fix: read Redis URL from environment variables for Docker deployment (fixes #385)

[octo-patch]

Fixes #385

Problem

The Celery broker and backend URLs were hardcoded to redis://localhost:6379/0 in api.py:

celery_app = Celery("tasks", broker="redis://localhost:6379/0", backend="redis://localhost:6379/0")

When running in Docker, the Redis service is reachable at redis://redis:6379/0 (via the Docker network using the service name), not localhost. The docker-compose.yml already correctly passes REDIS_URL=${REDIS_BASE_URL:-redis://redis:6379/0} as an environment variable to the backend container, but the code was ignoring it entirely.

This causes the backend container to silently fail to connect to Redis during startup, resulting in the symptom reported in #385: the container shows Device set to use cpu and then never reaches the Uvicorn startup logs.

Solution

Read the Redis URL from environment variables before falling back to localhost:

1. REDIS_URL — set by docker-compose.yml for Docker deployments
2. REDIS_BASE_URL — the user-facing variable name documented in .env.example
3. redis://localhost:6379/0 — fallback for local (non-Docker) development

_redis_url = os.environ.get("REDIS_URL", os.environ.get("REDIS_BASE_URL", "redis://localhost:6379/0"))
celery_app = Celery("tasks", broker=_redis_url, backend=_redis_url)

Testing

• Local development: no change in behaviour (falls back to localhost)
• Docker deployment: REDIS_URL=redis://redis:6379/0 is picked up from the environment, matching the existing docker-compose.yml configuration