fix(6244): the instance keeps its own SECRET_KEY in /data, and docker-compose.yml is the compose users deploy from

With no SECRET_KEY set, Grimoire fell back to a fixed placeholder. The
quickstart compose substituted change-me-in-production, so the README's
own update step (pull, up -d) recreated the container on that placeholder,
and every credential sealed under the key the user started with stopped
opening.

- config: with neither SECRET_KEY nor SECRET_KEY_FILE, the first start
  writes a random key to DATA_DIR/secret_key (mode 0600, published with a
  hard link so concurrent starts agree), and later starts read it back.
  An explicit key wins and leaves the file alone. A missing
  SECRET_KEY_FILE or an empty key file stops the start instead of
  switching keys. With no data directory (a run outside the container)
  the process gets a temporary key and says so.
- app: warns on a public placeholder key, on a temporary key, and when
  /data is not a mounted volume.
- Dockerfile creates /data.
- docker-compose.yml is now the deploy compose (published image, both
  volumes, nothing to set). The build-from-source compose is
  docker-compose.dev.yml, and docker-compose.quickstart.yml is gone. The
  Swarm stack file gains the /data volume and keeps its explicit key,
  because a Swarm can move the app away from a node-local volume.
- README: a Deploy section with the compose inline, SECRET_KEY optional,
  "keep both volumes", and updates that keep the key. A test holds the
  README's copy equal to the file.
- Installs that ran on the change-me-in-production placeholder get a key
  of their own on the new compose; Settings then asks once for saved
  forge tokens and the backup secret.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

.env.example

@@ -4,4 +4,6 @@ POSTGRES_USER=fabled
 POSTGRES_PASSWORD=fabled
 POSTGRES_DB=grimoire
 # PGDATA_VOLUME=grimoire_pgdata
-SECRET_KEY=dev-secret-change-me
+# Grimoire keeps its own SECRET_KEY in the /data volume. Set one here only to
+# use a key you manage yourself; it then wins over the kept one.
+# SECRET_KEY=

Dockerfile

@@ -56,5 +56,9 @@ ENV APP_BUILD_KEY=$BUILD_KEY
 ENV APP_CHANNEL=$BUILD_CHANNEL
 ENV APP_COMMIT=$BUILD_COMMIT
 
+# The instance's own SECRET_KEY, the model cache and diagnostics live here.
+# Mount a volume on it so they outlive the container.
+RUN mkdir -p /data
+
 EXPOSE 5000
 CMD ["sh", "-c", "alembic upgrade head && hypercorn 'grimoire.app:create_app()' --bind 0.0.0.0:5000 --keep-alive 600"]

Makefile

@@ -3,22 +3,22 @@
 # --- Docker ---
 
 build:
-	docker compose build
+	docker compose -f docker-compose.dev.yml build
 
 up:
-	docker compose up -d
+	docker compose -f docker-compose.dev.yml up -d
 
 down:
-	docker compose down
+	docker compose -f docker-compose.dev.yml down
 
 logs:
-	docker compose logs -f
+	docker compose -f docker-compose.dev.yml logs -f
 
 health:
 	curl -s http://localhost:5000/api/health
 
 migrate:
-	docker compose exec app alembic upgrade head
+	docker compose -f docker-compose.dev.yml exec app alembic upgrade head
 
 # --- Local dev (requires: uv sync --extra dev and cd frontend && npm ci) ---
 

README.md

@@ -8,17 +8,57 @@
 
 Claude records what it learned, decided and built in Grimoire, and that record reaches the next session, so a solution is recalled instead of reinvented. Notes, tasks, plans, rules, lessons and reusable code live in one place. Claude reaches them through a built-in MCP endpoint and the bundled Claude Code plugin, and people read and edit the same records in the web UI. There is no language model inside Grimoire: Claude is the only assistant.
 
-## Quick start
+## Deploy
 
-You need Docker with Compose. Download [`docker-compose.quickstart.yml`](docker-compose.quickstart.yml), then:
+You need Docker with Compose. Save this as `docker-compose.yml` (it is the [`docker-compose.yml`](docker-compose.yml) in this repository), then run `docker compose up -d` beside it:
 
-```bash
-SECRET_KEY=$(python3 -c 'import secrets; print(secrets.token_hex(32))') \
-  docker compose -f docker-compose.quickstart.yml up -d
+```yaml
+# Fabled Grimoire: the app and its database.
+#   docker compose up -d
+# Then open http://localhost:5000. The first account registered becomes admin.
+#
+# Keep both volumes. app_data holds the instance's own SECRET_KEY, which seals
+# saved credentials; pgdata holds everything else.
+
+services:
+  app:
+    image: git.fabledsword.com/bvandeusen/fabledgrimoire:latest
+    ports:
+      - "5000:5000"
+    environment:
+      DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${POSTGRES_PASSWORD:-grimoire}@db:5432/${POSTGRES_DB:-grimoire}"
+      LOG_LEVEL: "${LOG_LEVEL:-INFO}"
+      # Behind an HTTPS reverse proxy:
+      # BASE_URL: "https://grimoire.example.com"
+      # SECURE_COOKIES: "true"
+      # TRUST_PROXY_HEADERS: "true"
+    volumes:
+      - app_data:/data
+    depends_on:
+      - db
+    restart: unless-stopped
+
+  db:
+    # Postgres 17 with the vector extension Grimoire needs.
+    image: pgvector/pgvector:pg17
+    stop_grace_period: 120s
+    volumes:
+      - pgdata:/var/lib/postgresql/data
+    environment:
+      POSTGRES_USER: ${POSTGRES_USER:-grimoire}
+      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-grimoire}
+      POSTGRES_DB: ${POSTGRES_DB:-grimoire}
+    restart: unless-stopped
+
+volumes:
+  app_data:
+  pgdata:
 ```
 
 Open `http://localhost:5000`. The first account you register becomes the admin.
 
+Nothing has to be set first. On its first start Grimoire makes its own `SECRET_KEY` and keeps it in the `app_data` volume, so restarts, reboots and updates all use the same key.
+
 To connect Claude, create an API key under **Settings → MCP Access**, then in Claude Code:
 
 ```
@@ -30,34 +70,34 @@ The plugin asks for your Grimoire URL and the API key. Settings → MCP Access s
 
 ## Configuration
 
-Set these in the environment or a `.env` file beside the compose file:
+Everything is optional. Set these in the environment or in a `.env` file beside the compose file:
 
-| Variable | Needed | What it does |
-|----------|--------|--------------|
-| `SECRET_KEY` | Yes | Signs sessions and seals stored credentials. Set it once and keep it |
+| Variable | When | What it does |
+|----------|------|--------------|
 | `BASE_URL` | Behind a proxy | The public URL, used in email links and SSO redirects |
 | `SECURE_COOKIES` | Over HTTPS | `true` when Grimoire is served over HTTPS |
 | `TRUST_PROXY_HEADERS` | Behind a proxy | `true` to read the client address from the proxy's headers |
-| `POSTGRES_PASSWORD` | Recommended | The database password (defaults to `grimoire`) |
+| `POSTGRES_PASSWORD` | Before the first start | The database password (defaults to `grimoire`; the database is reachable only inside the compose network). Postgres takes it once, when the volume is created |
+| `SECRET_KEY` | Rarely | A key you manage yourself, instead of the one Grimoire keeps. It wins over the kept key; remove it to go back |
 
 Everything else (registration, email, SSO, git forge connections, retrieval tuning, backups) is in **Settings**, or listed in [Configuration](docs/configuration.md).
 
 ## Setup pitfalls
 
-- **Keep your `SECRET_KEY`.** Forge tokens and the backup bucket's secret are sealed under it. If it changes, Settings asks you to enter those credentials again.
-- **The embedding model downloads on first use**, into the `/data` volume. Without internet access, semantic search falls back to keyword search until the model is in place.
+- **Keep both volumes.** `pgdata` holds everything Grimoire stores. `app_data` holds the instance's key, and forge tokens and the backup bucket's secret are sealed under it. If the key is lost, Settings asks you to enter those credentials again; nothing else is lost.
+- **The embedding model downloads on first use**, into the `app_data` volume. Without internet access, semantic search falls back to keyword search until the model is in place.
 - **Behind a reverse proxy, turn off response buffering** for `/mcp`, because MCP responses stream. [Configuration](docs/configuration.md#behind-a-reverse-proxy) has an Nginx example.
 
 ## Updating and backups
 
 ```bash
-docker compose -f docker-compose.quickstart.yml pull
-docker compose -f docker-compose.quickstart.yml up -d
+docker compose pull
+docker compose up -d
 ```
 
-Database migrations run when the app starts. Claude Code updates the plugin from its marketplace; run `/reload-plugins` after an update.
+Database migrations run when the app starts, and the key stays in its volume. Claude Code updates the plugin from its marketplace; run `/reload-plugins` after an update.
 
-Everything Grimoire stores is in the database. **Settings → Data → Scheduled backups** writes a daily backup to a folder or an S3-compatible bucket, and `python -m grimoire.restore` brings a fresh instance back from one. See [Backups](docs/backups.md).
+**Settings → Data → Scheduled backups** writes a daily backup of the database to a folder or an S3-compatible bucket, and `python -m grimoire.restore` brings a fresh instance back from one. See [Backups](docs/backups.md).
 
 ## Features
 
@@ -92,7 +132,8 @@ The full list is in [Features](docs/features.md).
 
 | What you see | Why | Fix |
 |--------------|-----|-----|
-| `SECRET_KEY is set to the insecure default` at startup | `SECURE_COOKIES=true` while `SECRET_KEY` is still the built-in default | Set `SECRET_KEY` |
+| `SECRET_KEY is set to the public placeholder` at startup | `SECURE_COOKIES=true` while `SECRET_KEY` is `dev-secret-change-me` | Remove `SECRET_KEY`, or set a real one |
+| Settings asks for forge tokens or the backup secret again | The instance's key changed: the `app_data` volume was replaced, or `SECRET_KEY` was set or removed | Enter them again, or restore the previous key |
 | `Can't locate revision identified by '…'` at startup | An older image is running against a database a newer one has already migrated | Run the newer image; migrations only move forward |
 | The plugin's tools are missing in Claude Code | The plugin isn't installed, or Claude Code hasn't reloaded it | `/plugin` to install or update it, then `/reload-plugins` |
 | Search finds only exact words | The embedding model hasn't downloaded yet | Give the app internet access once, or wait for the first download to finish |
@@ -100,8 +141,7 @@ The full list is in [Features](docs/features.md).
 ## Development
 
 ```bash
-cp .env.example .env   # then set SECRET_KEY
-docker compose up --build
+docker compose -f docker-compose.dev.yml up --build
 ```
 
 [Development](docs/development.md) covers checks, migrations, the plugin and CI. [Architecture](docs/architecture.md) describes how it is built, and the [API reference](docs/api-reference.md) lists every REST endpoint.

docker-compose.dev.yml

@@ -1,36 +1,35 @@
-# Fabled Grimoire from the pre-built image:
-#   docker compose -f docker-compose.quickstart.yml up -d
-# Then open http://localhost:5000 (the first account registered becomes admin)
-# and connect Claude from Settings → MCP Access.
+# Grimoire built from this checkout, for development:
+#   docker compose -f docker-compose.dev.yml up --build
 
 services:
   app:
-    image: git.fabledsword.com/bvandeusen/fabledgrimoire:latest
+    build: .
     ports:
       - "5000:5000"
-    environment:
-      DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${POSTGRES_PASSWORD:-grimoire}@db:5432/${POSTGRES_DB:-grimoire}"
-      # Set SECRET_KEY in the environment or a .env file beside this one.
-      SECRET_KEY: "${SECRET_KEY:-change-me-in-production}"
-      LOG_LEVEL: "${LOG_LEVEL:-INFO}"
-    volumes:
-      - app_data:/data
     depends_on:
       - db
     restart: unless-stopped
+    volumes:
+      - app_data:/data
+      # Or a bind mount, for direct host access to the app's data:
+      # - ./data:/data
+    environment:
+      DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${POSTGRES_PASSWORD:-grimoire}@db:5432/${POSTGRES_DB:-grimoire}"
+      # Optional: a SearXNG instance to list as the web-search backend in Integrations.
+      # SEARXNG_URL: "http://searxng:8080"
 
   db:
     # Postgres 17 with the vector extension Grimoire needs.
     image: pgvector/pgvector:pg17
     stop_grace_period: 120s
+    restart: unless-stopped
     volumes:
       - pgdata:/var/lib/postgresql/data
     environment:
       POSTGRES_USER: ${POSTGRES_USER:-grimoire}
       POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-grimoire}
       POSTGRES_DB: ${POSTGRES_DB:-grimoire}
-    restart: unless-stopped
 
 volumes:
-  app_data:
   pgdata:
+  app_data:

docker-compose.prod.yml

@@ -3,10 +3,15 @@ services:
     image: git.fabledsword.com/bvandeusen/fabledgrimoire:latest
     environment:
       DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${DB_PASSWORD}@db:5432/${POSTGRES_DB:-grimoire}"
+      # A Swarm can start the app on another node, away from the key kept in
+      # a node-local /data volume, so a stack sets its key itself (or mounts
+      # a Docker secret and sets SECRET_KEY_FILE to its path).
       SECRET_KEY: "${SECRET_KEY}"
       LOG_LEVEL: "${LOG_LEVEL:-INFO}"
       TRUST_PROXY_HEADERS: "true"
       SECURE_COOKIES: "true"
+    volumes:
+      - app_data:/data
     networks:
       - backend
 
@@ -26,6 +31,7 @@ services:
 volumes:
   pgdata:
     name: ${PGDATA_VOLUME:-grimoire_pgdata}
+  app_data:
 
 networks:
   backend:

docker-compose.yml

@@ -1,33 +1,40 @@
+# Fabled Grimoire: the app and its database.
+#   docker compose up -d
+# Then open http://localhost:5000. The first account registered becomes admin.
+#
+# Keep both volumes. app_data holds the instance's own SECRET_KEY, which seals
+# saved credentials; pgdata holds everything else.
+
 services:
   app:
-    build: .
+    image: git.fabledsword.com/bvandeusen/fabledgrimoire:latest
     ports:
       - "5000:5000"
+    environment:
+      DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${POSTGRES_PASSWORD:-grimoire}@db:5432/${POSTGRES_DB:-grimoire}"
+      LOG_LEVEL: "${LOG_LEVEL:-INFO}"
+      # Behind an HTTPS reverse proxy:
+      # BASE_URL: "https://grimoire.example.com"
+      # SECURE_COOKIES: "true"
+      # TRUST_PROXY_HEADERS: "true"
+    volumes:
+      - app_data:/data
     depends_on:
       - db
     restart: unless-stopped
-    volumes:
-      - app_data:/data
-      # Or a bind mount, for direct host access to the app's data:
-      # - ./data:/data
-    environment:
-      DATABASE_URL: "postgresql+asyncpg://${POSTGRES_USER:-grimoire}:${POSTGRES_PASSWORD:-grimoire}@db:5432/${POSTGRES_DB:-grimoire}"
-      SECRET_KEY: "${SECRET_KEY:-dev-secret-change-me}"
-      # Optional: a SearXNG instance to list as the web-search backend in Integrations.
-      # SEARXNG_URL: "http://searxng:8080"
 
   db:
     # Postgres 17 with the vector extension Grimoire needs.
     image: pgvector/pgvector:pg17
     stop_grace_period: 120s
-    restart: unless-stopped
     volumes:
       - pgdata:/var/lib/postgresql/data
     environment:
       POSTGRES_USER: ${POSTGRES_USER:-grimoire}
       POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-grimoire}
       POSTGRES_DB: ${POSTGRES_DB:-grimoire}
+    restart: unless-stopped
 
 volumes:
-  pgdata:
   app_data:
+  pgdata:

docs/architecture.md

@@ -74,7 +74,7 @@ Every database wait on startup has a deadline. Migrations wait for the database
 - **OIDC** single sign-on with PKCE, optionally replacing local logins ([SSO / OAuth](sso-oauth.md)).
 - **Rate limits** on register, login and forgot-password, per client IP (`TRUST_PROXY_HEADERS` reads the IP from the proxy's headers).
 
-Credentials Grimoire stores for outside services (forge tokens, the backup destination's S3 secret) are sealed under a key derived from `SECRET_KEY`.
+Credentials Grimoire stores for outside services (forge tokens, the backup destination's S3 secret) are sealed under a key derived from `SECRET_KEY`, which the instance makes on first start and keeps in `/data/secret_key` unless one is set.
 
 ## Repository layout
 

docs/backups.md

@@ -31,7 +31,8 @@ differently. It does **not** hold:
 - **Credentials:** API keys and forge connections. After a restore, issue new
   API keys and add forge connections again. (Forge tokens are also sealed under
   `SECRET_KEY` in the database itself, so a copy of the database doesn't carry
-  them in readable form either.)
+  them in readable form either. The key lives in `/data/secret_key`, outside
+  the database.)
 - **Logs and notifications.**
 - **The backup destination's own S3 secret.**
 

docs/configuration.md

@@ -11,7 +11,8 @@ Any variable marked *secret* can also be read from a file: set `<NAME>_FILE` to
 | Variable | Default | Description |
 |----------|---------|-------------|
 | `DATABASE_URL` *(secret)* | `postgresql+asyncpg://grimoire:grimoire@localhost:5432/grimoire` | PostgreSQL connection string, using the `asyncpg` driver |
-| `SECRET_KEY` *(secret)* | `dev-secret-change-me` | Signs sessions and seals stored credentials. Set it before first use and keep it (see below) |
+| `SECRET_KEY` *(secret)* | kept in `DATA_DIR` | Signs sessions and seals stored credentials. Unset, the instance makes its own on first start and keeps it in `DATA_DIR/secret_key`. Set it only to manage the key yourself; it wins over the kept one (see below) |
+| `DATA_DIR` | `/data` | Where the instance keeps its own `SECRET_KEY`. Mount a volume here |
 | `BASE_URL` | `http://localhost:5000` | Public URL, used in email links and OIDC redirects |
 | `SECURE_COOKIES` | `false` | `true` when Grimoire is served over HTTPS |
 | `TRUST_PROXY_HEADERS` | `false` | `true` behind a reverse proxy, to read the client IP from `X-Forwarded-For` / `X-Real-IP` |
@@ -56,27 +57,27 @@ Admins can set SMTP in **Settings**; these are the defaults until they do.
 
 ## Running it
 
-### Quick start
+### Docker Compose
 
-[`docker-compose.quickstart.yml`](../docker-compose.quickstart.yml) runs the published image with its own database:
+[`docker-compose.yml`](../docker-compose.yml) runs the published image with its own database, and needs nothing set:
 
 ```bash
-SECRET_KEY=$(python3 -c 'import secrets; print(secrets.token_hex(32))') \
-  docker compose -f docker-compose.quickstart.yml up -d
+docker compose up -d
 ```
 
 Open `http://localhost:5000`. The first account registered becomes the admin; after that, registration stays closed until an admin opens it in Settings or sends an invite.
 
+Update with `docker compose pull && docker compose up -d`. The key and the data stay in their volumes.
+
 ### From source
 
 ```bash
-cp .env.example .env   # then set SECRET_KEY
-docker compose up --build
+docker compose -f docker-compose.dev.yml up --build
 ```
 
 ### Docker Swarm
 
-[`docker-compose.prod.yml`](../docker-compose.prod.yml) is a stack file: the app and database on a private overlay network, with `SECURE_COOKIES` and `TRUST_PROXY_HEADERS` on. Provide `DB_PASSWORD` and `SECRET_KEY` in the environment:
+[`docker-compose.prod.yml`](../docker-compose.prod.yml) is a stack file: the app and database on a private overlay network, with `SECURE_COOKIES` and `TRUST_PROXY_HEADERS` on. A Swarm can move the app to another node, away from a node-local `/data` volume, so the stack sets its own key: provide `DB_PASSWORD` and `SECRET_KEY` in the environment (or mount a Docker secret and set `SECRET_KEY_FILE`):
 
 ```bash
 DB_PASSWORD=... SECRET_KEY=... docker stack deploy -c docker-compose.prod.yml grimoire
@@ -89,9 +90,9 @@ The shipped files leave out healthchecks and restart policies on purpose. The ap
 ### Volumes
 
 - **Database** (`pgdata`): everything Grimoire stores.
-- **App data** (`/data`): the embedding model cache and diagnostics. Losing it costs a re-download, nothing else.
+- **App data** (`/data`): the instance's own `SECRET_KEY`, the embedding model cache and diagnostics. Losing it costs the key: Settings then asks for saved forge tokens and the backup secret again, and the model downloads again.
 
-Back up the database with **Settings → Data → Scheduled backups** ([Backups](backups.md)).
+Back up the database with **Settings → Data → Scheduled backups** ([Backups](backups.md)). A database backup doesn't carry the key; to move sealed credentials to a new instance, copy `/data/secret_key` with it.
 
 ## Behind a reverse proxy
 
@@ -115,7 +116,7 @@ With `TRUST_PROXY_HEADERS=true`, make sure the proxy replaces any `X-Forwarded-F
 
 ## Security checklist
 
-- **Set a strong `SECRET_KEY`, and keep it.** Forge tokens and the backup destination's S3 secret are sealed under a key derived from it, so a copy of the database alone can't read them. If the key changes (say, a restore onto an instance with a different key), Settings marks those credentials and asks for them again; nothing else breaks. The app logs a warning at startup while the default key is in use, and refuses to start with it when `SECURE_COOKIES=true`.
+- **Keep the key.** Forge tokens and the backup destination's S3 secret are sealed under a key derived from `SECRET_KEY`, so a copy of the database alone can't read them. Unset, the instance makes a random key and keeps it in `/data/secret_key`, readable by its owner only; keep that volume. If the key changes (say, a restore onto an instance with a different key), Settings marks those credentials and asks for them again; nothing else breaks. The app warns at startup when `SECRET_KEY` is a public placeholder or `/data` is not a mounted volume, and refuses to start with `dev-secret-change-me` when `SECURE_COOKIES=true`.
 - **`SECURE_COOKIES=true`** whenever Grimoire is served over HTTPS.
 - **Registration** is open only for the first account. Invite people, or open it briefly from Settings.
 - **Signing everyone else out**: changing a password ends that user's other sessions; Settings → Account has a button for it too.

docs/development.md

@@ -3,9 +3,9 @@
 ## Running it
 
 ```bash
-docker compose up --build          # app + PostgreSQL at http://localhost:5000
-docker compose up --build app      # rebuild after a change
-docker compose down -v             # stop and wipe the database
+docker compose -f docker-compose.dev.yml up --build       # app + PostgreSQL at http://localhost:5000
+docker compose -f docker-compose.dev.yml up --build app   # rebuild after a change
+docker compose -f docker-compose.dev.yml down -v          # stop and wipe the database
 ```
 
 The image builds the frontend, so a frontend change needs a rebuild too. For quicker frontend work, run Vite against the running app; it proxies `/api` to port 5000:
@@ -32,7 +32,7 @@ Unit tests run without a database. Tests marked `integration` need a real Postgr
 The container runs `alembic upgrade head` on every start, after waiting for the database to accept connections.
 
 ```bash
-docker compose exec app alembic revision -m "short_description"
+docker compose -f docker-compose.dev.yml exec app alembic revision -m "short_description"
 ```
 
 - Number revisions sequentially (`0123_...`). Revision ids must fit in 32 characters, the width of `alembic_version.version_num`.

src/grimoire/app.py

@@ -1,11 +1,12 @@
 import logging
+import os
 import time
 import traceback as tb_module
 from pathlib import Path
 
 from quart import Quart, g, jsonify, make_response, request, send_from_directory
 
-from grimoire.config import Config
+from grimoire.config import PLACEHOLDER_SECRET_KEYS, SECRET_KEY_FILENAME, Config
 from grimoire.routes.admin import admin_bp
 from grimoire.routes.api import api
 from grimoire.routes.auth import auth_bp
@@ -62,10 +63,23 @@ def create_app() -> Quart:
     app.config["SESSION_COOKIE_SAMESITE"] = "Lax"
     app.config["SESSION_COOKIE_SECURE"] = Config.SECURE_COOKIES
 
-    if Config.SECRET_KEY == "dev-secret-change-me":
+    if Config.SECRET_KEY in PLACEHOLDER_SECRET_KEYS:
         logger.warning(
-            "SECRET_KEY is set to the default value — session cookies are insecure. "
-            "Set SECRET_KEY or SECRET_KEY_FILE for production use."
+            "SECRET_KEY is set to a public placeholder, so session cookies are insecure. "
+            "Remove it to let the instance keep its own key in %s, or set a real one.",
+            Config.DATA_DIR,
+        )
+    if Config.SECRET_KEY_SOURCE == "ephemeral":
+        logger.warning(
+            "No SECRET_KEY is set and %s does not exist, so this process is using a "
+            "temporary key. Sessions and saved credentials will not survive a restart.",
+            Config.DATA_DIR,
+        )
+    elif Config.SECRET_KEY_SOURCE == "data" and not os.path.ismount(Config.DATA_DIR):
+        logger.warning(
+            "%s is not a mounted volume. The instance's key (%s) and the model cache "
+            "are lost when the container is recreated; mount a volume there.",
+            Config.DATA_DIR, SECRET_KEY_FILENAME,
         )
 
     if not Config.TRUST_PROXY_HEADERS:

src/grimoire/config.py

@@ -1,4 +1,13 @@
 import os
+import secrets
+from pathlib import Path
+
+# The file in the data directory that holds the instance's own SECRET_KEY.
+SECRET_KEY_FILENAME = "secret_key"
+
+# Placeholder keys that sample files and older compose files have shipped. A
+# key anyone can read in this repository signs nothing safely.
+PLACEHOLDER_SECRET_KEYS = frozenset({"dev-secret-change-me", "change-me-in-production"})
 
 
 def _read_secret(env_var: str, secret_file_var: str, default: str) -> str:
@@ -16,13 +25,76 @@ def _read_secret(env_var: str, secret_file_var: str, default: str) -> str:
     return default
 
 
+def load_or_create_secret_key(data_dir: Path) -> str | None:
+    """The instance's own SECRET_KEY, kept in `data_dir` and made on first use.
+
+    The first call writes a random key to `<data_dir>/secret_key`, readable by
+    its owner only; every later call, from this process or another, reads that
+    same key back. The file is published with a hard link, so two processes
+    starting together agree on one key and a crash never leaves an empty file.
+    Answers None when `data_dir` doesn't exist, which is a run outside the
+    container. An empty key file raises rather than being replaced, because a
+    new key would orphan everything sealed under the old one.
+    """
+    if not data_dir.is_dir():
+        return None
+    path = data_dir / SECRET_KEY_FILENAME
+    if not path.exists():
+        tmp = data_dir / f".{SECRET_KEY_FILENAME}.{os.getpid()}"
+        fd = os.open(tmp, os.O_WRONLY | os.O_CREAT | os.O_TRUNC, 0o600)
+        try:
+            with os.fdopen(fd, "w") as f:
+                f.write(secrets.token_hex(32) + "\n")
+                f.flush()
+                os.fsync(f.fileno())
+            try:
+                os.link(tmp, path)
+            except FileExistsError:
+                pass
+        finally:
+            tmp.unlink(missing_ok=True)
+    key = path.read_text().strip()
+    if not key:
+        raise RuntimeError(
+            f"{path} is empty. Restore it from a copy, or delete it to start "
+            "with a new key (saved forge tokens and the backup secret are then "
+            "entered again)."
+        )
+    return key
+
+
+def _secret_key() -> tuple[str, str]:
+    """SECRET_KEY and where it came from: "env", "data" or "ephemeral".
+
+    An explicit key (SECRET_KEY or SECRET_KEY_FILE) wins and leaves the data
+    directory's key file alone, so unsetting it returns to that key. With no
+    explicit key the instance keeps its own in DATA_DIR. With neither, the
+    process gets a random key of its own that dies with it.
+    """
+    explicit = _read_secret("SECRET_KEY", "SECRET_KEY_FILE", "")
+    if explicit:
+        return explicit, "env"
+    if os.environ.get("SECRET_KEY_FILE"):
+        # Falling through to another key would orphan what the named one sealed.
+        raise RuntimeError(
+            f"SECRET_KEY_FILE={os.environ['SECRET_KEY_FILE']} is missing or empty."
+        )
+    kept = load_or_create_secret_key(Path(os.environ.get("DATA_DIR", "/data")))
+    if kept:
+        return kept, "data"
+    return secrets.token_hex(32), "ephemeral"
+
+
 class Config:
     DATABASE_URL: str = _read_secret(
         "DATABASE_URL",
         "DATABASE_URL_FILE",
         "postgresql+asyncpg://grimoire:grimoire@localhost:5432/grimoire",
     )
-    SECRET_KEY: str = _read_secret("SECRET_KEY", "SECRET_KEY_FILE", "dev-secret-change-me")
+    # Where the instance keeps its own SECRET_KEY. /data is a volume in every
+    # shipped compose file, beside the model cache and diagnostics.
+    DATA_DIR: str = os.environ.get("DATA_DIR", "/data")
+    SECRET_KEY, SECRET_KEY_SOURCE = _secret_key()
     SECURE_COOKIES: bool = os.environ.get("SECURE_COOKIES", "").lower() in ("1", "true", "yes")
     LOG_LEVEL: str = os.environ.get("LOG_LEVEL", "INFO")
     LOG_RETENTION_DAYS: int = int(os.environ.get("LOG_RETENTION_DAYS", "90"))
@@ -93,8 +165,9 @@ class Config:
             errors.append(f"BASE_URL='{cls.BASE_URL}' must start with http:// or https:// when OIDC is enabled")
         if cls.SECRET_KEY == "dev-secret-change-me" and cls.SECURE_COOKIES:
             errors.append(
-                "SECRET_KEY is set to the insecure default but SECURE_COOKIES=true indicates "
-                "a production deployment. Set SECRET_KEY or SECRET_KEY_FILE before starting."
+                "SECRET_KEY is set to the public placeholder 'dev-secret-change-me' but "
+                "SECURE_COOKIES=true indicates a production deployment. Remove SECRET_KEY "
+                "to let the instance keep its own key in DATA_DIR, or set a real one."
             )
         if errors:
             raise ValueError("Configuration errors:\n" + "\n".join(f"  - {e}" for e in errors))

tests/test_config_secret_key.py

@@ -0,0 +1,83 @@
+"""The instance keeps its own SECRET_KEY in the data directory.
+
+A deployment that never sets SECRET_KEY must get the same key on every start,
+or everything sealed under the first one (forge tokens, the backup secret)
+stops opening after the first container recreate.
+"""
+import stat
+from unittest.mock import patch
+
+import pytest
+
+from grimoire import config
+from grimoire.config import SECRET_KEY_FILENAME, load_or_create_secret_key
+from grimoire.services import secrets
+
+
+def test_the_first_start_writes_a_key_only_its_owner_can_read(tmp_path):
+    key = load_or_create_secret_key(tmp_path)
+
+    path = tmp_path / SECRET_KEY_FILENAME
+    assert key and len(key) == 64
+    assert path.read_text().strip() == key
+    assert stat.S_IMODE(path.stat().st_mode) == 0o600
+    assert [p.name for p in tmp_path.iterdir()] == [SECRET_KEY_FILENAME]
+
+
+def test_every_later_start_reads_the_same_key_and_what_it_sealed_still_opens(tmp_path):
+    first = load_or_create_secret_key(tmp_path)
+    with patch.object(secrets.Config, "SECRET_KEY", first):
+        sealed = secrets.seal("forge-token")
+
+    second = load_or_create_secret_key(tmp_path)
+
+    assert second == first
+    with patch.object(secrets.Config, "SECRET_KEY", second):
+        assert secrets.unseal(sealed) == "forge-token"
+
+
+def test_no_data_directory_means_no_kept_key(tmp_path):
+    assert load_or_create_secret_key(tmp_path / "absent") is None
+
+
+def test_an_empty_key_file_is_refused_rather_than_replaced(tmp_path):
+    (tmp_path / SECRET_KEY_FILENAME).write_text("\n")
+
+    with pytest.raises(RuntimeError, match="is empty"):
+        load_or_create_secret_key(tmp_path)
+    assert (tmp_path / SECRET_KEY_FILENAME).read_text() == "\n"
+
+
+def test_an_explicit_key_wins_and_leaves_the_kept_one_alone(tmp_path, monkeypatch):
+    monkeypatch.setenv("DATA_DIR", str(tmp_path))
+    monkeypatch.delenv("SECRET_KEY_FILE", raising=False)
+    monkeypatch.delenv("SECRET_KEY", raising=False)
+    kept, source = config._secret_key()
+    assert source == "data"
+
+    monkeypatch.setenv("SECRET_KEY", "an-operator-key")
+    assert config._secret_key() == ("an-operator-key", "env")
+
+    monkeypatch.delenv("SECRET_KEY")
+    assert config._secret_key() == (kept, "data")
+
+
+def test_a_secret_key_file_that_is_missing_stops_the_start(tmp_path, monkeypatch):
+    monkeypatch.setenv("DATA_DIR", str(tmp_path))
+    monkeypatch.delenv("SECRET_KEY", raising=False)
+    monkeypatch.setenv("SECRET_KEY_FILE", str(tmp_path / "no-such-secret"))
+
+    with pytest.raises(RuntimeError, match="SECRET_KEY_FILE"):
+        config._secret_key()
+    assert not (tmp_path / SECRET_KEY_FILENAME).exists()
+
+
+def test_with_nowhere_to_keep_a_key_the_process_gets_a_temporary_one(tmp_path, monkeypatch):
+    monkeypatch.setenv("DATA_DIR", str(tmp_path / "absent"))
+    monkeypatch.delenv("SECRET_KEY", raising=False)
+    monkeypatch.delenv("SECRET_KEY_FILE", raising=False)
+
+    key, source = config._secret_key()
+
+    assert source == "ephemeral"
+    assert key not in config.PLACEHOLDER_SECRET_KEYS and len(key) == 64

tests/test_readme_compose.py

@@ -0,0 +1,24 @@
+"""The compose file the README shows is the one users deploy from.
+
+The README carries docker-compose.yml inline, so a newcomer can copy it from
+the page. Both copies have to say the same thing.
+"""
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parent.parent
+
+
+def test_the_readme_shows_the_shipped_compose_file_verbatim():
+    compose = (ROOT / "docker-compose.yml").read_text()
+    readme = (ROOT / "README.md").read_text()
+
+    assert f"```yaml\n{compose}```" in readme
+
+
+def test_the_shipped_compose_needs_no_key_and_keeps_its_data():
+    compose = (ROOT / "docker-compose.yml").read_text()
+
+    settings = [line for line in compose.splitlines() if not line.lstrip().startswith("#")]
+    assert not [line for line in settings if "SECRET_KEY" in line]
+    assert "app_data:/data" in compose
+    assert "pgdata:/var/lib/postgresql/data" in compose