The files leave the volume the database sits on

Every attachment — a document scan, a fuel receipt, a workshop invoice, a
photo of a part's box — has lived inside pb_data, in a directory beside the
SQLite file. One volume held both, so neither could be sized, backed up or
moved without the other. PocketBase can keep those bytes in an S3 bucket
instead, and now it is told to.

Nothing on the way to a client changes, because an attachment was never a
storage URL to begin with: it is fetched from GET /api/{records}/{id}/file,
which re-checks car access and asks PocketBase for the bytes as the service
account. PocketBase streams from the bucket through that same endpoint rather
than redirecting to it, so the web app, the phone and the plugin cannot tell
which side of the switch they are on.

The bootstrap that already creates the collections now writes PocketBase's
files-storage settings too, from PB_S3_*, on every boot and only when they
differ from what is already there — then asks PocketBase to prove it can reach
the bucket, and says so in the log either way. Two asymmetries are deliberate.
A read of the settings masks the stored secret, so a rotation of the secret
alone is invisible from here and needs another PB_S3_* to move with it. And it
never turns S3 back off: files already written to a bucket are reachable only
while PocketBase still points at it, so dropping the configuration would strand
them rather than undo anything.

Each deployment shape is one compose file with an .env example of the same
name, not a base plus an overlay to remember — six of each per folder, for
Docker and Docker-AIO alike: the plain one, .seaweedfs, .s3, and the three prod
twins. The SeaweedFS files run master, volume, filer and gateway as one process
and a one-shot init container beside it, because PocketBase never issues a
CreateBucket and SeaweedFS will not conjure one on first upload. The credentials
do double duty there — the gateway's only identity is also what PocketBase
authenticates with. In the all-in-one that gateway is a second container rather
than a fourth process under supervisord: keeping the object store inside the
image, on the volume the files are being moved off, would have defeated the
point and would have meant rebuilding.

Files uploaded before the switch are not carried across; PocketBase copies
nothing, and both READMEs say so where an operator will read it.

The TLS overlay and its Caddyfile go. The section they served stays, without
them: nothing in the stack terminates TLS any more, so it now names the four
variables to set in front of whichever proxy already does — TRUST_FORWARDED_PROTO
being the one that decides whether a charger is believed about how it arrived.

Unexercised: this was written on a machine without Docker, so the pinned
SeaweedFS image, the bucket-create and the settings write have not been run
against a live stack. The Go side builds, vets and tests clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Docker/README.md

@@ -78,6 +78,68 @@ instead. PocketBase runs as root, so a root-owned host directory is fine.
 > `WEBAPP_URL` and `CORS_ALLOW_ORIGINS` in `.env` to change them permanently —
 > in this stack the compose environment wins over anything the panel writes.
 
+## File storage (SeaweedFS / S3)
+
+Uploaded files — document scans, service and refill receipts, workshop invoices,
+part photos — live inside `pb_data` by default, next to the database. Two further
+compose files put them in an S3 bucket instead, so the blobs and the database can
+be sized, backed up and moved independently. Nothing else changes: an attachment has
+always been fetched through the API Server (`GET /api/service-records/{id}/file`),
+never from a storage URL, so the Web App, the phone app and the Home Assistant
+plugin cannot tell the difference.
+
+Each shape is one self-contained compose file — nothing to layer, nothing to
+remember — with an `.env` example of the same name:
+
+| Shape | From the registry | From source |
+|---|---|---|
+| **Local storage** — the default, unchanged | `docker-compose.prod.yml` | `docker-compose.yml` |
+| **SeaweedFS in this stack** | `docker-compose.prod.seaweedfs.yml` | `docker-compose.seaweedfs.yml` |
+| **An S3 endpoint outside it** | `docker-compose.prod.s3.yml` | `docker-compose.s3.yml` |
+
+So `docker-compose.prod.seaweedfs.yml` is configured from
+`.env.prod.seaweedfs.example`, `docker-compose.s3.yml` from `.env.s3.example`,
+and so on:
+
+```sh
+cp .env.prod.seaweedfs.example .env    # then edit it — PB_S3_* have no defaults
+docker compose -f docker-compose.prod.seaweedfs.yml pull
+docker compose -f docker-compose.prod.seaweedfs.yml up -d
+```
+
+Set `PB_S3_ACCESS_KEY` and `PB_S3_SECRET` first — both storage files refuse to
+start without them. The SeaweedFS ones add a `seaweedfs` container (master,
+volume, filer and S3 gateway in one process, on its own `seaweed_data` volume)
+plus a one-shot `seaweedfs-init` that creates the bucket, because PocketBase never
+issues a `CreateBucket` of its own. The external-S3 ones add no containers at
+all: set `PB_S3_ENDPOINT`, and create the bucket yourself.
+
+On every boot the API Server's bootstrap writes PocketBase's *Files storage*
+settings from those variables, then asks PocketBase to prove it can reach the
+bucket. Watch for it in the log:
+
+```
+[api] bootstrap: ✓ file storage → S3 (drivervault at http://seaweedfs:8333)
+[api] bootstrap: ✓ S3 storage reachable
+```
+
+A boot that finds the settings already correct logs `• file storage already on S3`
+and writes nothing.
+
+Two things to know before turning it on:
+
+- **Existing files are not migrated.** PocketBase copies nothing when the setting
+  flips, so attachments uploaded before the switch stop resolving. Copy
+  `pb_data/storage/<collectionId>/<recordId>/<file>` into the bucket root, keeping
+  that layout, *before* enabling it — or start from a stack with no attachments.
+- **Going back to the plain compose file is not an off switch.** It leaves
+  PocketBase pointed at
+  the bucket, deliberately: files already written there are reachable only while
+  it is. Move them back and turn it off in PocketBase's own admin UI. For the same
+  reason a rotation of `PB_S3_SECRET` alone is invisible to the bootstrap —
+  PocketBase masks the stored secret on read — so change another `PB_S3_*` value
+  alongside it, or set it in the admin UI.
+
 ## Charger control (OCPP)
 
 Chargers in own/proxy mode dial in to `/ocpp/{serial}`, authenticating with a
@@ -93,34 +155,26 @@ A plaintext `ws://` puts the control token on the wire in the clear, so
 
 ### With a public hostname and TLS
 
-`docker-compose.tls.yml` adds Caddy in front of the stack: one hostname, a
-certificate issued on first boot, and everything behind it spoken to over the
-compose network. Browsers and chargers arrive at the same name.
+Nothing in this stack terminates TLS. Put your own reverse proxy in front of the
+Web App port, give it a certificate and a hostname, and set four things by hand:
 
 ```sh
 # in .env
-DV_DOMAIN=drivervault.example.com
-DV_ACME_EMAIL=you@example.com
-
-docker compose -f docker-compose.prod.yml -f docker-compose.tls.yml up -d
+OCPP_PUBLIC_URL=wss://drivervault.example.com
+OCPP_REQUIRE_TLS=true
+CORS_ALLOW_ORIGINS=https://drivervault.example.com
+TRUST_FORWARDED_PROTO=true
 ```
 
-The overlay sets the rest for you: `OCPP_PUBLIC_URL=wss://$DV_DOMAIN`,
-`OCPP_REQUIRE_TLS=true`, `CORS_ALLOW_ORIGINS=https://$DV_DOMAIN`, and
-`TRUST_FORWARDED_PROTO=true` on the Web App so the BFF passes Caddy's
-`X-Forwarded-Proto` to the API Server instead of overwriting it with its own
-plaintext hop. Point the charger's OCPP backend at the endpoint the panel then
-shows, with the control token as its authorization key.
+`TRUST_FORWARDED_PROTO` is the easy one to miss: without it the Web App's BFF
+overwrites the proxy's `X-Forwarded-Proto` with its own plaintext hop and every
+charger is rejected as insecure. Only set it when that proxy really is the only
+way in — otherwise a charger could claim `wss` over a plaintext connection.
 
-Two things the overlay cannot arrange: `DV_DOMAIN` must resolve to the host from
-the internet with ports 80 and 443 reaching it (Caddy needs `:80` for the ACME
-challenge), and the charger must be able to resolve that name too — behind NAT
-that usually means hairpin NAT or a split-DNS entry pointing it at the LAN
-address.
-
-Using a proxy you already run instead? Terminate TLS there, forward to the Web
-App port, and set the same four variables by hand — the `X-Forwarded-Proto` one
-included, or chargers will be rejected as insecure.
+Point the charger's OCPP backend at the endpoint the panel then shows, with the
+control token as its authorization key. The charger has to resolve that hostname
+too: behind NAT that usually means hairpin NAT, or a split-DNS entry pointing the
+name at the LAN address.
 
 ## Notes