Commit Graph

231 Commits

Author SHA1 Message Date
Maurice 1ea08dcf0a fix(db): refuse a Postgres boot that would leave a SQLite install behind
Setting TREK_DB_URL on an existing SQLite install booted a new, empty
instance beside its data. The Postgres bootstrap now refuses an empty
database, or a migrated one without a single account, while the data
directory's travel.db holds accounts, and names the move command. A file
that cannot be read counts as holding them. TREK_DB_IGNORE_SQLITE_FILE=true
starts anyway, with a warning; the file is opened read-only and only when
the database holds no TREK data.
2026-11-17 23:03:54 +01:00
Maurice a26db47cb6 refactor(db): keep the touched files within their size baselines
The users identity guard moves to repositories/_shared/user-identity.ts
(Users.repository.ts re-exports UserIdentityTakenError), the flag text pass
of the entity generator to scripts/generate-entities-flags.ts, and the
Wallet pass query takes the repository's EntityManager. The size baseline
of Users.repository.ts is lowered to what it holds now (1445 to 1421); a
few exports only their own file used are no longer exported.
2026-11-17 23:03:53 +01:00
Maurice 693acd5580 fix(db): reset the admin on Postgres and keep the key rotation off it
reset-admin.js opened data/travel.db with better-sqlite3 whatever the
engine, so on a Postgres install it created a stray SQLite file or reset
the admin in a leftover one. With TREK_DB_URL set it now resets the account
in Postgres through the built server's engine (its pool, SSL and write
lock, in one transaction), and refuses a TREK_DB_FILE beside the URL; the
SQLite branch is unchanged. migrate-encryption.ts cannot re-encrypt the
secrets in Postgres yet, so it now stops before it opens or writes any file
when TREK_DB_URL is set. The wiki says what each does on Postgres.
2026-11-17 23:03:52 +01:00
Maurice 54b843ae29 fix(db): store addons.enabled as the integer flag on Postgres too
addons.enabled was the one flag MikroORM's boolean type declared as a
Postgres boolean, while SQLite and every other flag hold 0/1, so an addon
migration's usual SET enabled = 1 would fail there and a Kysely read would
answer true instead of 1. The column now uses IntegerFlagType: the entity
still reads a boolean, both engines store 0/1, and the regenerated Postgres
baseline declares it bigint default 0. The entity generator renders the
type for the columns in BOOLEAN_COLUMNS.
2026-11-17 23:03:52 +01:00
Maurice 2630d7e8ec build(docker): ship the PostgreSQL 18 client tools from the PGDG repository
A client handles every server up to its own major and refuses a newer one,
so the image needs the newest supported major, 18, which Debian 13 does not
ship. The PGDG repository is added with its signing key kept in the
repository and its component pinned to 18.
2026-11-17 23:03:51 +01:00
Maurice 4a3ad562a7 fix(db): match a stay-linked booking on Postgres in the public stay check
publicStayExists cast reservations.accommodation_id straight to integer,
which Postgres refuses for the stored '14.0', so the calendar feed and the
share links failed for every trip once any booking was linked to a stay.
The correlation now goes through accommodationIdMatches, which compares
through the new dialect helper integerTextEqualsKysely, and the Wallet pass
context uses the same helper instead of its own numeric cast.
2026-11-17 23:03:50 +01:00
Maurice 0cdfb9eb42 ci(db): run the Postgres probe and lane on Postgres 14 and 18
TREK supports Postgres 14 to 18, and CI tested 17 only. One server-postgres
job per major now runs the probe and then the lane against the same service,
so the change keeps the job count of the two jobs it replaces. The lane's
global setup refuses a service whose major is not TREK_TEST_PG_MAJOR, and
the image's Postgres boot moves to 18, the major its client tools target.

scripts/ci/workflow-jobs.test.mjs fails when a job other than the scan is
missing from ci-ok's needs, and changed-areas pins that migrations,
repositories, the engine and the lane's files always run the job.
2026-11-17 23:03:50 +01:00
Maurice 293a48c26a fix(db): keep check:pg-baseline speaking once the baseline is frozen
A frozen baseline still has its manifest verified, and the check now lists the
migrations that run on Postgres after the cut and what holds them. Every
problem it reports says what to do about it.
2026-11-17 23:03:50 +01:00
Maurice 80fca10784 feat(db): hold migrations after the Postgres cut to engineDdl in lint:dialect
A migration newer than the baseline's through (read from
assets/postgres/baseline.json) fails on a schema statement outside engineDdl,
a plain string column for addColumnIfMissing or a SQLite-only helper, and on a
SQLite spelling in an engineDdl postgres member. The advice names the two
steps, and the baseline cannot hold an allowance for these rules.
2026-11-17 23:03:49 +01:00
Maurice 830cd6e3b5 feat(db): print the Postgres side of a new migration with gen:pg-migration
The script builds a scratch database the way a boot does (the baseline, then
the migrations after its cut) up to the target migration, compares it with the
entities by PARITY-PG's rules and prints the remaining statements as the
postgres member of engineDdl. The scratch database is always dropped.
2026-11-17 23:03:49 +01:00
Isaias Tavares 2e3762afe1 feat(db): run TREK on PostgreSQL
TREK runs on PostgreSQL when TREK_DB_URL is set, and on SQLite as before
otherwise: an engine port under src/db/engine, the dialect helpers under
src/db/dialect, a Postgres baseline for the migrations up to the cut,
pg_dump/pg_restore backups, a probe of every repository method against a
real server and a Postgres test lane in CI.
2026-11-17 23:03:33 +01:00
Mariano 33d074c786 feat(places): tags on places, owned by the trip
A place keeps its one category, which decides its marker, and can now carry
any number of tags as extra labels to filter by.

- Tags belong to the trip (tags.trip_id): every member sees the same list and
  whoever may edit the trip's places adds, renames and deletes. Personal tags
  stay what they were.
- The place form picks them on desktop and mobile through one shared hook,
  usePlaceTags, and can add a new tag to the trip on the spot.
- The existing category filter of the places list lists the tags as well.
- Tags show on the place (inspector pills, phone place sheet); the marker is
  untouched.
- MCP: create_place, create_and_assign_place and update_place take `tags`,
  and list/create/update/delete_trip_tag manage a trip's tags.
- /api/trips/:tripId/tags and the tripTags:changed event keep every client
  of the trip in step.
2026-11-17 08:44:10 +01:00
Maurice 6789ee1cb6 refactor(server): keep the grown files at their size baselines; centre the inbox dialog
- auth.service: how an admin form value is stored (validated, encrypted, a
  masked echo kept) moves to the app settings registry as storedAdminValue
  and isMaskEcho.
- reservations.service: a booking's origin (withoutOrigin, originColumns)
  moves to reservation-origin.ts, its input shapes to reservation-inputs.ts,
  exported on from the service so importers stay as they were.
- Reservations.repository: the joined projection, the origin lookup and the
  origin columns move to _shared/, like the Wallet pass context before.
- test-sql: MailIngestApplyService.prepare is marked as no SQL statement.
- The pending bookings dialog opens centred, and says 'nothing waiting' once.
2026-11-13 20:28:17 +01:00
Maurice 783f679124 chore(server): lower the eslint baseline to what the code holds now 2026-11-13 19:38:57 +01:00
Maurice 1aa586d829 test: type fixes the CI typecheck found, and the lowered test-new-service baseline
- mailInboxModel.test: actionsOf takes the entry alone since the AI reread left.
- booking-origin.test: its endpoint helper gives the sequence an endpoint has.
- test-new-service baseline follows booking-import's single service factory.
2026-11-13 19:20:09 +01:00
Maurice 7d3ed21a03 feat(mail-ingest): the transport settings in the admin form, each overridable from the environment
The IMAP box, the webhook secret and the address template join the admin
settings form (masked, encrypted, held by the operator on a managed install),
and MAIL_INGEST_* variables override them through the mailIngestConfig token.
The audit row masks every masked key, and key rotation covers the new secrets.
2026-11-13 17:43:48 +01:00
Maurice baa906fb6b feat(mail-ingest): tables for forwarded booking mails and their bookings
Five tables carry the mail import: forwarding addresses (token hash plus the
encrypted token), a person's switches, the readers' cursor and lease, every
mail as it arrived, and each booking read from it with its checks and fate.
Repositories read the trips a booking may go to by column, so the import
weighs a booking without reaching into five other domains.
2026-11-13 17:42:43 +01:00
Maurice 154563f34d feat(flight-live): follow booked flights live from keyless ADS-B feeds
A new flight-live domain polls the open feeds adsb.lol and adsb.fi (no key,
no cost; adsb.fi only on self-hosted installs, its terms are non-commercial)
for every booked flight leg in its watch window, derives phase, take-off,
landing, delay and an arrival estimate, and pushes changes as flight:live.
GET /api/trips/:tripId/flights/live, the MCP tool get_trip_flight_status,
notifications for take-off, landing and delays, and the admin switch
flight_live_enabled (FLIGHT_LIVE_ENABLED wins) come with it.
2026-11-13 15:29:03 +01:00
Maurice 71cc97822a fix(ci): the second round of the quick-win checks on the server
oidc_auto_redirect is managed-locked in the registry too, costs_mode is read
as the mode it is, the days of a dated trip are read through the factories,
and the dashboard's country order moves out of the atlas service into a
module of its own, so the service keeps its size.
2026-11-13 12:57:56 +01:00
Maurice 721a610a70 fix(ci): green the quick-win round on the server, the shared contracts and the plugin facts
The trips entity takes costs_mode as the generator writes it, the plugin
output contract publishes it, and the hand-built test schemas carry it. The
new MCP fields come from @trek/shared, the link tool tests join the collab
file that already mocks the database, oidc_auto_redirect is the operator's
on a managed install, and the tests follow the KML service, the dated trip
days and the zod pipe.
2026-11-13 12:40:15 +01:00
Maurice 3e98938c95 feat(export): places as KML for Google My Maps (#1866)
Google Maps takes no list from outside, but My Maps imports a KML layer.
The trip export dialog and the phone export sheet get "Places for Google
My Maps" (GET places/export.kml, MCP export_trip_kml), and a list's export
menu gets KML next to GPX. The KML codec gains a pin writer; the list file
mapping moves to collections.helpers to keep the service within its size.
2026-11-13 10:17:12 +01:00
Maurice a70eea5bf2 feat(costs): let a trip keep its costs as totals only (#1714)
Travellers who track what a trip costs but never split it saw an
Unfinished warning on every expense and a settle-up they had no use for.
A trip now has a costs mode (appended migration, trips.costs_mode, split
by default); totals leaves payers, shares and settling up out of Costs on
desktop and phone through the same gate as a trip of one, and saves new
expenses without a payer. PUT /api/trips/:id and update_trip take it on
budget_edit. The split chips move into their own component.
2026-11-13 10:17:03 +01:00
Maurice 736dc5dda9 chore(lint): lower the non-null assertion baseline the AirTrail test change freed 2026-11-12 17:47:05 +01:00
Maurice ae26a6c61c refactor(airtrail): save the settings in one transaction
The address and the key it may no longer carry now change together, as
they do for Dawarich, so the transaction ratchet drops the AirTrail entry.
The tests that pinned the old no-transaction shape now pin the new one.
2026-11-12 17:31:56 +01:00
Maurice 948c9ef57b fix(atlas): show Ireland's 26 counties instead of its four provinces
geoBoundaries has no county layer for Ireland: its ADM1 is the four
provinces and its ADM2 the 166 local electoral areas, so the Atlas showed
Ireland as four provinces next to a UK of counties.

The Atlas builder can now take a country's regions from another source.
Ireland's come from Tailte Éireann's 26 counties (CC BY 4.0), with Dublin,
Cork, Galway and Limerick each one county, under their ISO 3166-2:IE
codes. Those outlines come at full detail, so the builder simplifies them
itself and keeps the borders two counties share identical on both sides,
so no point between them falls into neither. ATLAS_ONLY rebuilds just the
named countries without downloading the whole world.

A migration clears Ireland's cached place regions so they are worked out
again as counties, moves visited rows that name a county onto its code,
and hides each county of a province a user had hidden. Provinces a user
marked as visited are kept as they are.

Fixes #2922
2026-11-11 21:26:31 +00:00
Luffy a3622e91b6 feat(vacay): implement Nager.Holidays v4 integration and adapt holiday data structure 2026-11-11 21:04:07 +00:00
John Jahn 8e8cff8105 fix(reservations): keep day_id on partial updates of time-only bookings
A partial update that left day_id out (a notes-only PUT, an MCP
update_reservation or update_transport call) re-derived the day from the
stored reservation_time whenever one was set. A time-only value such as
"12:00" resolves to no day, so the booking's day_id was written as null and
it vanished from its day in the planner. The end_day_id branch did the same
when a time-only reservation_end_time was sent.

The day is now derived only when the time carries a date that resolves to a
trip day; otherwise an omitted day_id or end_day_id keeps the stored value.
A dated time still re-pins the day, an explicit day_id null still clears it,
and a cleared end time still clears end_day_id. REST and both MCP update
tools share updateInTx, so the fix covers all three.

Tests: service unit cases (RESV-SVC-044..048), an e2e case per behaviour
on PUT /api/trips/:tripId/reservations/:id, and an update_reservation MCP
case, all seeded with the ORM factories.

Fixes #3025
2026-11-10 10:05:22 +01:00
Maurice 6b086eaed5 refactor: the login's code step and the reservation row shapes get modules of their own
Both files had grown past their size limit with this round's fixes. The
two-factor step moves to pages/login/MfaCodeStep.tsx and the reservation row
types to reservation-rows.ts, re-exported from the repository, and both size
baselines come down to match.
2026-11-10 09:55:36 +01:00
Maurice a2aeeb32c8 fix(ci): keep the reservations repository at its size baseline
The wallet pass commit had reformatted the whole repository file, which put
it 200 lines over its size baseline. It is back to its old form, in the format
baseline again, and the pass-context query lives in a helper of its own.
2026-11-07 12:07:44 +01:00
Maurice 354e30a075 fix(ci): wallet passes pass the server gates
- TREK_WALLET_PASS_URL read as a secret to the env check (the word PASS); it
  is TREK_WALLET_SIGNER_URL now, everywhere it is documented
- findPassContext compared the TEXT accommodation_id with an integer id, which
  Postgres refuses; it compares them as numbers now, as both databases accept
- the wallet-pass service test takes its subject from createTestModule instead
  of building it by hand, and no longer names the DOM-only BodyInit type
- Reservations.repository.ts is formatted, so it leaves the format baseline
- the wallet button test reads the translated texts the render helper gives
2026-11-07 11:45:00 +01:00
Maurice 4bc764eb6a test: hold the audit follow-ups to the test, lint, theme and size ratchets
The packing row tests read the TREK tooltip and the accessible name instead
of the title the badges no longer carry, the journey delete suite takes its
service from the module over a snapshot db and seeds through the row
factories, the storage notifier suite builds its subject once, the POI
layer follows the view only while it has POIs, the collections repository
reads the list's places through its own query, and the baselines the moved
code left smaller come down with it.
2026-11-05 15:05:59 +01:00
Maurice 402a02a302 fix(docker): say what the volume-root check leaves alone, and skip the walk for a custom fsGroup
The entrypoint comment, the Helm chart README, the wiki and the ssrfGuard
note now describe what the shipped check does.
2026-11-05 14:36:22 +01:00
Maurice 50ea9a74c9 fix(mcp): keep budget categories to the Costs keys, and say what a journal entry's visibility does
An assistant writing expenses through the MCP was told "e.g. Accommodation, Food,
Transport" and stored exactly that, so "Food" and "food" sat side by side in the
MCP totals while the Costs tab already read both as one. A typed category in any
case, or a common name like Hotel or Flight, is now stored as its fixed key on
every path (REST, MCP, linked booking costs), the totals the MCP reads group by
that key, and the tools name the real keys. The client reads categories through
the same shared mapping. Any other text is still kept as given.

The journal tools said that a private entry stays off the share link. It never
did: the share link shows every entry that is not a draft, whatever its
visibility, and the web editor has no visibility at all. The tools now say so
and point to is_draft, which is what keeps an entry to the contributors.

The wiki explains why the location button needs HTTPS and how to run TREK behind
an HTTPS tunnel while still using it over plain HTTP at home, and that Android's
photo picker drops a photo's location while Immich and Synology keep it.
2026-11-04 14:26:25 +01:00
Maurice 74864f2543 fix(places): keep the places repository at its size baseline
The insert's note now says that a longitude is stored on the real world, in
one line fewer.
2026-11-04 12:09:29 +01:00
Maurice c6149cd64d fix(costs): move the settlement currency setup into the money helpers and hold the size baselines 2026-11-03 22:16:41 +00:00
roopesh 1625a10424 fix(server): refuse unwritable database in maintenance scripts (#2933) 2026-11-01 22:26:40 +01:00
Maurice 1c6a175ebc fix(storage): queue what a delete left unnamed past every undo, and tell a copied database apart
A delete or a replace no longer deletes inside the request. The owner's call
queues the keys in storage_release_queue, due after RELEASE_DELAY_MS (26 h:
the assistant's 24 h approval window plus a margin), and StorageReleaseJob
takes the due rows every ten minutes, at most 200 a tick and 50 keys a
question, asks every reference source again and deletes what none names.
It refuses what the nightly pass refuses (StorageReclaimGuardService: the
ledger, the required sources, a shared location, another installation's
owner marker, cached per category), and waits while a restore or a backup
runs or another storage job holds the slot. An assistant undo or a late
write that names the old value again keeps its object.

The install id lives in data/.trek-install-id, outside any backup and never
written by a restore. At boot and after every reopen the database is
reconciled with it: a database naming another installation (a boot restore,
a restart after a failed restore, a copied travel.db) loses its switch,
confirmations, reports, marks and queue before anything reads them.

A confirmation now stores the count the admin was shown and when; it covers
up to one and a half times that count (at least 100 more) for 14 days, is
used up once the share is back under a fifth, and a count the last report
no longer holds is refused as reclaim_confirm_stale. The admin state folds
a live confirmation in, so the prompt goes at once.

A pass reads each source once for every category and drops each category's
keys once judged; right before a batch it reads the category's sources
again, whole and in memory, instead of a LIKE per key. A narrowed read holds
at most 50 LIKE terms. A pass held off by a backup reports backup_running,
and a category the ledger never reclaims reports not_reclaimed even after a
stop.
2026-11-01 18:11:43 +01:00
Maurice d17457de56 fix(storage): delete what a delete or a replace left unnamed right away again
Leaving every replaced cover, place image and deleted journey photo to the
nightly pass kept them publicly fetchable by URL for a week at least, and
for good while deletion was off. Owners now hand the values a delete or a
replace left behind to StorageReleaseService once their write committed.
It asks every registered reference source, with the same parser as the
nightly pass, and deletes at once what no row names any more: an object a
trip copy, a journey or a saved place still shows stays, which the old
inline deletes got wrong.

The release holds the pass's guards (reclaimable categories at their root
only, never an owner marker, never while a restore runs, never with a
required source missing or an overlapping location), never fails the
request and logs at debug. It runs for trip covers (update, upload, trip
delete), collection covers and saved-place images, place images (update,
bulk update, delete), journey covers and journey or gallery photos
(their thumbnail goes with a file that went), avatars (change, delete,
account deletion), trip files (permanent delete, empty trash) and collab
note and chat attachments. A trip file row now goes before its bytes, so a
failing backend no longer fails the delete. The objects of a deleted trip's
places and files are left to the nightly pass: their rows are other
domains'.

PlacesService drops the collection places repository nothing read any
more. Size, warning and boundary baselines follow the shrunk files.
2026-11-01 16:18:21 +01:00
Maurice db0fbc1ccd fix(storage): stop deleting covers, place images and journey photos another row may still show
Replacing a trip or list cover, changing or deleting a place or saved-place
image and deleting a journey photo removed the bytes at once, without asking
whether a trip copy, a journey, a saved place or another photo row still
named them (#2924). Those paths now leave the object in place and the reclaim
job removes it once no row names it. A journey photo the user deletes takes
its local photo row with it when nothing else links it, so its bytes become
reclaimable instead of being kept by a row nobody sees.

A place image copied from a trip file is still dropped at once when the
update that would name it fails: that copy was never anyone else's.

Size and warning baselines follow the shrunk files.
2026-11-01 13:59:28 +01:00
Maurice 29298dcfe5 feat(storage): keep a mark per stored object no row names
storage_reclaim_marks records since when the reclaim pass has seen an object
unreferenced, keyed by category and key. A mark is never moved forward, so
its age is how long the object has been unreferenced without a break; it goes
once the object is named again or gone. Owned by the storage domain.
2026-11-01 13:37:44 +01:00
Maurice 7faf771f9c fix(docker): look at the volume roots only, and say when chown is refused
The ownership check still read every inode of the volumes on each start.
It now looks at the dirs and what sits directly in them, hands a volume
over once when it is not node's, and logs a warning instead of walking a
volume that refuses chown (NFS with root_squash).
2026-11-01 11:41:56 +01:00
Maurice 51b991f851 fix(docker): the container starts on large volumes and under runAsUser 1000
The entrypoint no longer runs chown -R over data and uploads on every start;
it only hands over files that are not node's yet, and logs when it does.
Started as a non-root user, it skips chown and gosu and runs Node directly.
The chart exposes podSecurityContext and securityContext for that setup.
2026-11-01 11:41:56 +01:00
Maurice f4e524f3c3 place-enrichment: the details lookup is a standalone cache fill
The service reads a community place's details before the result cache and
every other place's after it, so the lookup has two call sites of which one
runs. It only fills the details row cache and turns a failure into no
details, so it is tagged @txStandalone and the enrich baseline drops to the
four writes it holds without it.
2026-10-31 06:45:13 +00:00
Maurice c28604b891 planner: the details refresh reads through a repository, and the form drops what moved out
The fresh lookup behind New place details goes through placeDetailsRepo
(online-only, like the service identity) instead of api/ directly. The form
loses the Translate type and the comment that moved with the head band's
pills, which takes it under its line limit and out of the size baseline; the
maps service suite's test-size entry follows its shorter count.
2026-10-31 03:25:56 +00:00
Maurice 65abea6910 place-contrib: settle the first CI round
- The pause setting's type comes from its schema, so nullable fields read the
  same under the server's compiler settings.
- The place form passes the origin of a field through to the hook.
- Serial steps run through inOrder (src/nest/common/in-order.ts) and
  independent ones through Promise.all, instead of an await inside a loop. The
  bulk update in places.mcp.ts goes the same way, so its baseline drops by one.
- The status a call without a key gets is a named constant, not an envelope
  literal.
- A picture test no longer inherits the previous test's mock answer.
- The entity counts include the three new tables.
- The log keeps its failure per query instead of resetting it in the effect,
  and usePlaceForm sets its picture through one helper.
2026-10-31 03:25:55 +00:00
Maurice 182ed56fa4 places: let an admin share the details users typed with the TREK Places API
Every shareable field of a place now records where its value came from: typed
by a user, filled by a provider, written by the assistant, a plugin or an
import. Only values a user typed are ever sent, and only after an admin agreed
to the current terms for the whole installation, field by field.

The new place-contrib domain keeps the decision in app settings, queues values
in an outbox a cron job sends from, logs what the service made of each one and
can withdraw everything this installation sent. Development, test and demo
instances, an own TREK_PLACES_URL, TREK_SERVICES=off and
TREK_PLACES_CONTRIBUTE=off never send anything. Index results carry the
community version, so the details caches refresh when it changes.
2026-10-31 03:25:55 +00:00
Vini f56f6ae11e perf(atlas): keep the admin0 country index as Int16 arrays and send countries/geo from disk 2026-10-31 03:25:13 +00:00
Vini c94ff7421a keep looked-up regions as packed Int32 arrays under a 16 MiB budget 2026-10-31 03:25:12 +00:00
Maurice 2dbcc39832 chore(lint): lower the no-explicit-any baseline after the typed region lookup 2026-10-31 03:25:09 +00:00
Vini 7277e35a96 serve each country's regions from its own file, so the server no longer holds them all
The Atlas server no longer keeps the admin-1 regions of every country in memory. Each country's regions are now a pre-gzipped file of their own, sent to the map as they are.

**Why:** `GET /addons/atlas/regions/geo?countries=…` built `admin1Store` (every country's region text, kept until restart), then joined the requested countries, `JSON.parse`d them, and Nest stringified them again. On the 512 MB – 1 GB boxes TREK runs on, the first z≥5 view could OOM the server (same shape as #1576).

**What changes:**
- `server/assets/atlas/admin1/<ISO_A2>.geojson.gz`: 198 files, 14.85 MB, the same 3,447 features as `admin1.geojson.gz`.
- New `GET /addons/atlas/regions/geo/:code` with a Zod contract (`regionGeoParamsSchema` in `@trek/shared`: two letters, any case). It sends the country's file as it is, with `Content-Encoding: gzip` and the current 1-day `Cache-Control`, as `countries/geo` does. A country without regions gets an empty FeatureCollection. Anything that is not two letters gets a 400.
- `useAtlas.ts` fetches one URL per uncached country and still rebuilds the layer once. Each country has a stable URL, so a country evicted by `REGION_CACHE_MAX` comes back from the browser cache.
- `getRegionFeatures` reads only the country it needs. `admin1Store` is removed.
- `?countries=` keeps working for cached PWA clients. It joins the per-country files as bytes, without a parse.
- The migrations are untouched. `admin1.geojson.gz` still ships, only for `Migration20200101021500` and `Migration20200101031500`. This costs ~15 MB of disk in the image and no RAM. `build-atlas-geo.mjs` writes both, and ATLAS-BUNDLE-008 keeps them equal.

**Memory** (real `atlas-geo` module, first z≥5 view over Europe = 49 country codes, JS heap + Buffers):

| | peak | kept after |
|---|---|---|
| before, first view (builds the store) | +372 MB | +61 MB |
| before, later views | +115 MB | 0 |
| after, `/regions/geo/:code` | +4.7 MB | 0 |
| after, legacy `?countries=` | +59 MB | 0 |

**Tests:**
- New: `atlas-region-files.test.ts` (moved from `atlas.service.test.ts`, plus join and dedupe cases), the `regionGeoParamsSchema` spec, ATLAS-BUNDLE-007 (the file format the byte join depends on), ATLAS-BUNDLE-008 (bundle and per-country files are equal), ATLAS-016/017/018 (the new route over HTTP, including the 400).
- The `getVisitedRegions` tests (ATLAS-UNIT-019..024) used to pass only when `getRegionGeo`'s `beforeAll` had built the store first. They now also pass on their own.
- The shared, server and client gates pass. Baselines go down only where this change removed code (`no-console`, `no-explicit-any`, `no-swallowed-catch`, file sizes).

**Trade-offs:**
- A first zoom over Europe sends ~49 small parallel requests instead of 1. Over HTTP/1.1 the browser queues them 6 at a time. The total bytes are the same (4.4 MB gz).
- The per-country URLs are unversioned and cached for a day, like `countries/geo`. After an upgrade that changes a country's regions, a browser can show the old shapes for up to 24 h.

Approved on Discord in #github-pr: <paste the thread link here>
Related: #1576, #2559

- [x] Bug fix
- [ ] New feature
- [ ] Breaking change
- [ ] Documentation update

- [ ] I have read the [Contributing Guidelines](https://github.com/liketrek/TREK/wiki/Contributing)
- [ ] My branch is [up to date with `dev`](https://github.com/liketrek/TREK/wiki/Development-environment#3-keep-your-fork-up-to-date)
- [x] This PR targets the `dev` branch, not `main` *(wiki-only PRs are exempt)*
- [x] I have tested my changes locally
- [x] I have added/updated tests that prove my fix is effective or that my feature works
- [x] I have updated documentation if needed
2026-10-31 00:40:23 +00:00