# Observability

## Logging

**Library**: Loguru 0.7.3

**Sinks**:

| Sink   | Level   | Filter                             | Destination                          |
|--------|---------|-------------------------------------|--------------------------------------|
| File   | INFO+   | All                                 | `Logs/log_loader_{YYYYMMDD}.txt`     |
| Stdout | DEBUG   | INFO, DEBUG, SUCCESS only           | Container stdout                     |
| Stderr | WARNING+| All                                 | Container stderr                     |

**Format**: `[HH:mm:ss LEVEL] message`

**Rotation**: Daily (1 day), 30-day retention (file sink only)

**Async**: File sink uses `enqueue=True` for non-blocking writes

## Health Checks

| Endpoint     | Method | Response           | Purpose         |
|-------------|--------|--------------------|------------------|
| `/health`   | GET    | `{"status": "healthy"}` | Liveness probe  |
| `/status`   | GET    | `{status, authenticated, modelCacheDir}` | Readiness/info |

## Metrics

No metrics collection (Prometheus, StatsD, etc.) is implemented.

## Tracing

No distributed tracing is implemented.

## Gaps

- No structured logging (JSON format) — plain text only
- No request-level logging middleware (request ID, duration, status code)
- No metrics endpoint
- No distributed tracing
- Log directory `Logs/` is hardcoded — not configurable via environment