doctor

Spree stack diagnosis

Run every check below in order against this project. Do not stop at the first failure — run all of them, then report. Judge each against its healthy state.

Pre-flight — detect the project flavor:

• docker-compose.yml at the root (create-spree-app): run the Docker checks as written. The Rails app lives in server/ (older projects: backend/ — use whichever exists; call it <app> below). apps/dashboard/ and apps/storefront/ are optional.
• No Docker wiring but a Rails app with spree gems at the root (classic app): skip checks 1–2, run everything else natively — replace spree exec <cmd> with plain <cmd> from the app root, check the web on the port bin/dev/bin/rails s uses (default 3000), and read Gemfile.lock at the root.
• Neither: not a Spree project — say so and stop.

Command routing (Docker flavor): prefer spree exec <cmd>; if the spree CLI isn't available, fall back to docker compose exec -T web <cmd>. If .env contains SPREE_PATH, this is a monorepo edge project — spree dev/spree build/spree upgrade refuse there by design; remedies should say pnpm server:dev / pnpm server:build (run from the monorepo root) instead. On classic apps, remedies use native forms (bin/dev, bin/rails db:migrate, bundle exec rake).

Stack shape (create-spree-app): services web (Puma and the Solid Queue job supervisor in-process — there is no separate worker by default), postgres, mailpit. Optional: meilisearch (commented out in the template until enabled), a bin/jobs worker service if the user split jobs out (SOLID_QUEUE_IN_PUMA=false), and redis/sidekiq only if they swapped the job backend. Read docker-compose.yml and <app>/Gemfile.lock first and only run the optional checks for what's actually there.

Checks

1. Docker daemon — docker info --format '{{.ServerVersion}}' Healthy: prints a version. Remedy: start Docker Desktop / OrbStack.

2. Containers — docker compose ps -a --format '{{.Service}}: {{.State}} ({{.Status}})' Healthy: web and postgres running (web healthy once booted), mailpit running. Plus any optional services defined in the compose file (meilisearch, a jobs worker, redis). Exited containers are normal if the user just hasn't booted; the remedy is spree dev (foreground, Ctrl+C stops) — or pnpm server:dev in monorepo mode. Note whether the compose file has a build: section (ejected, builds from <app>/) or uses the prebuilt ghcr.io/spree/spree image — code changes in <app>/ only take effect on an ejected stack.