api

The FastAPI control surface — every operation MarketService supports, exposed over HTTP, plus a small set of endpoints for observing/poking the bot thread directly.

Files

• app.py — create_app(): wires the FastAPI app, stashes shared dependencies (sessionmaker, aodb, gmi, scraper, bot_handle) on app.state, registers a single exception handler that maps every MarketError subclass to an HTTP status code, and mounts each router in routes/.
• deps.py — FastAPI dependency providers, most importantly get_service() (builds a per-request MarketService, including a ChatSink that’s a no-op unless the bot thread is up) and call_on_bot() — the bridge that lets a request handler run a coroutine on the bot thread’s event loop via asyncio.run_coroutine_threadsafe, then awaits the result from FastAPI’s own loop.
• schemas.py — Pydantic request/response models for every route.
• routes/ — One module per resource:
  • health.py — /healthz (always 200, status reflected in the body) and /readyz (reflects whether the bot thread has connected — stays false forever in API-only mode, so it’s not used for the container’s liveness/readiness probes; see the Helm chart).
  • items.py — item search/lookup, proxied through aodb.
  • watch.py — start/stop watching an item, current order book.
  • users.py — player registration, watchlist subscriptions and per-subscription filters, per-player stats.
  • admin.py — bulk untrack, admin stats lookups.
  • status.py — overall market/watch-list summary.
  • bot.py — bot connection status, poll:trigger/autotrack:trigger to force a cycle on demand, and tell to send a chat message directly.

How it’s used

main.py calls create_app() with real implementations of every dependency and runs it under uvicorn on the FastAPI event loop, while the bot (if configured) runs on its own separate thread — see bot/README.md for the concurrency model that call_on_bot() bridges.