/health returns HTTP 200 even when the service is DOWN

[nangelovv]

Bug

The /health endpoint returns HTTP 200 even when the health check fails. Liveness/readiness probes (Docker HEALTHCHECK, Kubernetes, load balancers) only inspect the status code, so they treat a downed service as healthy.

Cause

In app/routes/document_routes.py the handler returns a tuple:

@router.get("/health")
async def health_check():
    try:
        if await is_health_ok():
            return {"status": "UP"}
        else:
            logger.error("Health check failed")
            return {"status": "DOWN"}, 503        # Flask idiom
    except Exception as e:
        ...
        return {"status": "DOWN", "error": str(e)}, 503

return body, 503 sets the status code in Flask, but FastAPI does not interpret a returned tuple that way. FastAPI serializes the whole tuple through jsonable_encoder, so the body becomes the JSON array [{"status": "DOWN"}, 503] and the HTTP status stays at the default 200.

Reproduction

docker compose up -d
docker compose stop db   # make the health check fail
curl -s -o - -w '\nHTTP %{http_code}\n' http://localhost:8000/health

Observed:

[{"status":"DOWN"},503]
HTTP 200

Expected: HTTP 503 with body {"status":"DOWN"}.

Impact

With a constant 200, a failed /health never marks the container unhealthy, never triggers a restart, and never removes the instance from a load-balancer pool — the outage is masked.

Fix

Return a JSONResponse with an explicit status_code=503 for both the unhealthy and exception branches. PR incoming.