty page, because hasMore
counts the raw page length while the filter shrinks the rendered set.
The client-side filter was display-only, applied in searchPhotos and
getAlbumPhotos — both picker-listing paths. Nothing guarded persistence
or rendering: getOrCreateTrekPhoto stores any id it is handed, pipeAsset
forwards Immich's 400/404 verbatim, and the photo grid is a plain <img>
with no onError. So syncAlbumAssets, which filtered `type === 'IMAGE'`
only, could persist a hidden IMAGE as a permanently broken tile. It now
applies the same guard, extracted as isVisibleAsset().
Albums keep their filter rather than requesting `timeline` visibility:
albums legitimately contain archived assets, and both the v2 album body
and the v3 album search return them.
Does not address tiles already persisted before this — those still render
broken and need a separate fix.
Refs #1474
* docs(memories): correct Immich version boundaries in the hidden-asset comments
Verified against the v1.120.0 → v3.0.0 OpenAPI specs and server source. The
previous comments said "Immich v2 hard-defaulted metadata search to timeline
visibility". That is true only for 1.133–1.144.
- `visibility` was added in 1.133.0. Before that, searchAssetBuilder applied
`.$if(options.isVisible !== undefined, ...)` with no default, so pre-1.133
servers returned hidden assets too. #1474 was therefore not purely a v3
regression.
- Those servers strip the `visibility: 'timeline'` filter rather than
rejecting it: Immich validates with `whitelist: true` and no
`forbidNonWhitelisted`. So the request stays valid, the filter is a no-op,
and isVisibleAsset() is the ONLY guard there. Say so, so it does not get
removed later as redundant.
- `albumIds` only exists from 1.135.0. Because unknown properties are stripped,
an albumIds search against an older server would silently drop the album
filter and return the entire library as the album's contents. Feature
detection on `assets` (present through 1.144.1, absent on v3) makes that
unreachable; a version probe with a wrong boundary would not.
Also cite Immich's own enum, which documents AssetVisibility.Hidden as
"Video part of the LivePhotos and MotionPhotos".
Comments only — no behavior change.
* feat(plugins): dashboard trip-card badges + a mock-host driver for plugin tests
Two additions that round out the plugin platform's breadth and its authoring DX.
tripCardProvider hook (hook:trip-card-provider): a plugin returns small declarative
badges for the dashboard trip cards. The dashboard fetches all visible cards in one
call; the host access-checks every tripId for the acting user, bounds each field
(label/value length, enum tone, http/https/mailto-only url), caps the count and drops
any badge for a card that wasn't requested — plugin JS never runs on the dashboard.
Rendered as text chips under the card meta; labelled + gated in all 22 locales.
createMockHost now exposes run(def) — the other half of a plugin unit test. Where the
ctx recorders capture what a plugin read, run() fires its own entry points (route, job,
scheduled, event, plugin-event, deleteUserData, exportUserData, provider hooks) against
the same mock ctx, and host.scheduled surfaces the timers it armed. A handler the plugin
didn't declare throws a clear error instead of a silent no-op.
* feat(plugins): include plugin data + code in backups, applied on restart
A TREK backup archived travel.db + uploads + the encryption key, but each plugin's
own SQLite file — the ONLY copy of the user data it holds — and its installed code
lived in separate trees that were never captured, so a restore left the plugins rows
with no data or code behind them.
createBackup now adds plugins-data/ (each plugin's db + WAL sidecars, so SQLite
recovers a consistent snapshot) and plugins-code/ (skipping dev-links by realpath, so
an author's linked source is never bundled). Restore can't swap those live — the
runtime holds each plugin db open — so it STAGES the extracted trees beside the live
ones and the runtime swaps them in at the next boot, before it opens anything. Same
"applies on restart" model the bundled encryption key already uses: no plugin quiesce,
no swap under open handles, no new admin setup. Older archives without the trees restore
exactly as before.
* fix(plugins): audit — runtime robustness, security & data-lifecycle fixes
Fixes from an adversarial audit of the plugin system, host/runtime side:
Robustness:
- getPluginDataDb recreated a handle a terminal-failure dispose had closed but
left cached, so a re-enabled plugin's db:own threw on every call — recreate
when the cached handle is shut.
- ctx.ws.broadcast* now carry _inv, so the host can bind the acting user (the
capability was silently refused, i.e. dead, without it).
- ctx.events.emit swallows a rejected emit instead of crashing the child into a
terminal 'error'; an uncaught throw AFTER activation is treated as a crash
(restart with backoff), not a load failure.
- A crash-respawned child gets the same activation deadline as a first activation
and the buffered-event queue is cleared on the timeout path, so a hung onLoad
after a crash can't peg a core and orphan events forever.
- Expired buffered events are pruned by the reaper, not only at flush; the
scheduler + erasure sweeps scope their LIMIT window to ACTIVE plugins so a
backlog for inactive plugins can't starve deliverable work.
Security / integrity:
- Unix-domain-socket / named-pipe connects are refused by default in the egress
guard (a host-local pivot to docker.sock / DB sockets), under the same policy
as private IPs.
- db.tx refuses transaction-control statements (a raw COMMIT would break its
atomicity) and caps rows across the WHOLE batch, not per statement.
- plugin_capability_audit is retention-capped per plugin (chain-safe: retained
rows stay self-verifying), so it can't grow unbounded in the shared db.
- A cap of 0 in TREK_PLUGIN_AI_PER_DAY / _NOTIFY_PER_DAY now disables the broker
instead of falling back to the default.
GDPR data lifecycle:
- Account deletion now erases host-side per-user plugin tables (config, OAuth
tokens/state) and enqueues the own-db erasure from the CORE path, so it works
even when the runtime is disabled or pre-boot; guest deletion does the same.
- uninstall keeps a pending erasure when data is retained (deleteData=false);
erasure delivery is no longer grant-re-checked (a queued erasure is a duty);
export flags installed-but-inactive plugins as pending instead of omitting them.
Backup/restore:
- Plugin DBs are WAL-checkpointed before archiving (no torn/stale snapshots).
- Restore applies the staged trees immediately by quiescing the plugins (no
unbounded gap where a later unrelated restart would revert diverged data);
the swap is content-level (safe on a volume-mounted root) and preserves
dev-links; the decompressed-size cap is operator-raisable.
* fix(plugins): audit — hook-output hardening, dashboard slot & mock-host parity
- Map-marker and atlas-layer tones were validated on String(tone) but emitted
raw, so a non-string tone (an object with a matching toString) slipped through
and crashed the client that renders it — check the raw value against the enum.
- View-contribution column/action caps are now PER ENTITY, not per view, so a
plugin's columns no longer vanish from every table row past the first 20; the
dashboard trip-card badge cap is per card (≥ one on every visible card).
- A reservation-detail widget no longer also renders as a context-free dashboard
sidebar card (the inline filter was missing that slot).
- mock-host matches the real host: it ignores asUserId on trip reads (bind the
acting user), throws on a wrong user-scope notify target instead of coercing,
enforces the scheduler caps, and detects RETURNING as a read in db.tx — so a
passing author test can't hide a production RESOURCE_FORBIDDEN.
* feat(plugins): full ctx parity in the dev server + fire jobs/events/hooks locally
The trek-plugin dev server injected only ~6 of the ~35 ctx areas, so any plugin
touching ctx.costs/packing/files/notify/ai/settings/scheduler/meta/oauth/db.tx/…
hit a TypeError in local dev while the same code passed mock-host tests and worked
installed. It also could only exercise routes.
Delegate every non-db-own capability to a grant-enforcing mock host (the same one
unit tests use) while keeping the real node:sqlite for db:own and dev-native ws
capture + logging — so the whole surface works in dev with the exact production
permission rules. dev-fixtures.json now takes the createMockHost options shape, so
you can seed the full surface. New GET /__dev/fire/<kind>[/<name>][/<fn>] fires a
job, scheduled timer, event subscription, GDPR handler or provider hook against the
dev ctx, closing the "can't test non-routes locally" gap.
* feat(plugins): wire the photoProvider + calendarSource hooks to real core consumers
Both hooks were declared, typed and documented but NO core code ever invoked them,
so an author could build, mock-test and install a photo or calendar plugin that
silently did nothing. Give each a real consumer that fans out to it, exactly like
the other eight provider hooks:
- GET /api/plugin-photos/search (+ /sources, /item) aggregates photoProvider results
for the picker — {id, title?, thumbnailUrl, fullUrl, takenAt?}, thumbnail/full URLs
http/https-only (they become <img src>), per-source count capped, failing source
skipped.
- GET /api/plugin-calendar?start=&end= aggregates calendarSource events for the
signed-in user — {id, title, start, end, allDay} ISO, count capped, failing source
skipped, sensible default window.
Both run with the acting user bound. The SDK interfaces now pass ctx as the last arg
(so a source can reach ctx.settings/oauth/http for its backend), and the wiki marks
them live instead of "reserved — no core consumer".
* feat(plugins): close the create-heavy API asymmetries importers/sync hit
Core services implemented these but plugins had no path to them, so the flagship
importer/sync integrations hit real walls. Added, each reusing the EXISTING grant
(no new consent):
- ctx.trips.removeMember(tripId, userId) — reconcile DEPARTURES, not just additions
(db:write:members + member_manage). Never removes the owner (that would orphan the
trip); ownership transfer stays a separate deliberate action.
- ctx.journal.createJourney({title, subtitle?, trip_ids?}) / deleteJourney(journeyId)
— an importer can now bootstrap the journal it fills with entries and clean it up
(db:write:journal), instead of only appending to journals a human created first.
Wired end-to-end (envelope → rpc-host → create-rpc-host reusing tripService/
journeyService → both SDK copies → mock-host) and documented. (trips.delete needs its
own destructive permission + consent copy and collab edit/delete + collections.delete
remain — tracked as small follow-ups.)
* feat(plugins): strip emojis from plugin-rendered text so it matches TREK's lucide UI
Plugin authors (especially AI-generated ones) sprinkle emojis into the declarative
text TREK renders in its OWN chrome — hook contributions (badges, columns, warnings,
PDF sections, map-marker/atlas labels, journal rows, place details, trip-card badges,
calendar + photo titles) and notifications — which clashes with TREK's lucide-only icon
language.
A shared stripEmoji() removes emojis (incl. flag/ZWJ/variation-selector sequences) and
tidies the leftover whitespace, applied at the render boundary in every hook-contribution
normalizer and in notify.send — so no matter what a plugin returns, the text TREK draws
stays emoji-free. It does NOT touch a plugin's own sandboxed /ui frame (the author's to
design), and it leaves photo ids verbatim (they round-trip to getById). The validate CLI
warns when a manifest name/description contains emojis, nudging authors to the declarative
`icon` field (a lucide name) instead.
* fix(plugins): harden the restore-apply path — regressions from the backup/dev fix pass
A final audit of the fix pass caught three regressions clustered in the two newest
surfaces; the restore path could both crash the server and destroy data.
- CRITICAL: a restore quiesces plugins via supervisor.shutdownAll() AFTER closeDb(), but
shutdownAll killed children without first marking them stopped, so each child 'exit'
took the CRASH path and wrote crash-accounting rows into the now-closed core DB — the
throw escaped an EventEmitter listener as an uncaughtException and killed the whole
process mid-restore. shutdownAll now marks every entry stopped and drops it from
`running` BEFORE the kills (so onExit early-returns), and the onStatus/onLog DB hooks
are wrapped in try/catch (also covers the stderr→onLog path). This also stops a normal
shutdown from logging phantom "crashed" rows.
- HIGH: swapContents cleared live entries then MOVED staged ones in, so a crash mid-move
permanently deleted a plugin's only data copy (staging was already emptied, so a retry
couldn't restore it). It now COPIES each staged entry over the live one and only deletes
staging at the very end — `staged` stays the complete source of truth, making the whole
operation crash-idempotent.
- HIGH: the dev server lost the actingUserId=1 default in the mock-host refactor, so a
fresh scaffold refused every user-bound capability. Restored.
* fix(plugins): final-audit medium/low findings
- GDPR export flags an active plugin whose export errored/timed out as `pending`
instead of silently omitting it (collectUserExport now returns a discriminated
result), so a data-access export never reads complete while missing data.
- Account deletion also enqueues an erasure for plugins UNINSTALLED with retained
data (an orphan data dir) — a same-id reinstall now honours the deletion instead
of re-adopting the user's data forever.
- oauth.getToken returns null in a userless context (matching the SDK/mock contract)
instead of throwing RESOURCE_FORBIDDEN a background caller can't handle.
- Crash-backoff restart is identity-guarded (+ the timer is tracked and cleared like
the activation timer), so a disable + re-enable during the backoff window can no
longer respawn a ghost child from the replaced entry.
- db.tx transaction-control guard strips leading comments first, so `/* */COMMIT`
can't slip past the start-anchored check and break batch atomicity.
- createJournal inherits its cover only from a trip that was actually LINKED
(access-checked), closing a cross-tenant cover-image read on plugin + REST paths.
- trip-warnings drops a null array element instead of losing ALL of that provider's
warnings; plugin-activity floors a non-integer ?limit so it can't 500.
- The trek-plugin dev server binds loopback only and refuses cross-site requests to
its side-effectful /__dev/fire endpoints (it serves real routes + no-auth dev
actions).
* fix(plugins): clear no-misleading-character-class in the emoji stripper
The character class listed the ZWJ, variation selectors and combining keycap
marks as members, which eslint reads as an accidental combined grapheme and
rejected on CI. Pull the emoji glyphs out into Extended_Pictographic /
Regional_Indicator alternatives so only the joiner/selector code points stay in
the class, with a scoped disable where the rule still can't tell them apart.
While here, reset lastIndex before the /g regex is reused in hasEmoji() so a
second call can't resume mid-string and miss a leading emoji.
* fix(security): trip-scope note-file deletion and guard the LLM base URL
Two reported issues:
- deleteNoteFile only matched on the note id and file id, so a member of trip A
could delete a file attached to a note in trip B by guessing its id. Thread the
trip id through the service and controller and scope the delete to it, the way
every other collab operation already does.
- The LLM extraction clients fetched the user-configured base URL directly, so a
user could point it at the cloud-metadata endpoint (169.254.169.254) and read
the echoed error body. Route both clients through a new safeFetchLlm() that
blocks the link-local/metadata range while still allowing a local or LAN Ollama
(loopback and private ranges stay reachable), pinned to the resolved IP so a
hostname can't rebind to the metadata address after the check.
* fix(security): route every LLM client through the SSRF guard
The base-URL SSRF fix covered the openai-compatible and anthropic clients but
missed the native Ollama /api/chat client and the /api/tags + /api/pull model-
management calls, whic…
* fix(plugins): repair plain-HTTP egress and forward the private-egress opt-out
Two pre-existing bugs in the plugin egress guard, found by running a plugin
against a real service end to end.
1. Every plain-HTTP request a plugin made was refused, whatever host it had
declared. Node pre-normalises `net.connect()` args into an [options, cb]
array and passes THAT array as the single argument; undici's plain-HTTP
connector takes this path, its TLS connector does not. classifyConnect read
`host` off the array, got undefined, and fell back to 'localhost' — so a
fetch to a declared, public host was rejected with the nonsense message
"localhost is not in the plugin's declared hosts". It failed closed, so it
was never a security hole, and it went unnoticed because the only shipped
egress plugin uses HTTPS. unwrapConnectArgs() unwraps the normalised form
before anything reads host/path.
2. TREK_PLUGIN_ALLOW_PRIVATE_EGRESS could never have any effect. The guard that
reads it runs INSIDE the child, whose env is scrubbed to a four-entry
whitelist that never included it — so a documented setting (wiki/
Environment-Variables.md) was wired to nothing, and no plugin could reach a
self-hoster's LAN service no matter what the operator set. Forwarded only
when set, so the default stays the secure block-private policy.
Regression tests cover the normalised form in both directions: the real host is
now resolved, and an undeclared host, a private IP and a unix socket are all
still refused when passed that way.
* feat(notifications): let a plugin register a notification channel
TREK's four channels (in-app, email, webhook, ntfy) were a closed set:
notificationService.send() dispatched with four copy-pasted `if` blocks and no
provider abstraction, so a fifth channel meant editing eight files by hand. A
plugin could produce a notification via ctx.notify.send(), but never deliver
one.
A plugin now registers a channel with `hooks.notificationChannel` +
`hook:notification-channel` on a plain `type: 'integration'` — not a new manifest
type, so the TREK-Plugins registry schema and both its CI gates are untouched.
Core refactor
- New channel registry (services/notifications/): email/webhook/ntfy become
ExternalChannel providers wrapping the EXISTING send functions — no delivery
logic is rewritten, only relocated. In-app deliberately stays out: it writes
typed rows with scope/target/callbacks, not a rendered title+body, the same
line shared/ already draws with i18n/externalNotifications.
- The event text is now rendered once per recipient instead of once per channel.
- The channel set is open: NotifChannel becomes a string, the matrix is
registry-derived, and the UI columns are server-driven. The DB column was
already bare TEXT and the Zod contract already a string record — only the
TypeScript and the two UIs were ever closed.
The hook runs USERLESS. Every other hook is user-initiated, so actingUserId falls
out of the request; a notification is host-initiated for an ARBITRARY recipient,
so ctx.settings.get() would return undefined. The host resolves the recipient's
decrypted scope:'user' settings itself and passes them as an argument. That is
what lets a channel plugin be handed someone's push token WITHOUT being handed
the right to read their trips as them.
Enabling the plugin is the opt-in: a plugin channel is not gated on the admin's
`notification_channels` list. A built-in always exists in code and needs an
explicit switch; a plugin channel only exists because an admin enabled that
plugin. (Nothing could write a `plugin:` id into that CSV anyway, and the admin
toggle rebuilt it from three booleans, silently dropping anything else — so
requiring a second opt-in meant the channel could never be turned on at all.)
Also fixed, found while building this:
- Plugin settings keys were unvalidated, so a field named `__proto__` or
`constructor` resolved off Object.prototype: a REQUIRED field with such a name
reported as configured for every user who had configured nothing — enough, for
a channel, to be dispatched to everyone with no credentials. Keys are now
constrained at install and the config blob is parsed null-prototype, so it is
impossible even for an already-installed plugin.
- A `select` field's options were cast straight through, so the obvious
`["1","5"]` form rendered every dropdown entry BLANK (the client reads
value/label). Now coerced, and malformed options are rejected.
Also adds: operator-supplied egress hosts (a plugin talking to a self-hosted
service can't name the operator's host at publish time, so an admin adds it
post-install and the runtime re-spawns the child with the widened allow-list —
only for a plugin that DECLARED operatorEgress, and only an admin, never a user);
settings-page actions (a "Test connection" button, user-initiated so
ctx.settings.get() returns the clicking user's own value); and a Gotify-shaped
notification-channel template in the SDK.
Verified end to end against a real Gotify container, not just in tests.
* docs(wiki): document the plugin notification-channel surface
Covers the pieces added in the previous commits, in the pages a reader would
actually reach for:
- Plugins.md (the admin-facing page) had none of it: notification channels,
settings actions, and a full "Allowed hosts" section — including what
operator-supplied egress deliberately does NOT let anyone do.
- Plugin-Development.md: the notificationChannel hook (and why it is the one hook
with no acting user), settings-page actions, operatorEgress, and the manifest
reference rows.
- Plugin-Cookbook.md: a "become a notification channel" recipe and a
"Test connection button" recipe.
- Plugin-Permissions.md: hook:notification-channel, operatorEgress under the
outbound section, and settings actions under "not a permission".
- Notifications.md: plugin channels alongside the four built-ins.
* fix(sdk): allow empty egress if and only if operatorEgress is true
* ci: don't run repo-specific workflows on forks
Guard release, publish, wiki-deploy and issue/PR-triage workflows with a
`github.repository` check so they no-op in forks instead of failing or
acting on the fork's own issues, PRs, tags and registries.
Also skip the Docker Scout scan for pull requests from forks: Docker Hub
secrets are never exposed there, so the login step could not succeed.
Tests and lint stay ungated — they need no secrets and are the gate for
incoming fork PRs.
* feat(sdk): add missing methods in mock-host
* fix(airports): rebuild the json file
* fix(airports.json): add small airports too
* fix(public transit): only show public transit option when a trip has actual dates
* fix(plugins): reap a queued erasure only once the plugin's data is gone
The orphan reap deleted every queue row whose plugin had left the registry, but
uninstall(deleteData=false) removes the plugins row while deliberately keeping the
data dir AND the queued erasure so a same-id reinstall can still honour it. The reap
now deletes a row only when the plugin's data dir is actually gone; a deleteData=true
uninstall already clears the rows itself.
* fix(backup): snapshot the core DB and swap restores atomically
createBackup archived travel.db via the archiver's lazy live-file read, so a WAL
auto-checkpoint firing mid-stream could write a torn database into the zip. It now
VACUUM INTOs a point-in-time snapshot and archives that, the same guarantee plugin
DBs already get. restoreFromZip swapped the DB by unlink-then-copy, which on an
interrupted restore could leave no valid travel.db; it now copies to a temp file and
renames it into place (atomic), dropping the stale -wal/-shm sidecars first.
* fix(deploy): Recreate strategy for the SQLite volume, pin the root compose image
The Helm Deployment had no strategy, so the default RollingUpdate would start a second
pod holding the same ReadWriteOnce PVC before the old one exits — a Multi-Attach
deadlock or two writers on one SQLite file. Default to Recreate (overridable for
ReadWriteMany). The root docker-compose.yml pinned trek:dev, a tag no workflow builds,
so a clone-and-up at the release tag ran a stale image; pin it to :latest like the README.
* fix(security): re-validate LLM endpoint fetch redirects per hop (GHSA-fmq9)
safeFetchLlm left undici's default redirect:'follow', so a configured LLM
endpoint could 302 to http://169.254.169.254/ and reach cloud-metadata
credentials — the DNS pin does not cover an IP-literal redirect hop, since
net.connect skips the pinned lookup for a literal IP. Follow redirects
manually now, re-resolving/re-checking/re-pinning each hop (allowing LAN/
localhost as before). Also block the Alibaba metadata IPs directly.
* fix(plugins): throttle the plugin log channel to prevent host-thread DoS
The per-plugin RpcRateLimiter only guarded the ctx.* (req) channel; ctx.log.*,
stdout/stderr and unknown evt topics reached a synchronous INSERT+prune on the
host thread unthrottled, so a while(true) ctx.log.error(...) loop could freeze
the instance. Route every plugin-driven log path through a per-plugin log token
bucket; excess lines are dropped with a summary line on resume.
* fix(plugin-sdk): serve dev /ui frame at /ui/index.html so relative assets resolve (#1526)
The dev server embedded the plugin UI as <iframe src="/ui"> (no trailing
slash), so a multi-file build's relative asset URLs (./assets/x.js from Vite
base:'./') resolved against the origin root -> /assets/x.js -> 404, even though
the files are served at /ui/assets/*. The real host loads the frame at
/plugin-frame/<id>/index.html where the same relative URLs resolve correctly, so
dev now matches it by loading /ui/index.html (and also serves /ui/ as index.html).
* i18n: improve Russian translations (#1539)
* v3.4.0 (#1527)
* fix(plugins): unknown column
* fix(mcp): reuse MCP sessions instead of creating one per tool call
The /mcp CORS layer never set exposedHeaders, so Access-Control-Expose-Headers
was absent and browser-context MCP clients (Claude Desktop connectors,
Claude.ai, MCP Inspector) could not read Mcp-Session-Id off the initialize
response. Unable to echo it back, every request looked like a fresh initialize:
one McpServer and one session per tool call, until the per-user cap returned a
429 and the integration died until the container was restarted.
The idle sweep was not at fault — it expires on lastActivity with a 1h default,
so sessions born seconds apart are nowhere near expiry, hence 'cleaned 0'.
- expose Mcp-Session-Id, MCP-Protocol-Version and WWW-Authenticate
- evict a user's least-recently-active session at the cap rather than
refusing the request, so a client that cannot persist its session id (or a
proxy that strips the header) can never wedge the server
- close the McpServer/transport orphaned by every session-less non-initialize
POST, which was leaked: never mapped, never swept, never closed
- return the cap error as JSON-RPC so clients surface the real reason
- warn on session-less POSTs to make a header-stripping proxy diagnosable
* chore(deps): declare @modelcontextprotocol/sdk ^1.29.0
Matches the version already resolved in the lockfile; no dependency-tree change.
* docs(mcp): add the reverse proxy specs for MCP
* fix(plugins): stop the row ⋯ menu being clipped by the sidebar
The menu was an in-flow `absolute` div, and its ancestor (PageSidebar) is
`overflow-hidden` — which clips absolutely-positioned descendants regardless
of z-index. With enough plugins installed a row sits low enough that its menu
runs past the sidebar's bottom edge and gets chopped, taking Delete with it,
so the plugin could no longer be uninstalled from the UI.
Portal the menu to <body> and position it `fixed` against the ⋯ button,
flipping upward when the bottom is tight and re-anchoring on scroll/resize.
A fixed child of <body> has no overflow ancestor, so nothing can clip it.
Closes#1523
* feat(plugins): surface author-signature status and add a scoped re-trust override
TREK has always verified an author's Ed25519 signature and TOFU-pinned the key on
first install, but none of it was ever shown: a successfully-installed UNSIGNED
plugin looked identical to a signed one, and a signature-refused update left the
plugin quietly pinned at its old version with the reason dying in a toast.
Give the four refusal conditions machine-readable codes (SIGNATURE_MISSING /
_INCOMPLETE / _KEY_CHANGED / _INVALID), persist a refusal on the plugin row so the
admin list keeps showing it, and badge Signed/Unsigned in the list and in Discover.
Only SIGNATURE_KEY_CHANGED is overridable — an author can legitimately rotate a key;
a signature that does not verify means the bytes are not what the author signed, and
there is no story where waving that through is right. The override re-pins and
updates in ONE call (POST :id/retrust): a re-pin that waited for a follow-up /update
would leave the plugin pinned to a key no install had ever been verified against if
that second call never came. The artifact must still verify under the new key, so a
re-trust moves the pin from one verified key to another.
assertRetrustable re-derives the condition server-side, so the UI hiding the button
is a convenience, not the control, and it echoes back the full key the admin was
shown so a re-key since the dialog rendered is refused. The rotation is written to
the admin audit log with both fingerprints — after an incident, "which key did we
move from, and to what?" is the question a single key cannot answer.
* fix(plugins): let a plugin's frame reach the hosts an admin added for it
The frame's connect-src was built from the manifest's http:outbound grants alone, but
the child's egress guard is the UNION of those and the hosts an admin added after
install for an operatorEgress plugin (a self-hosted Gotify, an ntfy — hosts the author
cannot know in advance). So such a plugin WITH A UI could call the operator's host from
its server and was CSP-blocked in its own iframe.
Match the frame to the child. The admin consented to these hosts at install and the
child already reaches them, so this widens no trust boundary that isn't already crossed.
Both sources stay validated on the way in, and the interpolation filter is unchanged.
* fix(plugins): report a rejected events.emit on the plugin's own log stream
The host can reject an emit (an undeclared event name, a rate limit). The rejection must
not escape — a detached rejection crashes the child and terminally disables the plugin
over one bad emit — but swallowing it silently left an author with no way to discover
that `emits` was missing from their manifest. Surface it as a warning instead.
* feat(plugin-sdk): expose the raw request body for webhook signature checks
A webhook author must run their HMAC over the exact bytes the sender signed. `body` is
the PARSED value, and re-serializing it will not reproduce those bytes — key order,
whitespace and unicode escaping all differ — so the signature never matches. Document
`rawBodyBase64`, which the host already sets on auth:false routes.
* fix(plugin-sdk): refuse to overwrite a released artifact, and keep the packed zip
A released artifact is IMMUTABLE: the registry pins its sha256, so overwriting the bytes
of a release already in the registry breaks the checksum for everyone who installed that
version — they can no longer install or update it. The old code blanket-caught every
`gh release create` failure (auth, network, a bad repo) and turned it into a --clobber
upload. Probe for the release explicitly and refuse unless --force.
Also stop deleting plugin.zip on the way out. It is the exact bytes the release and the
entry's sha256 pin were computed from; a re-pack on another machine or SDK version can
differ (CRLF, walk order), so anyone re-running `entry`/`sign` afterwards must hash THAT
file, not a rebuild.
* fix(plugin-sdk): don't fail submit when the fork already has an upstream remote
`gh repo clone` of a fork may already have wired `upstream`, in which case a bare
`remote add` exits non-zero and took the whole submit down. Set it either way.
* fix(plugin-sdk): make the dev server behave like the host
Three ways `dev` lied to an author about how their plugin would run in production:
- ctx.db: one try/catch wrapped both the node:sqlite probe AND opening the database, so
an mkdir/permission failure silently degraded to the in-memory stub — which swallows
every write while reporting success. A db:own plugin "worked" in dev and persisted
nothing. Probe separately: fail loudly on a real error, degrade only on old Node, and
say plainly that the stub discards writes.
- notificationChannel: the host fires it with no acting user and hands the recipient's
decrypted settings in as a separate `config` argument — send(msg, config, ctx) /
test(config, ctx). Firing it like an ordinary hook passed `ctx` where `config` belongs,
so a channel plugin read its settings off ctx and was broken in production.
- /preview pinned tripId 42 while the scaffold seeds trip 1, so the widget's first
trek:invoke hit assertMember(42) and 500'd. Preview against a trip that exists.
* fix(plugin-sdk): reject unknown flags, wire up create's new flags, and add --help
`parse()` accepts any --x, so a flag a command does not read was silently dropped:
`create --template notification-channel` cheerfully scaffolded a blank plugin. Silently
ignoring an author's explicit instruction is worse than refusing it, so unknown flags are
now an error, and create actually forwards --template/--egress/--required-addons.
A bare --permissions used to split the string "true" into a permission literally named
`true`; listFlag() now treats a valueless flag as absent.
Since an unknown flag is now fatal, `--help` has to exist: it is intercepted before the
flag check and prints usage on stdout with exit 0.
* chore(plugins): drop three unused eslint-disable directives
The no-console rule is not enabled for these files, so the directives were dead and
eslint reports them as unused.
* fix(plugin-sdk): bring preflight back in step with the registry's gates
preflight exists to tell an author what TREK-Plugins' CI will say before they open the
PR. The registry now verifies author signatures, and preflight didn't — so it drifted
into the one failure mode it must never have: a false green. An author trusts a green.
- Verify the signature against the artifact bytes. preflight already downloads them for
the sha256 check, so this costs one call. Without it, signing with the wrong key (or
re-packing after signing) sails through and is caught at review.
- Check the signature SHAPE (checkSignatureShape): a key with no signed version, a
signature with no key, a malformed key or signature. TREK refuses to install a
half-signed entry, so such an entry is dead on arrival.
- Default apiVersion to 1 before comparing. It is OPTIONAL in the manifest — install/
manifest.ts and `entry` both default it — so a manifest that legally omits it was
failing preflight with "manifest apiVersion undefined != entry 1" while the registry
passed it. A false RED, which teaches authors to ignore the tool.
- Check requiredAddons/pluginDependencies parity, and operatorEgress without an
http:outbound grant.
The verifier is a port of the host's install/verify-signature.ts (the registry has its
own port); sign.ts's verifyArtifact only understands the bare key/signature pair the SDK
itself emits and cannot judge a minisign key. A test pins all three to the same verdicts.
* feat(plugins): enforce compatibility range
* chore: bump sdk version
* test(e2e): repair the trip-creation specs
create-trip and trip-planner have been failing for a while — long enough for
three separate UI changes to drift past them, which nothing caught because CI
runs vitest only and never invokes Playwright.
Each failure was masking the next:
- The release-notice modal greets a freshly seeded user and its backdrop
swallowed the click on .add-trip-card. Added a shared dismissSystemNotices()
helper (the X only shows on the notice's last page, so it has to page through
first).
- .modal-backdrop no longer exists — the class was namespaced to
.trek-modal-backdrop so content blockers stop hiding it.
- input[type=text].first() is no longer the Title field: the cover-image search
inputs now sit above it, so the specs were typing the trip name into the photo
search box and creating an untitled trip that never matched getByText(title).
* fix(planner): gate drag & drop on pointer type, not viewport width (#1432)
The 3.2.1 fix disabled drag on "mobile", but nothing in the client has ever
detected touch — "mobile" was inferred from viewport width, at four independent
breakpoints. A tablet is a coarse-pointer device at a *desktop* width, so an
iPad (820-1366px) fell on the wrong side of all four, which is why iPhone was
fixed and iPad was not:
- useTripPlanner's isMobile (<768px) is what disarmed `draggable`, so on iPad
rows stayed draggable and a scroll swipe became an HTML5 drag.
- TripPlannerPage hardcoded isMobile={false} on the desktop PlacesSidebar, so
its drop handlers and the drop-to-import overlay could never be disabled —
that overlay is the reported symptom.
- The arrow-button reorder fallback was revealed only below 767px, leaving iPad
with no drag *and* no fallback.
- touchDragPolyfill loaded drag-drop-touch at >=1024px, synthesising drags from
touchmove — on a landscape iPad that re-armed the very gesture hijack 3.2.1
removed.
Adds useIsTouch() ((pointer: coarse)) as a signal separate from isMobile: layout
stays width-driven, so the iPad keeps the desktop two-pane planner, while every
drag affordance is gated on isMobile || isTouch. The reorder arrows now show on
coarse pointers, and the polyfill loads only on hybrid laptops
((pointer: fine) and (any-pointer: coarse)), which also removes the #1440
phantom-dblclick map zoom on tablets.
Guarded by unit tests plus an e2e spec on real WebKit in an iPad Pro 11 context
— the engine matters, since every browser on iPadOS is WebKit underneath, which
is why the reporter hit this in all three they tried.
* test(planner): guard the day-plan reorder arrows against phantom clicks
The hover rule that reveals the reorder arrows had been dead since the
TypeScript migration: it targeted `.place-row:hover .reorder-btns`, and neither
class exists — the component renders `.reorder-buttons` inside an unclassed row.
The buttons sat at opacity:0 on desktop with nothing to reveal them.
That was not merely invisible. opacity:0 still hit-tests, so every itinerary row
and day note carried an invisible, fully clickable target that silently
reordered the trip:
opacity: "0", visibility: "visible", pointerEvents: "auto"
elementFromPoint(centre) -> BUTTON, inside .reorder-buttons
The repair — pointer-events: none while hidden, plus a working
.dp-row:hover/:focus-within reveal (focus-within so the buttons are also
reachable by keyboard, which the file's JS-hover pattern cannot do) — lives in
index.css and DayPlanSidebar.tsx. Both files also carry the #1432 drag gating,
so they went with that commit rather than being split mid-file; this commit is
the regression guard for them.
E2E rather than unit, because jsdom does not evaluate :hover. Against the old
CSS it fails on exactly the right assertion: "hidden arrows must not swallow
clicks" — expected false, received true.
* fix(pdf): keep the header gap on day-header overflow pages (#1531)
The repeated <thead> day header (#1471) carried no gap before the first
card on overflow pages: the 12px sat in .day-body's block-start padding,
which a fragmented box only paints on its first fragment. Move it to the
thead cell so it repeats with the header; the day's first page renders
pixel-identically.
* fix(budget): keep "no one paid yet" when editing an expense (#1533)
The ExpenseModal payer initializer fell back to the current user whenever the
edited item had no payer, so reopening an expense saved with "no one paid yet"
silently reselected "You" — and re-saving then recorded the current user as the
payer, corrupting the balances. An absent payer on an existing expense is a
deliberate value, so only a brand-new expense defaults to me.
* feat(sdk): support for plugin icons
* fix(planner): keep the places filter applied and visible across tab switches
The category and all/unplanned/tracks filters lived twice: a local copy in
the sidebar driving the checkboxes and the list, and a page-level copy
driving the map markers, synced only when a checkbox was clicked. Switching
planner tabs unmounts the sidebar, so the local copy reset to "All
Categories" while the markers stayed filtered — and the only way out was
toggling any category on and off again (#1541).
The filter now lives once in the trip store, next to selectedDayId: it
survives the Plan tab unmounting (and the mobile places sheet closing),
keeps both sidebar instances in agreement, and resets when another trip
loads.
* feat(airtrail): import connecting flights as one multi-leg booking
AirTrail flights always imported as separate single-leg reservations, so a
layover could not be expressed and the connection country ended up counted
as visited in Atlas (#1535).
The import picker now detects connection chains among the listed flights —
each leg departing from the airport the previous one landed at, onward
within 24 hours, and never back to the origin (an out-and-back is a return,
not a connection) — and offers to import each chain as one flight with
layover stops, on by default. The joined booking keeps per-leg airline,
flight number, times and seat in metadata.legs, files every leg on its own
trip day, and mirrors the first/last leg flat, exactly like the manual
multi-leg form. With the connection stored as a stop endpoint, the existing
Atlas role filter excludes the layover country on its own.
AirTrail has no multi-leg flight a joined booking could round-trip to, so
it imports detached from live sync, with every source flight id recorded in
metadata.airtrail_ids — the picker and the server-side dedupe both treat
those legs as imported, per leg, even across trip members. The server
re-validates each requested chain and falls back to individual imports when
it does not actually connect.
* fix(airtrail): stop syncing a booking once it grows extra stops
A linked flight that becomes multi-leg locally no longer matches the single
AirTrail flight it was imported from: pushing would rewrite that flight to
span the whole route, and the next pull would flatten the layover chain
back to a plain from/to. Both sync directions now detach the link instead —
the same state a joined import starts in, surfaced by the existing "Not
synced" badge.
TransportModal also carries metadata.airtrail_ids through re-saves (like it
already does for transit itineraries and day positions), so editing a
joined booking cannot cost it its import dedupe and get its legs re-offered
in the picker.
* fix(planner): default a new accommodation to checking out the next day
The hotel picker pre-filled "Apply to days" with the same day for check-in
and check-out — a stay that ends the day it begins. New accommodations now
default to the following day for check-out; the last trip day keeps the
same-day range, and editing still seeds from the stored range.
* docs(wiki): document the AirTrail import and connection joining
* fix(plugins): fold resolvePluginIcon into PluginIcon
pluginIcon.ts and PluginIcon.tsx resolve to the same file on the
case-insensitive filesystems dev checkouts commonly sit on (Windows,
macOS) — tsc sees both casings of one module and fails with TS1149 in
every importer. Keep the resolver and the component in one module.
* feat(sdk): add update verification
* chore: remove test files
* fix(sdk) rework the sdk helpers and DX/UX
* fix(sdk) harden dev environment
* fix(budget): add back the multi payer selection
* fix(map): stop MapLibre mouse rotation from reversing near mid-screen
Since MapLibre 5's camera rewrite, the right-button rotate handler flips
its sign whenever the cursor sits above a mid-screen line it derives by
re-projecting the map center. That line drifts with the bearing by a
fraction of a pixel, so inside the 100px band around the screen center a
steady horizontal drag lands alternately above and below it — every
processed movement reverses the previous one and the camera ping-pongs in
place instead of rotating (#1545). A real hand crossing the line mid-drag
flips the rotation direction outright. maplibre-gl 4.x rotated from plain
horizontal movement and had none of this.
Passing aroundCenter: false opts the handler out of the around-center
mode and restores the 4.x/mapbox-gl behaviour: horizontal drag rotates,
vertical drag pitches, in one continuous motion. Applied to all three GL
map builds (planner, journey, settings preview); mapbox-gl keeps its
options untouched.
* fix(budget): settle in the trip's real currency, not always EUR
The settlement route read `currency` off the row returned by canAccessTrip,
whose SELECT never included the column. A cast hid the mistake, so trip.currency
was always undefined and the settlement was told every trip is in EUR.
Balances are netted in the trip currency and converted to the display currency
once. With the trip mislabelled EUR, expenses in the trip's own currency still
cancelled out, but an expense booked in a foreign currency was divided by its
frozen rate into trip-currency units, then converted again as if those were
euros — inflating balances by the EUR/trip rate (~27x for a RUB trip with a USD
expense, #1543). MCP was unaffected: it reads the currency with its own SELECT.
Select the currency in canAccessTrip so the read is real. The settlement maths
was correct all along; it was simply being lied to about the base currency.
Fixes#1543
* feat(trips): let users set the trip currency, and rebase the budget when it changes
The trip currency is the base every expense and settle-up is netted against, but
the only picker for it lived in the legacy Budget addon panel — so on the Costs
panel a trip was stuck with whatever it was created as. Add the field to the trip
form, on create and on edit, gated on trip_edit. The REST and MCP write paths
already accepted `currency`; only the form was missing.
Changing it is not a rename, though: an expense's frozen `exchange_rate` is
"units of its currency per 1 trip currency", and `currency = NULL` means "the
trip's own", so both are relative to the outgoing base. Swapping it out from
under them redenominates the implicit rows (9 000 RUB becoming 9 000 EUR) and
leaves the frozen rates pointing at a currency the trip no longer uses — the same
mismatch that inflated #1543.
So rebaseTripCurrency() runs first, while the old currency is still on the row:
it pins the implicit rows to the outgoing currency and re-freezes every rate
against the incoming one, for expenses and settle-up transfers alike. No stored
amount is rewritten — each keeps the figure the user typed, in the currency they
typed it in, and its real-world value survives the switch.
Also covers the #1543 data as a settlement regression test.
* feat(budget): give settle-up payments their own currency
A transfer settling a shared bill can be made in any currency — paying a rouble
debt in euros is normal — and the server has stored `currency` + a frozen
`exchange_rate` on every transfer since #1445, re-freezing it on edit. The UI
just never let anyone choose one, so a payment silently inherited whatever the
viewer's display currency happened to be.
Add the picker to the payment modal, mirroring the expense modal, and reopen an
existing payment in the currency it was actually recorded in. The ledger row now
shows a foreign payment as `$30.00 -> 27,00 EUR` like a foreign expense does,
instead of stamping the display currency's symbol onto the raw stored number.
The Settle buttons on the suggested flows keep sending the display currency:
those amounts are computed in it.
* feat(settings): make the display currency optional, falling back to the trip's
Costs already resolved `default_currency || trip.currency || 'EUR'`, but the
setting could never actually be empty: the store seeded 'USD', so a user who had
never touched it silently had every trip converted into dollars, and the picker
offered no way to unset it.
Seed it empty and lead the picker with a "Trip currency" option, so an unset
preference means "show each trip in its own currency" instead of forcing them all
through one code. An explicit empty value persists and beats the admin-set
instance default — it is a deliberate choice, not an absence.
This also brings the public share's fallback to life: the share payload has
resolved sharer's currency -> trip currency since #1361, but the trip-currency
branch was unreachable while every owner had a currency forced on them.
Plugins are handed `formats.currency` as a concrete code, so PluginFrame now
resolves the same chain rather than passing an empty string through the bridge.
* docs(wiki): explain the three currencies and how they relate
Trip currency, expense currency and display currency answer three different
questions, and nothing said so: the trip currency wasn't documented at all (it
had no picker until now), and Budget-Tracking conflated the other two while
still claiming 47 currencies and a display currency that always came from
Settings.
Add a Currencies page as the one place they're defined together — the trip
currency as the accounting base, the expense currency as the receipt with its
rate frozen at entry, the display currency as presentation-only — plus what
happens when a trip's currency changes, which currency a public share renders
in, and what belongs to the Costs addon versus the trip itself.
Rewrite Budget-Tracking's currency section against it, document the currency
field in Creating-a-Trip and the display currency in Display-Settings (which
never mentioned it), and note the sharer-or-trip fallback in Public-Share-Links.
* feat(help): serve the in-app wiki from disk instead of fetching GitHub
The /help pages fetched their markdown from raw.githubusercontent.com at
runtime, so a self-hosted install was served docs from main rather than the
version it was actually running, and help was unusable without network access.
The wiki/ directory was in the repo the whole time; the bundled-snapshot
fallback the code reached for was gitignored and never populated by any build
step, so it was dead code.
Read wiki/ straight from disk instead. server/{src,dist}/services both sit
three levels under the repo root, so a single __dirname anchor resolves in dev,
a built source install, vitest and Docker with no copy or build step. GitHub is
kept strictly as a fallback for when the directory cannot be resolved, decided
once at load by probing for _Sidebar.md: a page missing from a present wiki is
a genuine 404, since falling back per-file would reintroduce the version skew
this removes.
Ship wiki/ in the image: .dockerignore excluded it outright, so the COPY alone
would have produced an image with no wiki, and the GitHub fallback would have
masked that at runtime. Add a real path-containment check on assets now that
the path becomes a filesystem read rather than a URL, and document
TREK_WIKI_DIR as an off-by-default escape hatch across the deployment configs.
* fix(plugins): serve frame assets root-relative and cross-origin loadable
res.sendFile(absolutePath) resolves against the rewritten req.url under
the Nest ExpressAdapter and 404s spuriously (files-download already
works around the same trap), which broke every plugin frame document.
And helmet's CORP: same-origin made the browser drop the opaque-origin
frame's own script/style subresources, so a multi-file plugin client
could never boot. Serve root-relative and mark frame responses
cross-origin — sandbox + per-plugin CSP stay the isolation boundary.
* feat(plugins): let a plugin ship its own settings page
A plugin that declares capabilities.settingsUi: true gets its
client/settings.html framed as a card under Settings -> Plugins — same
opaque-origin sandbox and postMessage bridge as its widget, sized via
trek:resize. Hosts that predate the flag strip it at install, so old
instances keep working and simply don't show the card.
* feat(map): open maps framed on their places (builds on #1393) (#1556)
* fix(map): fit MapLibre routes reliably
* fix(map): default planner map to world view
* fix(map): only await route geometry when a route is actually pending
The fit armed a pending route-refit slot on every fitKey change, even with no
route drawn, and only ever cleared it when a route arrived. So a route toggled
on much later — after the user had panned elsewhere — was mistaken for the
awaited geometry and yanked the camera back, and only on the first toggle.
Arm the slot only when a route is already on screen: updateRouteForDay lays down
straight lines in the same batch as the fit and upgrades them to real geometry a
moment later, so an empty route at fit time means none is coming.
* fix(collections): open the empty collection map on the world view
A collection with no mappable places centred on Paris, the same hardcoded
default this branch removes everywhere else.
* feat(map): open the map framed on its places
A trip in Japan opened on the world view at 0,0 and only then animated a fitBounds
flight across the planet — the hardcoded default was the map's answer to a question
its own places already answer.
Each renderer now derives its opening camera from the places it receives, at
construction: MapView for Leaflet, MapViewGL for MapLibre and Mapbox (whose zoom
runs one level below Leaflet's, measuring against a 512px world tile rather than
256px). Doing it at construction is what confines it to load — the map is built
once, and by then the trip's places are in hand. Nothing recomputes it afterwards,
so the camera stays where the user leaves it, and the opening fit stands down
rather than overruling the gentler zoom a lone place opens at.
A trip with no coordinates still falls back to the world view. Collections and the
public shared-trip page frame themselves the same way.
* refactor(settings): drop the default map centre and zoom
Nothing reads them now that every map frames itself on its own places, and a
home-city default was the wrong answer for the next trip on the other side of the
world. The style preview keeps a fixed location of its own: it needs a city to show
label density and 3D buildings, which open ocean cannot.
* fix(map): frame the map the way each renderer can actually draw
Two defects the unit tests missed and running the app exposed.
MapLibre and Mapbox opened on Null Island at zoom 2 regardless of the places: the
effect that mirrors an external centre prop onto the camera also ran on mount, so
it jumped straight to the default nobody passed and threw away the camera the map
had just been built with. It now only responds to actual changes, which is what it
was for. Leaflet was unaffected — its controller already guarded on the centre
changing.
A trip spanning Sydney, Reykjavik and Santiago lost Sydney's marker entirely. The
narrowest arc containing all three crosses the antimeridian, and framing there is
only sound on a renderer that repeats the world: MapLibre and Mapbox draw a marker
on whichever copy is nearest the camera, Leaflet draws one world and puts the
marker at its absolute position — off-screen. Leaflet now spans the long way round,
as L.latLngBounds would. The test that should have caught this wrapped the x-offset
in its own projection helper, quietly assuming behaviour Leaflet does not have; it
now models each renderer's real wrapping.
---------
Co-authored-by: Azalea <noreply@aza.moe>
* feat(mcp): expose public transit planning tools (#1558)
* feat(mcp): add public transit planning tools
* refactor(transit): reuse local time conversion
* fix(mcp): harden transit journey validation
* refactor(transit): centralize itinerary processing
* fix(mcp): return the transit itineraries the provider actually offers
search_transit_routes validated each itinerary leg's mode against
SCHEDULED_TRANSIT_MODES, but that constant is the request-side filter
whitelist — the modes a caller may ask for — not the modes MOTIS can return.
Its default TRANSIT mode expands to TRAM,FERRY,AIRPLANE,BUS,COACH,RAIL,ODM,
RIDE_SHARING,FUNICULAR,AERIAL_LIFT,OTHER, and street legs can be BIKE/CAR/
RENTAL. Any itinerary carrying one of those failed the parse and was dropped
by the flatMap, so the tool reported fewer routes than exist — or none at all.
Against the live provider, Trondheim → Ålesund returns 5 itineraries and 3 of
them contain an AIRPLANE leg, so the tool silently discarded them; the web app
shows all 5, because it treats mode as a free string and renders anything
non-WALK as a transit leg.
Accept any mode token on a leg and keep the existing "at least one non-WALK
leg" rule as the real gate, which restores parity with the web app. Everything
downstream keeps its mode !== 'WALK' semantics, so a journey created over MCP
is identical to one created in the app.
Dropping an itinerary is still possible when provider data fails the remaining
consistency rules, but it is indistinguishable from "no routes exist" — so
search_transit_routes now reports a `dropped` count alongside the results.
---------
Co-authored-by: Uzini <43294422+Uziniii@users.noreply.github.com>
* fix: show map poi search controls on mobile (#1555)
* fix(pdf): use the trip's actual currency instead of hardcoded EUR (#1519)
* fix(pdf): use the trip's actual currency instead of hardcoded EUR
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(i18n): drop hardcoded currency text from pdf.costLabel across locales
Remove redundant EUR/currency references from pdf.costLabel translations
across 21 locales now that the PDF export correctly renders amounts
with their actual trip currency via formatMoney().
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(pdf): drop hardcoded euro-shaped icon from place price chip
svgEuro rendered a fixed € glyph next to the price chip regardless of
the trip's actual currency, undermining the currency fix. Swap for a
currency-neutral coin icon.
---------
Co-authored-by: Nguyen Trong Binh <nguytb15@VN1N07HO1CD1015.local>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
* fix(trips): pin place prices when the trip currency changes
A place price with no currency of its own means "the trip's own currency" —
that is how the PDF export and the place chips read it since #1519. But
rebaseTripCurrency() only pinned budget_items and budget_settlements, so
switching a trip's currency silently redenominated every implicit place
price: a €15 museum on a trip moved to JPY started reading as ¥15. Same
class of mismatch as #1543, on the places surface.
Pin priced, currency-less places to the outgoing currency inside the same
transaction. No amount is rewritten — the figure the user typed keeps its
real-world value, it just stops being ambiguous about which unit it is in.
Bump updated_at as well: it doubles as the optimistic-concurrency token
(#1135), so a client holding the pre-switch row can no longer write the
pin away.
Document place prices in the wiki's currency model, and drop the now-stale
"in EUR" from the PDF export's estimated-cost stat.
* fix(budget): re-split expenses when a member is removed
* fix(plugins): derive plugin req.user.isAdmin from role, not is_admin
* chore(test): make jsdom Storage work under Node's native Web Storage globals
* fix: recompute vacay calendar when week start changes (#1554)
* fix(websocket): handle socket 'error' events to prevent crash on malformed frames (#1584)
* fix(packing): apply templates into the active list tab (#1581)
* Fix packing list readability on mobile
* Localize packing quantity label in overflow menu
* fix(atlas): stream boundary GeoJSON instead of caching parsed bundles (#1585)
* fix(atlas): stream boundary GeoJSON instead of caching parsed bundles
* fix(atlas): scan admin1 features in one pass to avoid O(n^2) build
createFeatureSplitter re-scanned each partial Feature from the start on
every gunzip chunk, so a large feature (Canada ~5.5MB, ~340 chunks) made
the one-time admin1 store build O(n^2). That pushed the streaming build
(~4.2s) past the 15s test timeout under CI coverage/fork contention and
added the same latency to the first live region request.
Carry scan state (position, brace depth, string state) across chunks so
each character is examined once. Output is byte-identical (3228 features,
197 countries) and peak build memory is unchanged (~192MB RSS under a hard
512MB cap); build time drops to ~1.7s.
* fix(i18n): add packing.quantity to all locales
The packing.quantity key existed only in en (Qty) and de (Menge), so
i18n:parity:strict and the client parity test failed for the other 20
locales. Add the key everywhere; en/de are unchanged.
---------
Co-authored-by: jubnl <jgunther021@gmail.com>
* fix(trips): keep accommodations on their dates when the trip range shifts (#1288)
The v3.1.3 fix re-anchored dated bookings after a trip date change but
explicitly excluded hotels: day_accommodations has no absolute date columns,
so stays remained glued to positionally re-dated day rows and shifted with
the range. updateTrip now snapshots day dates before generateDays and a new
resyncAccommodationDays re-anchors each stay (and its linked hotel
reservation, restamping its stale reservation_time) to the days holding its
pre-change dates; out-of-range stays stay glued so whole-trip moves still
shift together. Unlinked dated hotels resync like any booking, and the
date-change block is wrapped in a transaction.
Changing the start date now also asks how plans should follow via a new
date_shift_mode field ('keep_bookings' default / 'shift_all', which reuses
the reorder/insert restamp path to glue everything), exposed in the trip
edit modal (all 22 locales), the shared contract, and the MCP update_trip
tool. Clients no longer show stale state: the initiator reloads
reservations + accommodations after saving, collaborators refetch on a
date-changing trip:updated, and reconnect hydration nudges the planner's
accommodations too.
* fix(extract): retry with json_object and surface AI import failures (#1546)
OpenAI-compatible providers that only support json_object (DeepSeek,
Mistral, some vLLM/llama.cpp) reject the json_schema response_format
with a 400, and the resulting error was swallowed silently: not logged
server-side and never rendered by the background-task widget, leaving
only a generic "no reservations" message.
- Retry the chat/completions request once with response_format
json_object when the json_schema attempt returns 400 (non-NuExtract
only); the system prompt already dictates the output shape
- Log swallowed llm-parse errors with an [llm-parse] tag so failures
show up in server logs
- Render task warnings under the empty-preview note in the background
tasks widget so the actual provider error reaches the user
* fix(costs): keep ticket item amounts visible on narrow screens (#1568)
* chore: update repo url
* chore: update repo url
* chore: update repo url
* chore: update repo url
* fix(security): block IPv6 transition addresses (NAT64/6to4/Teredo) in SSRF guard
An attacker-controlled DNS record pointing at a NAT64 (64:ff9b::/96),
6to4 (2002::/16), or Teredo (2001:0000::/32) address that embeds a
private IPv4 (e.g. 64:ff9b::a9fe:a9fe = 169.254.169.254) bypassed the
SSRF guard: none of the guard functions recognised these ranges, so on a
host that routes the transition prefix the connection reached the
embedded private target (cloud metadata / internal SSRF).
Add a shared embeddedTransitionIpv4() detector and re-apply each guard's
own blocklist to the extracted IPv4 in isAlwaysBlocked, isPrivateNetwork,
isLinkLocal (ssrfGuard.ts) and isBlockedIp (egress-policy.ts). A
transition address to a public IPv4 stays allowed so legitimate
IPv6-only egress is unaffected.
egress-policy.ts keeps the detector inline to preserve its dependency-free
contract for the isolated plugin subprocess.
* chore: Add star history
* Revert "chore: Add star history"
This reverts commit f9d5f75837.
* fix(map): draw transit routes even without other places on the day
The reservation overlays hide any transport whose from/to endpoints
project closer than a per-type pixel threshold (200px for transit) to
declutter tiny no-op straight connectors. A transit journey, though,
draws its real rail/bus alignment rather than a straight endpoint line,
so on a zoomed-out day — one with no other places to tighten the map
onto — its stations fall under the threshold and the whole route
vanishes.
Exempt a transit booking that carries real per-leg geometry from the
proximity gate in both renderers (Leaflet + MapLibre); a geometry-less
transit keeps the straight-arc declutter.
* fix(atlas): resolve region AND country by coordinates against the bundled polygons
Rebased onto dev's streaming atlas index (#1576): region resolution now
resolves a place's lat/lng directly against the same bundled admin1
polygons the client renders — offline, deterministic, and guaranteed to
match a bundle feature — rather than trusting Nominatim's address level,
which can name a subdivision the bundle doesn't carry (Barcelona's ES-B
province vs the bundle's ES-CT autonomous community) and never highlight.
Country resolution moves to the same coordinates-first order, so a place
stored "..., San Francisco, CA" no longer resolves to Canada.
admin1 is held as per-country GeoJSON text (never parsed whole, #1576),
so a country's regions are flattened to the compact Float64Array form and
cached on first use — only visited countries pay the parse. The stale GB
constituent-country rescue is removed, and a one-time migration clears the
re-derivable place_regions cache so every place re-resolves under the new
logic.
Closes#1547
* fix(client): convert mixed-currency day/trip cost totals instead of mislabeling raw sums (#1561)
Day headers, the plan sidebar footer and the PDF day/cover totals summed
raw place prices across currencies and labeled the result with a single
currency — a $2,730.27 hotel on a NOK trip read as "2730 NOK".
Totals now convert every amount into a base currency via the existing
frankfurter rates (sidebar: the user's display currency, falling back to
the trip's; PDF: the trip currency, resolved once before rendering so the
document is consistent), marked with "≈" when a conversion happened. When
a rate is unavailable (offline, blocked egress, unknown code) they fall
back to an honest per-currency breakdown ("2 500 kr + $2,730.27") instead
of folding foreign amounts into a mislabeled number. All-same-currency
trips make no FX request, so offline PDF export keeps working.
Also swaps the place inspector's hardcoded € chip icon for a neutral one
and formats the price via formatMoney in the place's own currency.
Note: day-header totals move from "50 EUR" to Intl formatting ("50 €").
* feat(atlas): let a visited region be hidden, cascading to the country when none remain (server)
Countries already have a hide/tombstone mechanism (hidden_countries, #1490):
a zero-count derived country can be dismissed and stays gone across reloads.
Regions had no equivalent — unmarkRegionVisited only ever deleted a
manually-marked visited_regions row, a no-op for the common case of a region
derived fresh from place_regions on every request, so there was no way to
dismiss one at all (place-derived or otherwise).
Adds the region-level counterpart:
- New hidden_regions table (user_id, region_code, country_code), mirroring
hidden_countries.
- getVisitedRegions() filters its result through it, the same way getStats()
already filters through getHiddenCountries().
- unmarkRegionVisited() now tombstones unconditionally (not just for a
manually-marked region — a region with a real place attached, e.g. one
misassigned by a border-simplification gap, is exactly the case this
exists for) and derives the country code from the region code's
"<country>-<rest>" prefix when there's no visited_regions row to read it
from.
- Cascade: after hiding a region, if the country has no other visible region
left (checked against place_regions + visited_regions, minus hidden_regions),
the country is hidden too via the existing unmarkCountryVisited.
- markRegionVisited() clears both tombstones on re-mark, so a region (and its
cascade-hidden country) can come back.
Note the cascade only has a visible effect on a country with no real place
attached to it — getStats' places-derived country entries are never
suppressed by hidden_countries (#1490's deliberate "reappears with a real
place" rule), so hiding every region of a country that DOES have real places
leaves the country visible, by the same existing design. Test coverage
reflects this.
Server-side only — client wiring (a way to trigger this from the map) is a
separate commit.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(atlas): let a visited region be hidden, cascading to the country when none remain (client)
Client wiring for the server-side hide-region feature (previous commit).
Clicking a visited region on the map used to open the hide/unmark
confirmation only when it was manually marked (visitedRegions[...].manuallyMarked);
a region derived from real place data instead opened the country-detail
view, with no way to dismiss it at all. Now any visited region offers the
same "Remove this region from your visited list?" confirmation regardless of
how it was derived — country details remain reachable via the country
search/sidebar, which was never gated on this in the first place.
The confirm handler's optimistic country-removal check dropped its
`&& r.manuallyMarked` filter on the remaining-regions count, matching the
server's unconditional cascade — but keeps the existing "only when the
country has zero real places/trips" guard, since a country backed by real
data is never actually hidden server-side (#1490) and removing it from the
UI early would just flash and reappear on the next reload.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* fix(settings): make fresh-instance unit defaults internally consistent
A brand-new instance — no admin-set defaults, no saved value — rendered
temperatures in Fahrenheit and times on the 12-hour clock while distances
defaulted to metric, a mix that matches no locale. The store also seeded
'fahrenheit'/'12h' while DisplaySettingsTab's fallback said 'celsius', so the
two paths disagreed about the intended default in the first place.
Default to one system — celsius / metric / 24h — matching the already-metric
distance_unit. The unit defaults now live in a single exported DEFAULT_SETTINGS
that both the store and DisplaySettingsTab's fallback read, so they can't drift
apart again. Admin-set user defaults (getAdminUserDefaults) and any value a user
has already saved still take precedence; only the code-level default changed.
* feat(i18n): complete Catalan (ca) translation
Adds Catalan as a supported language — the full shared/src/i18n/ca locale
(all domain files plus the notification texts), registered in
SUPPORTED_LANGUAGES and the client locale loader. Rebuilt onto current dev so
the locale is at full key parity with en (i18n:parity:strict clean).
* fix(map): stop real road-route fetches from dying under StrictMode
useTransportRoutes cached its AbortController in a ref that was
created once and aborted on unmount. React StrictMode's dev-only
mount->cleanup->remount cycle ran that abort during the *simulated*
cleanup, permanently poisoning the controller before the real mount's
fetch ever started — every road-routed booking (car/bus/taxi/bicycle)
silently fell back to a straight line in local dev, while production
builds (no StrictMode double-invoke) worked fine.
Fix: create a fresh AbortController per effect run instead of a
ref-cached singleton, and synchronously un-mark a job as "attempted"
in that same run's cleanup if it didn't settle before the cleanup
fired — so a StrictMode remount (or any other pre-completion
cancellation) retries instead of being skipped forever. Verified
against the real OSRM endpoint in a running dev server: all four
road-routed legs on a live trip now resolve with real road geometry
instead of straight lines.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): add reservation route-visibility util
Extracts the route-visibility filter that was hand-copied, identically,
in both MapView.tsx and MapViewGL.tsx into one pure, unit-tested
function: a reservation's route shows on the map when it's a transit
booking with the day-route toggle on, or its id is in the caller's
visible-ids set. isRoutableReservation (>= 2 endpoints) is exported
separately since callers besides the map filter need the same check.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): add per-trip connections-visibility resolution util
Pure storage/resolution logic for a trip's booking-route visibility
preference, keyed trek:visible-connections:<tripId>. Two modes:
'only' (nothing shown except the listed ids — today's existing
behavior; a legacy bare-array localStorage value parses as this mode
for backward compatibility) and 'all-except' (everything routable
shown except the listed ids). A trip with no stored preference falls
back to the account-wide default (all or nothing) without writing
anything, so flipping the account setting later never silently
overrides a trip with an explicit per-trip choice already recorded.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(settings): add always-show-booking-routes account setting
New map_always_show_routes account setting, defaulting to off, i18n'd
across all 22 supported locales. Lives in Display > Travel & map,
directly under Booking route labels — its closest sibling — using
that section's immediate-save On/Off pattern rather than a separate
toggle+Save flow, since it's a booking-display preference, not a map
render-config option.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* feat(map): per-trip booking-route visibility with account default and bulk toggle
Lets a user see a booking's route on the map without manually
toggling it per item, two ways:
- The account setting from the previous commit sets the default for
any trip that's never had its routes touched before.
- A new bulk "show all / hide all" button in the day-plan toolbar
flips a trip explicitly between the two connectionsVisibility modes,
independent of the per-item toggle (which still edits whichever
mode's id list is active, in both directions, including while the
account default is on).
useTripPlanner.ts resolves a trip's effective visible-connection ids
from connectionsVisibility.ts + the account setting + the trip's
routable reservations, and exposes it through the same
visibleConnections/toggleConnection contract MapView, MapViewGL and
DayPlanSidebar already had, plus allConnectionsShown/
toggleAllConnections for the new bulk control. MapView/MapViewGL
consume it via the shared reservationRoutes util instead of each
carrying their own copy of the filter.
The bulk toggle's tooltip gets its own map.showAllConnections/
hideAllConnections i18n keys (all 22 locales) distinct from the
per-item toggle's text, and matches the per-item toggle's active
(solid blue) styling rather than the toolbar's generic hover tint.
Manually verified end-to-end in a running dev server: the account
default seeding an untouched trip, the bulk toggle flipping a trip
between all-shown/all-hidden, and a single per-leg override while the
trip is in all-shown mode.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* Update documentation for booking visibility change
* fix(i18n): add Catalan keys for the booking-route toggle + fix the language-count test
#1483 added map.showAllConnections/hideAllConnections and
settings.alwaysShowRoutes/Hint to every locale that existed when it was
written; Catalan (#1418) landed just after and was missing them, which
broke i18n parity. Also bumps the SUPPORTED_LANGUAGES test to 23 entries
and asserts the Catalan one — the ca addition left that assertion stale.
* test(e2e): add a reproducible documentation screenshot run
The wiki screenshots had drifted three months and two releases behind the UI:
39 of 46 assets came from a single commit in April, and pages such as
Budget-Tracking carry a disclaimer that their own images are out of date.
Retaking them by hand is what created that drift in the first place.
This adds a `screenshots` Playwright project that captures them from a dev
build instead:
- seed.ts populates a demo trip over the REST API — deliberately in JPY, so the
3.4.0 currency work (per-trip currency, frozen FX rates, foreign-currency
settle-up) is actually visible rather than hidden behind EUR.
- shot.ts settles the page before capturing (fonts, images, transitions) so
captures don't catch skeleton loaders, and rewrites /auth/app-config to clear
dev_mode — the E2E backend runs NODE_ENV=development, which would otherwise
put a "Dev: Notifications" tab no real deployment has into every admin shot.
- promote.mjs downscales 2880px captures to 1600px on the way into
wiki/assets/, which is 70% smaller for detail the wiki never renders.
Tabs are located by their visible label, so a rename like Budget → Costs fails
the run loudly instead of quietly capturing the wrong panel.
npm run shots && npm run shots:promote
* docs(wiki): retake screenshots against 3.4.0
Regenerated with `npm run shots` from the dev build. Replaces 21 assets and
adds 13 surfaces that had no screenshot at all.
Notable corrections:
- Collections.md referenced assets/Collections.png, which was never committed —
the only broken image in the wiki. It now exists.
- The Budget panel became Costs in 3.3.0 (#1464); every budget image still
showed the old label.
- The trip-create dialog gained a Currency field in 3.4.0, absent from the old
TripCreate.png.
- Settings grew a Plugins tab and the admin sidebar a Plugins entry; neither
appeared in the old sidebar shots.
- Weather now renders in Celsius. A fresh instance defaults to Fahrenheit while
distance defaults to metric, so the seed pins both units — the mismatched
defaults are a separate bug, not fixed here.
Koffi is installed from the community registry rather than dev-linked, because
dev-link and sideload both stamp a badge on the plugin card that an ordinary
install never shows.
Total 3.6 MB for 34 images, against 26 MB for the 46 assets already in the
directory.
No wiki page text is touched here — pages still point at the old filenames
where those were kept, and the newly added images are not referenced yet.
* fix(help): make wiki links and anchors work in the in-app reader
The wiki pages are written in GitHub-wiki style, and 455 of their links across
81 pages use the bare relative form — `[Currencies](Currencies)` — against only
114 `[[..]]` links. processMarkdown only rewrote the latter, so in-app every one
of those 455 fell through to HelpPage's external-link branch, opened a new tab
and 404'd. Since 6c87bf2f serves the wiki from disk, that is the primary way
users read these docs.
Rewriting them in the renderer fixes every page at once and keeps the sources
GitHub-compatible, so contributors can keep writing either form and neither
target breaks. Rewriting the 81 files instead would have fixed today's pages and
left the trap open for tomorrow's.
Also:
- Code is protected from the rewrite. Plugin-Development.md documents
`actions[key](ctx)`, which reads as a markdown link and would otherwise be
corrupted into `actions[key](/help/ctx)` inside a verbatim snippet.
- `[[Page#anchor|Slug]]` no longer renders its anchor as visible link text.
- Headings carry GitHub-compatible ids, so the 22 in-page `](#anchor)` links
scroll instead of doing nothing.
* docs(wiki): correct settings, map and addon docs against 3.4.0
The wiki described settings that no longer exist and pointed at UI labels that
had been renamed. Each correction below was checked against the code.
- The Settings tab is labelled **General**, not "Display"
(shared/src/i18n/en/settings.ts:6). Every "Settings → Display" path was wrong.
The *display currency* setting keeps its name — only the tab was renamed.
- Colour mode is on the **Appearance** tab, not Display.
- The "Route calculation" setting does not exist: no hits for route_calc /
routeCalc / auto_route / calculateRoutes anywhere in client or server.
- Default map centre and zoom were removed in 3.4.0 (0f4766e1). Replaced with
what actually happens now, written from client/src/utils/mapViewport.ts:
every map frames itself on its own places, world view when a trip has no
coordinates.
- A third map provider, MapLibre GL / OpenFreeMap, was undocumented. It needs no
access token, which is the reason a reader would choose it over Mapbox.
- Budget-Tracking.md said the feature is called Costs everywhere and then told
the reader to open the "Budget" tab. The screenshot disclaimer is gone too —
the images now show Costs.
- The admin tab table was missing Plugins.
- The `packing` addon is seeded as "Lists", not "Packing list management".
- Install docs pinned `mauriceboe/TREK:3.0.15` as the exact-release example,
four minor versions stale.
* docs(wiki): document passkeys, calendar feeds, appearance, plugins and help
Five shipped features had no user documentation at all:
- **Passkeys** — WebAuthn enrolment and sign-in, admin policy, RP ID/origins.
Previously mentioned only in passing in Environment-Variables.
- **Calendar Feeds** — the subscribable per-trip and per-user ICS feeds. Note
Day-Plans-and-Notes documents only the one-off .ics *export*; the two are
cross-referenced so it is clear which is which.
- **Appearance Settings** — the whole tab, including the custom accent colour
and its contrast check.
- **Admin: Plugins** — installing from the registry, the pre-install permission
review, egress hosts, and what Reviewed/Signed/Unsigned actually guarantee.
- **In-App Help** — that the wiki ships in the image and is served from disk
since 3.4.0, with the GitHub fallback and TREK_WIKI_DIR.
One deliberate deviation from the brief: the Appearance settings are documented
as account-level, not per-device. They persist to /api/settings on the user
account with nothing in localStorage, so the dashboard widget picker's
desktop/mobile split still changes both from either device.
All five are listed in _Sidebar.md.
* docs(wiki): add screenshots for the remaining surfaces
Second capture pass, bringing the run to 42 screenshots. Adds the surfaces that
need more than a navigation to reach: collection and journey detail, MCP access,
two-factor setup, the settle-up payment dialog, and the trip file manager.
Two fixes to the harness itself, both of which had produced misleading images:
- The admin captures showed a "Dev: Notifications" tab that only exists when the
server runs NODE_ENV=development. The run now clears dev_mode in the
/auth/app-config response; switching the server to production instead would
have enabled HSTS and broken the run over http://localhost.
- The settle-up capture clicked "Settle up", which does not open a view — it
records the transfers. It zeroed every balance and photographed "Everyone's
square", and because the specs share one database it poisoned Costs.png in the
same run. Screenshot specs must not mutate state; it now captures the
"Add payment" dialog instead.
Koffi is installed from the community registry rather than dev-linked, since
dev-link and sideload both badge the plugin card in a way no ordinary install
does.
* test(e2e): capture the detail pages and dialogs
Adds the second wave of screenshot specs (collection/journey detail, MCP access,
2FA, settle-up dialog, files) and enables the mcp, documents and collab addons
in the seed so their surfaces render instead of 404ing.
* docs(wiki): point plugin authors at the agent skill and the registry
Plugin-Development jumped straight into scaffolding without saying that two
supporting resources exist. TREK-Plugins was referenced only in passing far down
the page, and Plugin-Skill — an agent skill that teaches Claude Code and other
SKILL.md-compatible agents to build and publish a plugin — was not mentioned
anywhere in the wiki.
Both are called out up front, with a note that neither is required: the registry
only matters once you want other instances to find your plugin.
* test(e2e): capture the four collab surfaces separately
One Collab.png illustrated chat, notes, polls and the What's Next widget, so at
most one of those four wiki pages showed the feature it described. Each now has
its own capture.
Two things the collab seed needed:
- The conversation is posted by three different people. Every collab write is
attributed to the acting user, and a single-voice chat log would misrepresent
the feature outright.
- Each member therefore gets its OWN request context, created with an explicit
`storageState: undefined`. Without that, newContext inherits the project's
storageState — the admin's trek_session cookie — and extractToken reads the
cookie BEFORE the Authorization header (server/src/middleware/auth.ts:9). The
posts still return 200; they are just all recorded as the admin. That is
exactly what happened on the first attempt, and the DB was the only place it
showed.
The collab view is not tabbed — CollabPanel renders all panels at once — so the
captures target cards by seeded content rather than clicking tabs or matching
headings, whose DOM text is 'Notes'/'Polls' while CSS renders them uppercase.
* docs(wiki): show every screenshot on the page it belongs to
Finishes the wiring the screenshot commits deliberately left out.
- Embeds the 14 images that were committed but displayed nowhere: the Costs
panel and settle-up dialog, the trip planner, transports, documents,
collection and journey detail, the notifications inbox, the Offline and
Account settings tabs, appearance, admin user defaults, registration and
password reset.
- Splits the four collab pages onto their own images. Chat, Notes, Polls and
What's Next each showed the same Collab.png until now; the overview shot moves
to Real-Time-Collaboration, which had no image at all.
- Day-Plans-and-Notes pointed at TripPlaner.png — one 'n'. It now uses the
correctly spelled file, and the misspelled one is deleted since nothing else
referenced it.
- Removes 45 dead '<!-- TODO: screenshot -->' markers whose screenshot had long
since been added. 9 remain, each on a page that genuinely still lacks the
image it asks for — so the marker means something again and the gap is
greppable, which is how this drifted unnoticed for three months.
Every asset in wiki/assets/ is now referenced by a page, and every image
reference resolves to a file.
* docs(wiki): add the four collab screenshots
Chat now shows a real three-person conversation rather than one voice talking to
itself, and the poll shows three separate votes across two options.
44 images, 4.6 MB total.
* test(help): point the asset test at the correctly spelled screenshot
The integration test hard-coded assets/TripPlaner.png — one 'n' — so deleting
the misspelled file broke it. It was the only thing keeping that filename
alive.
* docs(wiki): regenerate the screenshots on top of the rebased dev
ba3733da changed the fresh-instance defaults to celsius/metric/24h. Every
capture showing a clock — chat timestamps, bookings, day plans — and the General
settings tab itself were still on the 12-hour clock, so 39 of 44 images needed a
new run. Regenerating them is one command, which is the point.
The seed keeps pinning the units explicitly: it now matches the new defaults, but
stating them keeps the captures reproducible if a default moves again.
* docs(wiki): regenerate screenshots after rebasing onto dev
dev added a Catalan translation, an always-show-booking-routes account setting
and a bulk route toggle in the day-plan toolbar since the last run — all visible
on captures we ship. 15 of 44 images changed.
Also resolves the Map-Features conflict from 41d12e89: upstream's new bulk-options
section is kept, with the two 'Settings → Display' paths corrected to 'General'.
The tab is labelled General (shared/src/i18n/en/settings.ts:6), and upstream's own
i18n key for that section is settings.general.travelMap.
* chore: only allow manual trigger for the build&push
---------
Co-authored-by: Maurice <61554723+mauriceboe@users.noreply.github.com>
Co-authored-by: Dieter Blomme <dieterblomme@gmail.com>
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
Co-authored-by: sld272 <zjrdmczh@outlook.com>
Co-authored-by: Nguyen Trong Binh <nguytb15@VN1N07HO1CD1015.local>
Co-authored-by: Maurice <mauriceboe@icloud.com>
Co-authored-by: Pavel Zolotarevskiy <code@fxgn.dev>
Co-authored-by: Azalea <noreply@aza.moe>
Co-authored-by: Uzini <43294422+Uziniii@users.noreply.github.com>
Co-authored-by: Daniel <drmoreno271@gmail.com>
Co-authored-by: trongbinhnguyen <43725147+trongbinh15@users.noreply.github.com>
Co-authored-by: Konstantinos Thermos <info@subdee.org>
Co-authored-by: Konstantinos Thermos <subdee@users.noreply.github.com>
Co-authored-by: Lucas Español <lucas.espanol@tutanota.com>
Co-authored-by: fbnlrz <frlrnzn@gmail.com>
2026-07-18 22:13:52 +02:00
3057 changed files with 334158 additions and 61628 deletions
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
## What TREK is
A self-hosted, real-time collaborative travel planner. npm-workspaces monorepo with four workspaces:
- **`shared/`** (`@trek/shared`) — Zod schemas that are the **single source of truth** for API contracts, consumed by both server and client. Also owns all i18n locale files. Must be built before server/client typecheck or run.
- **`nest-mcp/`** (`@trek/nest-mcp`) — decorator-driven MCP tool/resource/prompt registration for NestJS, consumed by the server. Deliberately **extraction-clean**: no imports from `@trek/server`/`@trek/shared`, no TREK semantics.
- **`server/`** (`@trek/server`) — NestJS API (Express adapter), SQLite via `better-sqlite3`, WebSocket sync, built-in MCP server, sandboxed plugin runtime.
- **`client/`** (`@trek/client`) — React 19 + Vite + Zustand + Tailwind PWA, offline-first via Dexie/IndexedDB.
**`plugin-sdk/`** (`trek-plugin-sdk`) also lives in the repo but is **not a root workspace** — it has its own lockfile, is published to npm independently, and must ship standalone. Run its commands from `plugin-sdk/`, not the root.
Each package has its own `CLAUDE.md` with internals; this file covers the monorepo picture.
## Commands
Run from the repo root unless noted. Per the user's preference, **use `npm run ... --workspace=<pkg>`, not `npm -w <pkg>`**.
```bash
npm run dev # build shared + nest-mcp, then watch shared + server + client concurrently
npm run build # build shared → nest-mcp → server → client (order matters)
npm run lint # lint all four workspaces
npm run format # prettier --write all four
```
Tests (the user runs client tests as `cd client && npm run test`):
```bash
npm run test# all workspaces
npm run test --workspace=server # vitest run (server)
npm run test:unit --workspace=server # tests/unit only
npm run test:integration --workspace=server
npm run test:ws --workspace=server # websocket tests
npm run test:e2e # server e2e (boots Nest against temp SQLite)
npm run test:cov # coverage, server + client
cd client && npm run test# client vitest run
cd client && npm run e2e # Playwright
cd client && npm run lint:pages # enforce the Page pattern (see below)
cd client && npm run theme:lint # flag styling that bypasses appearance tokens (--strict in CI mindset)
npm run typecheck --workspace=server # tsc --noEmit (also client/shared)
```
Run a single test file or test:
```bash
npx vitest run tests/unit/nest/weather.controller.test.ts # (from server/ or client/)
npx vitest run -t "name of the test"# by test-name pattern
```
i18n parity (enforced in CI — every non-`en` locale must have identical file set + top-level keys):
```bash
npm run i18n:parity --workspace=shared # audit (exit 0)
npm run i18n:parity:strict --workspace=shared # CI gate (exit 1 on drift)
```
Dev ports: server listens on **3001**, Vite dev server proxies `/api`, `/ws`, `/uploads`, `/oauth`, etc. to `http://localhost:3001` (see `client/vite.config.js`).
## Server architecture
Nest owns everything: the strangler migration **completed 2026-08-10**, and the last seam — the pre-`app.init()` MCP/OAuth mount — moved behind the container 2026-08-11: **zero `*.bridge.ts` files remain**. Legacy `src/services/` is deleted (ESLint errors on any import of it) and all business logic lives in DI modules under `server/src/nest/<domain>/`. `server/src/nest/README.md` is the module blueprint. Every cron is Nest-owned (`server/src/nest/scheduling/`, `@nestjs/schedule`).
- **`bootstrap.ts` — `buildApp()`** is the single shared builder used by both production (`index.ts`) and the integration/e2e test harness. **Composition order is load-bearing**: global middleware → platform uploads → the discovery metadata middleware → static assets → the named body-parser wrappers are all registered on the raw Express instance **before `app.init()`**, because Nest's router throws `NotFoundException` on unmatched routes and won't fall through to anything registered after init. The transport surface (`/api/health`, `/mcp`, `/.well-known/*` documents, `/oauth/*`) is Nest controllers/middleware. The SPA `index.html` fallback and the global error envelope are `APP_FILTER`s (`SpaFallbackFilter`, `TrekExceptionFilter`).
- **`/mcp` bodies are raw by design** — bootstrap's parser wrappers exempt `/mcp` so the MCP SDK reads the untouched stream; never `@Body()` a `/mcp` route.
- **Per-domain module shape** (one folder under `server/src/nest/<domain>/`): `<domain>.controller.ts` + `<domain>.service.ts` + `<domain>.module.ts`, registered in `app.module.ts`. **`weather/` is the reference implementation** — copy its shape. The Zod contract for the domain lives in `shared/src/<domain>/<domain>.schema.ts`.
- **Auth is default-deny**: `GlobalAuthGuard` + `MfaPolicyGuard` run as `APP_GUARD`s; public routes opt out with `@Public()`, and a boot-time ratchet (`validate-route-guards.ts`) refuses unknown public routes. New domain logic goes in the module's `.service.ts` — never resurrect a `services/*`-style plain module.
- **Trip-scoped routes** must verify trip access (404 on no access) and the relevant permission (403) — the standard shape is `@UseGuards(JwtAuthGuard, TripAccessGuard)` + `@RequirePermission('<action>')` per handler (a few domains still check inline; see `roadmap.md`) — and forward `X-Socket-Id` to the WebSocket broadcast so the originating client doesn't echo its own change. Never gate a multipart upload with a guard (ECONNRESET instead of 403 — check in the handler).
- **DB** (`server/src/db/`): `better-sqlite3`, WAL mode, FK on. `database.ts` initializes; `schema.ts` creates tables; `migrations.ts` + `seeds.ts` run on boot. In `NODE_ENV=test` each vitest worker gets an isolated `:memory:` DB; the Playwright harness uses `TREK_DB_FILE` for a throwaway file DB.
- **WebSocket** (`server/src/websocket.ts`): JWT-authed, room-per-trip (`tripId → Set<WebSocket>`), heartbeat keep-alive, per-connection rate limiting, served on `/ws`.
- **MCP** (`server/src/mcp/`): OAuth 2.1-authenticated MCP server exposing TREK to AI assistants (150+ tools in `tools/`, resources, scopes, session manager). See `MCP.md`.
- **Addons** are admin-toggleable feature modules keyed by `ADDON_IDS` in `server/src/addons.ts` (mcp, packing, budget, documents, vacay, atlas, collab, journey, airtrail, llm_parsing, collections).
- **Plugins** (`server/src/nest/plugins/`): sandboxed third-party plugin runtime — one forked child process per plugin (`supervisor/`), permission-gated RPC (`host/rpc-host.ts` registers a handler only if the plugin holds the unlocking permission), install-time manifest/signature/egress checks (`install/`, `runtime/egress-policy.ts`). Author tooling lives in `plugin-sdk/`; the `trek-plugin-dev` skill covers plugin authoring.
## Client architecture
Offline-first, with a layered data flow. A Page never owns state directly:
- **Page pattern** (enforced by `lint:pages`, see `client/src/pages/PATTERN.md`): a `*Page.tsx` default export is a **wiring container** — it composes a co-located `use<Page>()` hook into JSX and must **not** call `useState/useReducer/useEffect/useLayoutEffect/useMemo/useCallback/useRef` in its own body. Stateful logic lives in the hook.
- **State**: Zustand stores in `client/src/store/`. The big one is `tripStore.ts`, assembled from **slices** in `store/slices/` (places, packing, todo, budget, reservations, files, …). `slices/remoteEventHandler.ts` applies incoming WebSocket events to local state.
- **Data flow (offline-first)**: `store → repo (client/src/repo/) → api (client/src/api/) | Dexie (client/src/db/offlineDb.ts)`. Repos check `navigator.onLine`: online they hit the REST API and upsert into Dexie; offline they read from Dexie. Writes go through `sync/mutationQueue.ts` — optimistic Dexie write, then on reconnect `flush()` replays REST with an `X-Idempotency-Key` header so retried writes don't double-apply. `sync/tripSyncManager.ts` and `sync/syncTriggers.ts` drive reconnection sync; `sync/tilePrefetcher.ts` caches map tiles.
- **Styling**: use the appearance tokens (`var(--token)` inline or `bg-[var(--...)]` Tailwind classes), never hardcoded color literals or numeric `fontSize` — `theme:lint` flags bypasses so user-chosen scheme/transparency/text-size keep working. Suppress intentional exceptions (map/PDF/brand colors) with a `theme-lint-disable` line comment.
- A route is "done" only once its contract lives in `shared/` and both sides import the inferred types. Edit the Zod schema in `shared/src/<domain>/`, rebuild shared, then both server (validation + DTO types) and client (typed requests) pick it up.
- **i18n locale files live in `shared/src/i18n/<locale>/`**, one file per domain. `en/` is canonical. When you add/change a translation key, **add it to every locale** or `i18n:parity:strict` fails CI. The client resolves keys via `client/src/i18n/TranslationContext.tsx` (`t(key)`), which only reads top-level string keys.
## Project direction (from the 2026 full-repo audit)
A verified audit of the whole repo lives in `audit_report/TREK_audit_reports/` (`00_MASTER_SUMMARY.md` is the entrypoint, `01_STEERING_ROADMAP.md` the plan). Its verdict shapes how all new code should be written:
- **Protect the crown jewels.** The client's offline-first data core (repos + mutation queue + Dexie), the out-of-process plugin sandbox, the Zod contract layer, and the auth primitives are production-grade. Extend them; never route around them or rewrite them in anger.
- **No "modern shell over legacy core."** The debt pattern is a modern frame delegating to the old approach underneath (historically Nest controllers over legacy `services/*` — that migration is done; on the client it's feature components calling the raw API past the offline core, still ~190 files). New code goes all-in on the target architecture; when touching a legacy seam, migrate it rather than adding another wrapper.
- **Single source of truth over manual synchronization.** Never add a hand-mirrored copy of state, schema, or contract (server↔Dexie↔Zustand, schema↔SQL, the plugin contract). Derive from one source, or guard the copy with a parity test that cannot silently skip.
- **DI over global mutable module state.** The `db` proxy, `ws` singleton, and event-sink globals are what force `require()`-to-dodge-cycles and load-bearing boot ordering. New server code injects its dependencies; new client code avoids `window`-global buses and mutable module state.
- **Use the runtime's native idioms.** Prefer React 19 features (`useOptimistic`, Actions, `use()`/Suspense) and Nest subsystems (DI providers, pipes, guards, `@nestjs/schedule`) over hand-rolled equivalents of the same machinery.
- **Fail closed, gates stay on.** Security switches default to safe; misconfiguration must refuse, not silently degrade. Never lower a quality gate (strict TS, lint severity, coverage, parity checks) to land a change — no new `any`, no new `eslint-disable`, no downgrading rules to `warn`.
## Conventions (from CONTRIBUTING.md)
- **Target the `dev` branch** for PRs, not `main` (exception: `wiki/`-only changes).
- **Discuss first**: outside contributions must be pitched in the `#github-pr` Discord channel before any code is written — undiscussed PRs are closed.
- **PRs follow `.github/PULL_REQUEST_TEMPLATE.md`** and require a linked issue (`Closes #N`) for bug fixes or an approved feature-request discussion for features — no issue/discussion, no PR.
- **Conventional commits** (`fix(maps): ...`, `feat(budget): ...`). Per user preference, **do not add Co-Authored-By Claude attribution**.
- One focused change per PR; no breaking changes; no unrelated reformatting/refactors. Tests required — the project holds **80%+ coverage** (the `src/nest/**` vitest coverage gate enforces ≥80%).
- When migrating/adding a route, **parity is law**: same URL, method, query/body, HTTP status, `Set-Cookie`, and JSON body — including bespoke error strings (e.g. reproduce `{ error: 'Admin only' }` exactly rather than relying on a generic guard message). Note Nest defaults POST to 201; add `@HttpCode(200)` where the legacy contract returned 200. Declare static sub-routes (`/reorder`, `/in-app/all`) **before**`:id` param routes.
@@ -10,7 +10,7 @@ Thanks for your interest in contributing! Please read these guidelines before op
4.**Target the `dev` branch** — All PRs must be opened against `dev`, not `main`. Exception: PRs that only modify files under `wiki/` may target any branch
5.**Match the existing style** — No reformatting, no linter config changes, no "while I'm here" cleanups
6.**Tests** — Your changes must include tests. The project maintains 80%+ coverage; PRs that drop it will be closed
7.**Branch up to date** — Your branch must be [up to date with `dev`](https://github.com/mauriceboe/TREK/wiki/Development-environment#3-keep-your-fork-up-to-date) before submitting a PR
7.**Branch up to date** — Your branch must be [up to date with `dev`](https://github.com/liketrek/TREK/wiki/Development-environment#3-keep-your-fork-up-to-date) before submitting a PR
## Pull Requests
@@ -39,8 +39,8 @@ feat(budget): add CSV export for expenses
## Development Environment
See the [Developer Environment page](https://github.com/mauriceboe/TREK/wiki/Development-environment) for more information on setting up your development environment.
See the [Developer Environment page](https://github.com/liketrek/TREK/wiki/Development-environment) for more information on setting up your development environment.
## More Details
See the [Contributing wiki page](https://github.com/mauriceboe/TREK/wiki/Contributing) for the full tech stack, architecture overview, and detailed guidelines.
See the [Contributing wiki page](https://github.com/liketrek/TREK/wiki/Contributing) for the full tech stack, architecture overview, and detailed guidelines.
| **Static API token** | `trek_` | Full access | No expiry | **Deprecated.** Triggers deprecation warnings in AI clients. Will be removed in a future release. |
| **Web session JWT** | — | Full access | Session-based | Used internally by the TREK web UI. Not intended for external clients. |
All methods require the `Authorization: Bearer <token>` header (strict scheme enforcement — `Bearer` required).
---
## OAuth Scopes
When connecting via OAuth 2.1, you grant specific scopes during the consent step. TREK registers only the MCP tools
| **Admin activation required** | The MCP addon must be enabled by an admin before any user can access it. |
| **Per-user scoping** | Each MCP session is scoped to the authenticated user. You can only access trips you own or are a member of. |
| **No image uploads** | Cover images cannot be set through MCP. Use the web UI to upload trip covers. |
| **Reservations are created as pending** | When the AI creates a reservation, it starts with `pending` status. You must confirm it manually or ask the AI to set the status to `confirmed`. |
| **Demo mode restrictions** | If TREK is running in demo mode, all write operations through MCP are blocked. |
| **Rate limiting** | 300 requests per minute per user (configurable via `MCP_RATE_LIMIT`). Exceeding this returns a `429` error. |
| **Per-client rate limiting** | Rate limits are tracked per user-client pair, so each OAuth client has its own independent rate limit window. |
| **Session limits** | Maximum 20 concurrent MCP sessions per user (configurable via `MCP_MAX_SESSION_PER_USER`). Sessions expire after 1 hour of inactivity. |
| **Token limits** | Maximum 10 static API tokens per user. Maximum 10 OAuth clients per user. |
| **Token revocation** | Deleting a static token or revoking an OAuth session immediately terminates all active MCP sessions for that token/client. |
| **OAuth scope enforcement** | Only tools matching your granted OAuth scopes are registered in the session. Calling an out-of-scope tool returns an error. |
| **Addon toggle invalidation** | When an admin enables or disables an addon, all active MCP sessions are invalidated and must be re-established. |
| **Real-time sync** | Changes made through MCP are broadcast to all connected clients in real-time via WebSocket, just like changes made through the web UI. |
| **Addon-gated features** | Some resources and tools are only available when the corresponding addon (Atlas, Collab, Vacay, Journey) is enabled by an admin. |
---
## Resources (read-only)
Resources provide read-only access to your TREK data. MCP clients can read these to understand the current state before
| `get_trip_summary` | Full denormalized snapshot of a trip: metadata, members, days with assignments and notes, accommodations, budget, packing, reservations, collab notes, to-dos, and poll/message counts. Use this as your context loader. |
### Compound Tools
Compound tools collapse common multi-step workflows into a single atomic call. Each one wraps two sequential operations in a database transaction — if the second step fails, the first is rolled back automatically.
> **When to use:** Only use compound tools when the place or item does not yet exist. If it already exists, call the individual tools (`assign_place_to_day`, `create_accommodation`, `set_budget_item_members`) directly.
| Tool | Wraps | Description |
|---|---|---|
| `create_and_assign_place` | `create_place` + `assign_place_to_day` | Create a new place and immediately assign it to a specific day. Accepts all `create_place` fields (`place_notes` instead of `notes`) plus `dayId` and optional `assignment_notes`. Returns `{ place, assignment }`. |
| `create_place_accommodation` | `create_place` + `create_accommodation` | Create a new place and immediately book it as an accommodation for a date range. Accepts all `create_place` fields (`place_notes` instead of `notes`) plus `start_day_id`, `end_day_id`, `check_in`, `check_out`, `confirmation`, and `accommodation_notes`. Also auto-creates a linked hotel reservation. Returns `{ place, accommodation }`. |
| `create_budget_item_with_members` | `create_budget_item` + `set_budget_item_members` | Create a budget item and optionally set which members are splitting it. Accepts all `create_budget_item` fields plus an optional `userIds` array. If `userIds` is omitted or empty, behaves identically to `create_budget_item`. Returns `{ item }` with members populated. |
**Scope requirements** match the underlying tools: `places:write` for `create_and_assign_place`, `trips:write` for `create_place_accommodation`, `budget:write` for `create_budget_item_with_members` (Budget addon required).
| `create_accommodation` | Add an accommodation (hotel, Airbnb, etc.) linked to a place and a check-in/out date range. |
| `update_accommodation` | Update fields on an existing accommodation (dates, times, confirmation, notes). |
| `delete_accommodation` | Delete an accommodation record from a trip. |
### Transport
Transport bookings (flights, trains, cars, cruises) support multi-stop `endpoints[]` — each endpoint has a `role` (`from`/`to`/`stop`), name, optional IATA `code` (for flights), coordinates, timezone, and local time. Use `search_airports` to resolve airport names to IATA codes before creating a flight.
| `create_transport` | Create a transport booking (`flight`, `train`, `car`, `cruise`) with optional endpoints, departure/arrival times, and confirmation details. Created as pending. |
| `update_transport` | Update an existing transport booking. Pass `endpoints[]` to replace the full stop list. Use `status: "confirmed"` to confirm. |
| `delete_transport` | Delete a transport booking from a trip. |
### Reservations
For flights, trains, cars, and cruises, use the **Transport** tools above. Reservations cover all other booking types.
| `create_reservation` | Create a pending reservation. Supports hotels, restaurants, events, tours, activities, and other types. Hotels can be linked to places and check-in/out days. |
| `update_reservation` | Update any field including status (`pending` / `confirmed` / `cancelled`). |
| `delete_reservation` | Delete a reservation and its linked accommodation record if applicable. |
| `reorder_reservations` | Update the display order of reservations (and transports) within a day. |
| `link_hotel_accommodation` | Set or update a hotel reservation's check-in/out day links and associated place. |
### Budget
> To create a budget item and set its members in one call, use [`create_budget_item_with_members`](#compound-tools).
See [`charts/README.md`](https://github.com/mauriceboe/TREK/blob/main/charts/README.md) for values.
See [`charts/README.md`](https://github.com/liketrek/TREK/blob/main/charts/README.md) for values.
<h2 id="install-as-app-pwa">Install as App (PWA)</h2>
@@ -331,6 +343,8 @@ The script creates a timestamped DB backup before making changes and prompts for
For production, put TREK behind a TLS-terminating reverse proxy. TREK uses WebSockets for real-time sync, so the proxy **must** support WebSocket upgrades on `/ws`.
If you use the MCP addon, the proxy must also pass the `Mcp-Session-Id` header through in both directions on `/mcp` — Nginx and Caddy do this by default, but a proxy that strips it makes every tool call open a new session instead of reusing one. See the [Reverse Proxy wiki page](https://github.com/liketrek/TREK/wiki/Reverse-Proxy) for details.
<details>
<summary>Nginx</summary>
@@ -368,6 +382,19 @@ server {
proxy_set_headerHost$host;
proxy_read_timeout86400;
}
# Only needed if you use the MCP addon. Responses are Server-Sent Events,
# so buffering must be off or tool results arrive late.
> with a report listing every offending variable. Boolean switches accept
> `true`/`false`, `1`/`0`, `on`/`off` and `yes`/`no` (any casing).
<details>
<summary><b>Full reference</b></summary>
@@ -403,6 +437,7 @@ Caddy handles TLS and WebSockets automatically.
| `ENCRYPTION_KEY` | At-rest encryption key for stored secrets (API keys, MFA, SMTP, OIDC). Recommended: generate with `openssl rand -hex 32`. If unset, falls back to `data/.jwt_secret` (existing installs) or auto-generates a key (fresh installs). | Auto |
| `TZ` | Timezone for logs, reminders and cron jobs (e.g. `Europe/Berlin`) | `UTC` |
| `TREK_WIKI_DIR` | Where the in-app Help pages (`/help`) read their content from. TREK ships its wiki and serves it from disk, so Help always matches the version you are running — you should not need to set this. Point it at your own directory to serve custom docs. If the path does not exist, Help falls back to fetching the public GitHub wiki (needs outbound network, and tracks the latest release). | bundled `wiki/` |
| `DEFAULT_LANGUAGE` | Default language on the login page for users with no saved preference. Browser/OS language is auto-detected first; this is the fallback. Supported: `de`, `en`, `es`, `fr`, `hu`, `nl`, `br`, `cs`, `pl`, `ru`, `zh`, `zh-TW`, `it`, `ar`, `id`, `tr`, `ja`, `ko`, `uk`, `gr` | `en` |
| `ALLOWED_ORIGINS` | Comma-separated origins for CORS and email links | same-origin |
| `FORCE_HTTPS` | Optional. When `true`: 301-redirects HTTP to HTTPS, sends HSTS, adds CSP `upgrade-insecure-requests`, forces the session cookie `secure` flag. Useful behind a TLS-terminating reverse proxy. Requires `TRUST_PROXY`. | `false` |
@@ -430,7 +465,7 @@ Caddy handles TLS and WebSockets automatically.
| `UNSPLASH_ACCESS_KEY` | Optional Unsplash Access Key for trip-cover and place-image search. Without one, TREK uses Unsplash's unauthenticated endpoint, which some datacenter/VPS IPs are blocked from. Get a free key at [unsplash.com/developers](https://unsplash.com/developers). Overrides any per-admin key set in Admin > Settings (where it can also be configured instead). | — |
| `MCP_RATE_LIMIT` | Max MCP API requests per user per minute | `300` |
| `MCP_MAX_SESSION_PER_USER` | Max concurrent MCP sessions per user | `20` |
| `MCP_MAX_SESSION_PER_USER` | Max concurrent MCP sessions per user. At the cap, the least-recently-active session is closed to make room | `20` |
</details>
@@ -456,4 +491,3 @@ for full third-party attributions.
## License
TREK is [AGPL v3](LICENSE). Self-host freely for personal or internal company use. If you modify and offer TREK as a network service to third parties, your modifications must be open-sourced under the same licence.
<Profile>TREK is a self-hosted, real-time collaborative travel planner. Plan trips together with interactive maps, budgets, bookings, packing lists, day-by-day itineraries and file management — every change syncs instantly across everyone in your group. Includes OIDC/SSO, TOTP MFA, dark mode, PWA support, multi-language UI and a modular addon system (Vacay, Atlas, Collab, Budget, Packing, Journey). Maintained by mauriceboe — support and bug reports via GitHub Issues.</Profile>
> **Note:** `chart.liketrek.com` is a custom domain (CNAME) for the GitHub Pages site at `https://liketrek.github.io/TREK` — both URLs serve the same repository. The github.io URL keeps working (it redirects to `chart.liketrek.com`), but the custom domain is the canonical one to use.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Maurice