Fix rate limiter Redis fallback for healthcheck

[railway-app]

Problem

The /health endpoint returns HTTP 500 on every healthcheck because the slowapi rate limiter middleware attempts to connect to redis://redis:6379 — the hardcoded internal Docker Compose hostname — when no Redis service exists in the deployment. The DNS lookup fails with redis.exceptions.ConnectionError: Error -2 connecting to redis:6379. Name or service not known, which propagates through the middleware stack and crashes the response. The previous guard (if os.getenv("RAILWAY_ENVIRONMENT") and redis_url.startswith("redis://redis:")) only fired when the RAILWAY_ENVIRONMENT variable was present, leaving a gap for Railway deployments where that variable may not be set.

Solution

Updated _rate_limit_storage_uri() in backend/app/middleware/rate_limiting.py to unconditionally return "memory://" whenever the resolved Redis URL starts with redis://redis:, removing the dependency on RAILWAY_ENVIRONMENT being set. Also reordered the env var lookup to check REDIS_URL (Railway's standard variable) before HYPERCODE_REDIS_URL, and fixed _with_redis_db to return "memory://" instead of re-introducing the bare hostname as a fallback. When a real Redis URL is provided via REDIS_URL or HYPERCODE_REDIS_URL, it is used as before; otherwise the limiter runs entirely in-memory, which is correct for single-instance deployments.

Changes

• Modified backend/app/middleware/rate_limiting.py

---

Generated by Railway