cesnimda
80fc813b61
Layer a staging overlay (docker-compose.staging.yml) on the base compose file: same Linux images and Production runtime as prod, differing only in the dev banner, capped sync, shifted ports (18080/18081), and an isolated project namespace so it never touches prod data. Wire -Staging into deploy/up.ps1 and deploy/down.ps1, add .env.staging.example, and track that template in git. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
deploy/
Holds the live secrets (deploy/.env) and one-command scripts to build and run the stack against them. deploy/.env is git-ignored — it is never committed.
deploy/.env
Required keys (see ../.env.example for the template):
POSTGRES_PASSWORD=...
GOOGLE_CLIENT_ID=...
GOOGLE_CLIENT_SECRET=...
AI_MODE=Disabled # Disabled | LocalOllama | CloudOpenAi
FRONTEND_ORIGIN=http://localhost:8081
Commands
Windows (PowerShell), from the repo root:
./deploy/up.ps1 # build + start (detached)
./deploy/up.ps1 -Foreground # stream logs instead of detaching
./deploy/up.ps1 -Proxy # also run the top-level nginx (everything on :80)
./deploy/logs.ps1 api # tail a service's logs
./deploy/down.ps1 # stop
./deploy/down.ps1 -Volumes # stop and wipe the DB + key volumes
Linux / Ubuntu production target:
./deploy/up.sh # build + start (detached)
./deploy/up.sh --foreground
./deploy/up.sh --proxy
Both scripts validate that deploy/.env exists and that the required keys are non-blank before invoking Docker, so a misconfigured env fails fast with a clear message instead of a half-started stack.
What it runs
docker compose --env-file deploy/.env up --build — Postgres, the API (auto-applies EF migrations on boot), and the frontend. The --env-file flag feeds the ${...} variables in docker-compose.yml.
After it's up:
- Frontend — http://localhost:8081
- API / Swagger — http://localhost:8080/swagger
Reminder: the Google OAuth redirect URI for this layout is http://localhost:8081/signin-google (see ../docs/GOOGLE_OAUTH_SETUP.md).