- list_budget_items lists a trip's expenses with their ids, payers and
splits, page by page so no page is cut down on its way to a model;
get_budget_per_person answers who paid and owes what. Both in a new
budget-items.mcp.ts, checked like their REST twins.
- get_wind_forecast reads the hourly wind, gusts and direction of a day.
- get_journey_tracks lists a journey's GPX tracks; the line for drawing
rides under a key the provider never receives.
- The Dawarich wishlist scan and the countries the recordings suggest
become tools, with their confirming writes, in DawarichAtlasMcp in the
Atlas group; the shared time window check refuses alike on REST and
MCP, and the model never sees when the user was last seen in a city.
- The journal's map, tracks and recorded days load with the journey
group instead of riding in the core of every step.
- The prompt points at list_budget_items for expense ids (the trip
summary never carried them) and names the shares and the wind tools;
each new read has its own live step.
The hiking costing moves to shared, so the planner and the server ask the
router the same way. The road trip router learns to plan a hike on foot
with that costing (Valhalla only: OSRM knows no trail difficulty), to read
heights along a line in chunks and to read the OSM ways a line follows,
all through its one paced queue. A hike router builds on it: the way, its
heights, distance, climb and DIN walking time, and the SAC grades of its
stretches from the trail grade service, each part best effort.
New MCP tools in HikePlanMcp, in the assistant's tours group:
- propose_hike plans a hike without saving it.
- create_hike saves a confirmed hike through the planner's own creation;
a router that does not answer leaves it pending for the planner, a way
that cannot be planned is refused and nothing is saved.
- update_tour changes a saved hike against its revision, keeping linked
stations and refusing what would silently drop drawn legs or stages.
- import_tour_gpx takes a GPX file's text in as tours, like the import.
Same permission checks as REST, broadcasts from the services.
In the assistant a new hike or city walk is an "Add tour" card with its
route and difficulty, undone through delete_place while the tour is
untouched, and a hike being planned shows as a live step.
- Tests that still knew the ChatGPT sign-in as personal only follow the
admin's shared plan: the service takes the caller with their role, the
admin tiles show ChatGPT, the controller tests cover the admin routes.
- The table refused an instance ChatGPT connection with a CHECK, so
sharing the plan failed. A new migration rebuilds ai_connections
without that check; the shipped migration stays as it was.
- useChatgptConnect grew past the function size limit: reading the
clipboard and noticing the return to the tab are helpers of their own.
- The file link factory hands its overrides over apart from the row,
like every other factory, which ends the deep type instantiation.
- Knip: exports only their own file used are private again, and the
booking cap reads MAX_WRITES_PER_STEP instead of an alias.
- lint:i18n-keys reaches the tools' own action names through a key
builder and the connection hints through literal keys.
- MMehrSheet stays in the assistant hints pair only.
The Anthropic request marked only the instructions and the trip card for
the prompt cache, so every step of a turn paid the whole history again.
It now also marks the end of the history, which the next step starts
from, and the end of what came before the user's question, which every
step of the turn sends unchanged, within the four marks a request takes.
Every provider now reports how much of the input it read from its cache
(Anthropic cache_read_input_tokens, OpenAI cached_tokens), the turn adds
it up, and the usage log line says it, so the saving can be measured.
The connection for everyone can now also be the admin's own ChatGPT sign-in.
Whether a plan may be shared that way is the self-hoster's call under their
plan and OpenAI's terms; the form says what it means before the sign-in.
The sign-in starts on the admin route and finishes on the shared one, which
refuses a sign-in for everyone to anyone who is not an admin. The ChatGPT
return reads the clipboard only once the paste handler exists.
- The sign-in opens in a small window over TREK instead of a new tab.
- Step 2 shows the address the window ends on, so the error page there is
expected rather than alarming.
- Back in TREK the copied address is read off the clipboard where the browser
allows it and the sign-in finishes; else the field has the cursor.
- The token and OAuth client factories hand their columns over apart from
the row, as the others do (TS2589).
OpenAI offers a fixed web redirect for Sign in with ChatGPT only to selected
commercial partners; open-source apps sign in through the 127.0.0.1 loopback
callback. The relay could never be switched on, so it goes again.
Once the TREK project has a client id registered with OpenAI, the sign-in
goes out with the relay page as its redirect address and the install's own
settings in its state; the relay sends the browser back there and the
settings finish the sign-in, nothing to paste. The relay sees the code but
cannot redeem it without the install's PKCE verifier. Without the project's
client id, or for an install without an https address, the loopback page
and the pasted address stay as they are.
relay/chatgpt-callback/index.html is the page to publish.
- A card for a tool without one of its own names what it works on: the place
of an assignment, the day, the note, the booking; its arguments sit under
the labels the other cards use, a new order reads as the names in order.
28 frequent tools carry their own action name in every language.
- The server's own steps without a card still refuse, as before.
- Tests: the rate limit is asked one request after another, the mention list
closes on the caret move a browser reports, the place factory hands its
category over as a plain id, eleven files formatted.
The assistant offered 25 hand-picked tools. It now offers every tool the MCP
registry holds, but a short denylist with a reason per tool (deleting the
trip, share, invite and feed links, membership changes, fetching an address
the model picks, documents to the provider, instance-wide categories and
holidays). The scopes widen to every addon but trips:delete, the share
scopes, file contents and plugins.
- A tool annotated read-only runs at once, every other one waits as a change
card; reads that fetch a planted address wait too.
- A turn starts with the trip's core tools; use_addon_tools loads tours, road
trip, journey, vacay, atlas, collections, trip chat, files, account or
plugin tools for the rest of the turn.
- A writing tool without a card of its own gets one from its arguments:
add, change or remove, the part of TREK it works in, the trip's days,
places and people named.
- The boot check fails on a tool that shares, invites or publishes and is
neither denied nor known to only take back.
The assistant tables gave Users three more relations, and a spread of
overrides checked against RequiredEntityData now ran tsc past its
instantiation depth (TS2589). Every factory passes its overrides as
createRow's own argument, the way rows.ts documents it.
- agent loop and decisions without await in loops: steps as recursion,
calls and applied changes in order through one promise chain, the
independent reads and claims in parallel; part mapping in its own helpers
- place card, header and weather strip without refs or components created
in render; long hooks and components split; state reset during render
- trip notes only in a turn's tool session, never an undo's; map view
reported only while the assistant can run
- place strip scrolls through the strip's own smooth scroll, no physical side
- tests follow the raised limits and the new card fields; toasts read through
a mocked useToast instead of a window global; baselines lowered where the
counts fell
- the assistant keeps facts for a trip (remember_for_trip) as a change card,
applied into the user's trip notes; existing expenses can be changed
- change card fields as a few icon chips with the rest folded under Details,
the day band marks only what changes, the week grid as compact tiles
- no map button on receipts of things that have no place on the map
- tests follow the composer's accessible name, the receipt's status icon and
the GL map's container; the hanging stream helper honours an early abort
- Postgres probe: qualified columns in the upsert updates; entities restored
to the generator's output; strict ratchet reads the SDK client as untyped
The three AI addons share one tile in the integration column; a dialog holds
the connection, the assistant's limits and AI parsing, which now reads with
the assistant's connection unless an install keeps its own. Existing parsing
configs stay where they are and keep winning.
The chat sees the map on screen, offers next questions under an answer,
shows its live step in the window head, previews waiting changes inside the
day list, checks whether a day still works, folds a multi-day plan into a
week grid, keeps per-trip notes and can estimate costs. An empty or cut-off
answer gets one push instead of ending silently, and the assistant's own
step and size limits are raised.
Adds an in-app assistant that reads and plans the open trip. Users connect
their own provider (Sign in with ChatGPT, or an Anthropic, OpenAI, Gemini or
OpenRouter key) or use the one the admin set up, including a custom
OpenAI-compatible server. It is an addon and only shows while a connection
is usable.
The chat runs on the existing MCP tools, pinned to the trip. Every change is
held as a change card the user applies or discards, and applied changes can
be undone. On the desktop it is a floating window that can be moved and
resized; on the phone it is fullscreen. Search results show as place cards
that can go onto a day, weather lookups as a strip, proposed places as
ghost markers on the map with a before/after switch, and an answer can be
shared into the trip chat. Booking files can be dropped in and are read
into change cards.
GET /api/journeys/:id/map-layers answers the planned places of the linked
trips (one per place, with every day the plan puts it on and the entry
written about it), the trips' GPX routes while the track switch is on, and
the located stays entries were taken over from. It goes through
journey_trips, so a trip nobody wrote about yet shows its plan too. Every
source is read on its own: one that fails says unavailable and the others
still draw. A position that is not one is left out, never moved.
GET /api/public/journey/:token/map-layers cuts the same on the server by the
link: nothing while the map is not shared, the plan and its routes only with
share_plan, the recorded days only with share_recordings and only beside a
published entry, stays only of published entries. Migration 046190 adds both
flags to journey_share_tokens, off by default, so a journey that is public
already shows what it showed before. The share link REST route and the MCP
share tool take them alike; get_journey_map_layers reads the layers over MCP
on the same access, a route by its size only.
Journey Settings get "Prepare recorded days automatically" under GPX Tracks,
off by default and set by the owner. With it on, the Dawarich background
check also reads the linked trips of the journey for the owner and every
editor with a connection of their own, within the same 400 day reach and
windows as the trip sync, and only for new and changed stays: a trip the
person travels on is not asked twice, and the stays of a trip they are not
on are filed under no trip. Nothing is written into the journey; the stays
wait at their day, and a stay taken over or left out is never offered again.
Under the switch the caller sees when their stays were last fetched, or why
not (no connection, no dated trip, background check off, the reason of the
last failure beside the state still shown), with a "Fetch now" that the
switch also runs when it is turned on. A day of the trips that is not final
yet carries a dot on its take-over button, on the phone in the stay sheet,
and the tooltip says whether it is still running, was fetched before it
ended, was not fetched yet or shows an older state after a failed refresh.
The day review says so for today too.
Server: migration 046180 adds journeys.auto_prepare_recordings and
journey_recording_prep (the last attempt, last success and last error per
journey and person). JourneyRecordingPrepService in the journey domain lists
the prepared trips and keeps the status; DawarichSyncService walks those
trips in the same loop as the person's own and records each journey's
outcome. GET /api/journeys/:id/recordings/prep answers the caller's status
under a @trek/shared contract, the MCP tool get_journey_recording_prep reads
the same, and update_journey takes auto_prepare_recordings like the PUT
contract. The column stays out of the plugin output. All 27 locales carry
the strings; the Journey, Dawarich and MCP tool wiki pages describe it.
A journey on a phone with stays on days that hold no entry offers each waiting day in the Dawarich stay sheet, so a day can be built from its stays alone; desktop and phone read the same days through openStayDatesWithin. A position picked for a group falls back to the longest stay with a place once that stay is unchosen or split off, a merged entry of stays without a position goes out unpinned instead of being refused, and a stay taken over meanwhile is never joined across. review_journey_day_stays is no longer marked read-only, since it refreshes the caller's stored stays. The already-taken notice no longer claims the stays were in this journal. An e2e case checks the drafts stay off the public share.
A journey day now has a list button in its header beside the route, with the
number of stays still waiting, and the row of the day's Dawarich stays gets a
"Take over day" button; on a phone the journey's More options offers it for
the day of the card shown. The dialog reads the day again from the caller's
own Dawarich and shows the route and the stays on one map, numbered in the
order they were lived and pinned in the colour of the day, beside a list with
times, durations, the stays Dawarich only guessed or is unsure about, the ones
without a position, the entries the journal has on the day and how many people
will see the drafts. When Dawarich does not answer, it says why and shows the
stays of the last sync.
Open stays start chosen, the same place twice in a row starts joined as one
entry under a title and a chosen map position, any two neighbours can be
joined or kept apart, a name can be corrected and a stay left out. "Day
report" turns every chosen stay into one entry that lists them with their
times, at a chosen stay or off the map. The day's route can go along when the
journey keeps none yet. Nothing is written before "Take over N drafts", and a
blank name or title holds it back with the reason. Stays left out are dismissed
and can be restored in the stay list.
The take-over writes drafts only, never touches what the journal holds, and is
one transaction: every stay is checked to be the caller's and of that day, the
entries, the stays they hold and the marked suggestions commit together, and
the entries are announced after the commit. A stay the journey holds already
is reported instead of written twice, so a retry after a lost answer is
harmless; a failed attempt reads the day again and keeps the choices, and a
route that fails after the drafts can be kept again alone.
Server: migration 046170 adds journey_entry_stays (one row per stay of an
entry with its own name, place and times, unique per journey, source, person
and visit, gone with its entry), owned by the journey domain through
JourneyEntryStaysService. The single stay accept records its stay there too,
over REST and MCP alike. DawarichSyncService.refreshDay reads one day's visits
and upserts them without flagging anything missing; DawarichDayStaysService
lists and settles a day's stays. JourneyDayImportService in journey-recordings
serves POST /api/journeys/:id/recordings/:date/stays/review and /import with
@trek/shared contracts, and the MCP tools review_journey_day_stays and
take_over_journey_day_stays run on the same service and checks. createEntry
can write without announcing for a caller that announces after its commit.
The desktop dialog and the phone sheet of the route (JRN-01) and of the
take-over share one frame each. All 27 locales carry the strings; the Journey
and Dawarich wiki pages and the MCP tool list describe it.
A day header in the journey timeline now carries the Dawarich mark for an
editor whose Dawarich answers (dimmed with the reason while offline). It opens
a dialog that names the journey and the day and reads that one local day from
the caller's own connection: a sketch of the route, distance, first and last
time, pieces, gaps without recording, modes and whether tracks or thinned
points answered. The day is cut in the zone Dawarich records in where its
answer shows one, else in this device's, and a picker cuts it in any other.
Nothing is kept before "Keep route". On a phone the same sheet opens from the
journey's More options for the day of the card shown.
What is kept is the preview that was seen: the save reads the day again and
refuses with recording_changed when its hash moved, and the dialog shows the
new version. The same day taken over twice keeps one version and announces
nothing; a different version is replaced only on "Replace", never touching an
entry or text. A failed read throws before any write, so the kept version
stays. Gaps stay gaps: points are split where silence ends elsewhere, and no
day is joined to the one before. The dialog says how many other people will
see the route; editors may keep their own recording.
Kept days show in the day header (distance, whose recording, since when,
still running) and on the desktop and phone journey maps in the colour of
their day's pins, winning over the live trail of their date, and stay drawn
without Dawarich. A trip's GPX track is now drawn dashed as the plan, recorded
routes solid. Shared links do not show recorded days. Remove route asks first
and works without a connection.
Server: migration 046160 adds journey_recorded_days (one row per journey,
local day and source, segments as JSON, zone, window, fetch time, hash,
importer kept past their account). The journey domain owns the rows through
JourneyRecordedDaysService; the new journey-recordings domain takes a day
over through DawarichTracksService.fetchDay, which reads past the cache, asks
a window wide enough for every zone and refuses with DomainErrors. REST under
/api/journeys/:id/recordings (list, preview, PUT and DELETE /:date) with
@trek/shared contracts, and MCP list_journey_recorded_days,
preview_journey_recorded_day, save_journey_recorded_day and
delete_journey_recorded_day on the same service and access. A new user event
journey:recording:updated keeps every open map current. JourneyTrack gains an
optional kind. All 27 locales carry the strings; the Journey and Dawarich wiki
pages and the MCP tool list say what it does.
A saved ski tour now opens with the line, heights, time, build and
planned legs it was saved with, read from the place this device keeps,
and nothing is asked of the piste router until a point, an option or the
discipline changes. Opened offline, while TREK Places is down or with
piste routing switched off, it no longer turns every leg straight, and a
rename saves the planned tour as it was. A kept line whose length is not
the tour's is another save's, and the tour is planned afresh.
- the save names the legs the router planned (ski_planned_legs, checked
against the tour's own legs); get_tour_plan reports each leg routed or
drawn from them instead of from the build alone, a tour that names none
counting as planned only with the router's time
- copies keep ski_options in every path: in the trip, taken over, kept
in a collection and made from one (collection_tours gets the column in
the same unshipped migration)
- the route summary shows the data build, the km of runs, the height
skied down on them and the lift rides, and the credit the server sends
in place of a fixed one; a tour opened as saved says so
- answers refused while piste routing is off, or planned on an older kept
copy, are not remembered, so they are asked again later
- terrain heights between known ones are filled in one walk along the line
- i18n: five summary keys in all 27 locales
The router thread now takes one job at a time from a queue kept on the
main thread, so a job's clock starts when the thread takes it up. A job
that times out ends only its own thread, and only while that thread is
still the current one; the jobs in line go to a fresh thread instead of
waiting out their own timers and killing the new thread in turn. A
thread that dies fails the job it held, and a thread that cannot start
fails each job on its own.
- the daily refresh fetches changed areas two at a time and stops once
the TREK Places breaker opens; the rest keep their older copies
- the winter index cap is on the unpacked JSON fetch hands over, so it
follows from the host's 8 MB gzip cap (64 MB) instead of matching it
- a build name is read to the rule a ski tour stores it by, so a build
TREK could not save is refused when the winter layer is read
- an unused import in the piste router contract is gone
A ski tour now saves the options its legs were planned with and the piste
data build they ran on, as one JSON column (a new migration, since the
ski migrations have shipped on this branch); the server keeps only the
options its discipline reads. Both stay off every other type. The refusal
of a ski time is lifted: the piste router works one out, and the planner
sends it only when every leg followed the piste data. get_tour_plan names
the legs of a tour the router planned as routed, and list_tours says what
the new fields hold.
A new Nest domain plans the legs of a ski and winter tour on the runs,
lifts, cross-country trails, ski touring and winter paths the Places host
builds from OpenStreetMap, one SGR1 graph per ski area. TREK fetches the
index once, downloads an area by its id when a planner first plans in it,
checks it against the index's SHA-256 and keeps it in data/ski; no
clicked point ever leaves the server. A leg whose area is slower than two
seconds answers loading, and an older copy stands in while the host is
down.
The reader decodes the writer's golden file byte for byte, including the
two edge flags it added (area_has_run, type_unknown). The search is A* in
a worker thread over a profile per discipline (alpine, snowboard,
cross-country by technique, touring, snowshoe, winter hike) with the
times and penalties of the cost model, up to five snap candidates per
point, lift stations as entry points, legs that go on from each other,
pieces of a split network stitched by their boundary nodes, other ways by
the penalty method, and a failed leg explained by relaxing one rule at a
time (difficulty, lift class, technique, riding down, uphill) without ever
handing out the relaxed line.
POST /api/trips/:tripId/ski/route, GET .../ski/area-near and
GET .../ski/network/:areaId sit behind Tours and the trip, with
plan_ski_route as their MCP twin. A daily job looks for a new build and
sweeps areas unused for 90 days. The admin switch ski_routing_enabled is
on wherever TREK Places is.
The ways and routes layers now count failures on a breaker of their own, so
busy or slow layer answers no longer send place search to the fallback for a
minute, and a layer's 404 no longer clears the count of search failures.
When a later batch of /v1/ways fails, the batches that answered are kept and
only the ids still open go to the Overpass mirrors.
The wiki, the compose file and the chart now say that the way ids of a planned
tour and the horse routes box go to the TREK Places API.
The trail grades under a tour (BIK-05, MAP-03, MAP-04, RUN-02) and the
horse routes around a map view (HOR-01) asked the public Overpass mirrors.
The places host now serves both from its own planet build: GET /v1/ways
for the tags of ways by id and GET /v1/routes for route relations per box,
each in Overpass element form.
trek-places.client gains trekPlacesWays (batches of 500, one after
another) and trekPlacesRoutes, both through getJson with its off switch,
breaker, timeout, size cap and instance header, each element validated on
its own. TrekPlacesLayersService in the maps domain hands their answers to
the tours services, or why there is none (switched off, layer not built,
failed), and logs that once per layer until it answers again. Both tours
services ask the layer first and fall back to the existing Overpass path;
the responses TREK sends its clients are unchanged.
A fresh install gets Tours enabled from the addon seeder. An install that
already has the row is switched on by a migration that runs once, so an
admin who turns Tours off afterwards keeps it off.
The get_tour_plan description named the estimate of a tour "such a field"
right after the values of a recording, so it could be read as one of them.
It now follows the sentence on derived fields and says outright that it is
one of them, worked out from distance and heights or the pace of a run.
The arrow between a tour's start and end points from the start to the end
in a right-to-left language as well. A tour just added to a day, which
carries no tour_place_id until the next refresh, now says its overrun like
the row above it. A phone row of an import with a total the travellers
entered shows that one time, the estimate joining its words as on the
desktop. The German overrun line reads naturally without an item name.
Maurice