tajniak81andClaude Opus 5
9bd5c523c4
The integration cascade stored its top layer differently from the two below it: org (L2) and user (L3) plugin config lived in PocketBase, in a pluginSettings field, while the global (L1) layer sat in a plugins.json next to the binary. That split was accretion rather than design - the file was the whole store in the v1 MVP, and the per-tenant layers were later built on PocketBase and layered on top of it instead of replacing it. It also cost something real. plugins.json was a second state store with different durability from pb_data: its own volume, its own ownership, its own backup. Losing pb_data is unmissable; losing api_data was silent, which is how "every plugin comes back disabled after a redeploy" happened. L1 now lives in the app_settings collection - one record keyed "global", holding its settings in a pluginSettings field, the same mechanism and the same field name the layers below use. The documents still differ in shape, because only L1 carries enable state and the registration of external plugins, but the storage is no longer a special case. The Manager grows a Store seam (PocketBase in production, file for the import, memory for tests) and, more importantly, a loaded gate. Settings in a database mean the store can be unreachable at boot - a cold stack, or a service account still to be set from the panel. That must not read as "no plugins configured", or the first save would write emptiness over real settings. So until a read succeeds the Manager stays unloaded, every mutation is refused, /api/admin/plugins* answers 503, and a background retry backs off to two minutes. The same gate covers a document that will not parse: it is never replaced by one built from an empty map, which is a stronger guarantee than the .corrupt backup it replaces. Writing to a store also revealed a hole in the previous fix. Classifying a save failure as errPersist was left to each Store, and a store that returned a plain error would fall through to the "saved, but the plugin failed to start" branch and be reported as a 200 - the same silent-success bug through a different door. The Manager now classifies, whatever the Store returns; a test pins it. Upgrades are automatic: on the first boot that finds no settings in the database, an existing plugins.json is imported and renamed to plugins.json.migrated. The import is refused if the store is merely unreachable, or if the file does not parse, so a stale or broken file can never overwrite live settings. /data is still needed - the panel rewrites .env there when it retargets PocketBase - but plugin settings no longer depend on it. 21 tests in internal/plugins cover both stores, including the production path against a fake PocketBase: create-then-update of the singleton, round-trip across a restart, an outage that leaves settings intact, a missing collection reading as not-ready rather than empty, and the import running exactly once. go build, go vet and go test ./... pass. Schema changes are mirrored into scripts/setup-pocketbase.mjs as that file requires. Not verified: no Docker CLI here, so no image was built and the bootstrap of app_settings against a real PocketBase is untested outside the fake. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
DriverVault — Docker (multi-container stack)
Three containers — PocketBase, API Server, Web App — on one compose
network. This is the deployment to use unless you specifically want everything
in a single image; for that see ../Docker-AIO.
Browser ─► Web App BFF (:8090) ──/api/*──► API Server (:8080) ─► PocketBase (:8070)
Only the Web App port is meant to be public. The API Server and the PocketBase
admin UI are published for convenience and, in the prod file, bound to
127.0.0.1 by default.
| File | Use |
|---|---|
docker-compose.yml |
builds from source in this repo — for development and local testing |
docker-compose.prod.yml |
pulls prebuilt images from the registry — for deployment |
.env.example / .env.prod.example |
copy to .env for the matching compose file |
pocketbase/ |
the PocketBase image (official release binary on alpine) |
Run it
cd Docker
cp .env.example .env # then edit — PB_ADMIN_* have no safe defaults
docker compose up -d --build
Production, from the registry:
cp .env.prod.example .env # then edit
docker compose -f docker-compose.prod.yml pull
docker compose -f docker-compose.prod.yml up -d
Then: web app on http://host:8090/, the API Server's superadmin panel on
http://host:8080/, PocketBase admin on http://host:8070/_/.
First boot
Both steps are idempotent, so restarts and upgrades are safe:
- PocketBase upserts its superuser from
PB_ADMIN_EMAIL/PB_ADMIN_PASSWORD. This is the only way to create the first superuser — the REST API cannot bootstrap it. The API Server then authenticates with the same credentials. - The API Server creates any missing collections and reconciles existing
ones, then creates the first app
superadminfromDRIVERVAULT_SUPERADMIN_EMAIL/_PASSWORDif no such user exists. SetPB_BOOTSTRAP=falseto skip once the database is established.
No manual setup-pocketbase.mjs step is needed here — the server runs the same
schema reconcile itself.
Volumes
| Volume | Holds |
|---|---|
pb_data |
the PocketBase SQLite database and uploaded files |
api_data |
the .env the API Server panel rewrites when a superadmin retargets the PocketBase connection (plugin settings live in pb_data, with everything else) |
Both default to Docker-managed named volumes. In the prod file, set PB_DATA /
API_DATA to absolute host paths for bind mounts instead.
The API Server serves as the unprivileged
appuser, so/datahas to be writable by it. Its entrypoint arranges that itself: it starts as root, takes ownership of/dataifappdoes not already hold it, then drops privileges. So a bind mount to a root-owned host path needs no manualchown, and neither does a named volume left over from an image that ran as root. The one case it cannot fix is a container forced to another user (user:in compose,docker run --user), where the entrypoint has no privileges tochownwith — prepare the host directory yourself there. PocketBase runs as root, soPB_DATAis unaffected either way.
Plugin enable-state and global config used to live in a plugins.json on this
volume, which made them the one piece of configuration a lost volume could erase
without anyone noticing. They are now in PocketBase, backed up with pb_data
like everything else. An existing plugins.json is imported automatically on the
first boot after the upgrade and renamed to plugins.json.migrated; keep the
volume mounted for that boot.
Charger control (OCPP)
Chargers in own/proxy mode dial in to /ocpp/{serial} on the API Server
port, authenticating with a per-charger control token in an OCPP Basic-auth
header. A plaintext ws:// would put that token on the wire in the clear, so
OCPP_REQUIRE_TLS defaults to true and non-TLS connections are rejected.
This stack serves plain HTTP, so to actually use charger control you need to
terminate TLS in a reverse proxy in front of it and set OCPP_PUBLIC_URL to the
public wss:// base (behind a proxy, deriving it from request headers is
unreliable). OCPP_REQUIRE_TLS=false is for trusted networks only. You will
also need API_BIND set so the proxy can reach the port.
Notes
docker-compose.ymlbuilds the API Server and Web App from../API Serverand../Web App, so run it from this directory with the repo checked out.- The Web App's Vue bundle is built with an empty
VITE_API_BASE, so the browser uses same-origin/apiand the BFF proxies it — no CORS in play. CORS_ALLOW_ORIGINStherefore only matters if a browser calls the API Server directly. Native mobile apps are not subject to CORS at all.