Operations¶

Managing Reflex in production.

💓 Health Monitoring¶

Basic Health Check¶

curl http://localhost:8000/health

{"status": "healthy"}

Detailed Health Check¶

curl http://localhost:8000/health/detailed

{
  "status": "healthy",
  "indicators": [
    {"name": "database", "status": "healthy", "latency_ms": 1.5},
    {"name": "event_queue", "status": "healthy", "message": "42 pending"},
    {"name": "dlq", "status": "healthy", "message": "0 in DLQ"}
  ]
}

📬 Dead-Letter Queue (DLQ)¶

Events that fail after max retries move to the DLQ for manual intervention.

List DLQ Events¶

python scripts/dlq.py list

Retry Events¶

python scripts/dlq.py retry <event-id>

python scripts/dlq.py retry-all

🔄 Event Replay¶

Replay historical events for debugging or reprocessing:

python scripts/replay.py

Use cases:

• Debugging agent behavior with specific events
• Reprocessing events after bug fixes
• Testing new trigger configurations

🔭 Observability¶

Reflex integrates with Logfire for observability.

Automatic Tracing¶

Traces are captured automatically for:

• HTTP requests
• WebSocket connections
• Event store operations
• Agent tool calls

Custom Spans¶

import logfire

with logfire.span("my-operation", key=value):
    # Your code here
    ...

Configuration¶

Set LOGFIRE_TOKEN in your environment:

LOGFIRE_TOKEN=your-token-here

📖 Runbook¶

High DLQ Count¶

Steps:

1. Check recent deployments for bugs
2. Review DLQ events: python scripts/dlq.py list
3. Check external service status
4. Fix root cause before retrying

High Event Latency¶

Steps:

1. Check agent loop logs for slow operations
2. Review Logfire traces for bottlenecks
3. Consider scaling horizontally
4. Check database query performance

Database Connection Exhaustion¶

Steps:

1. Check DB_POOL_MAX vs running instances
2. Verify PostgreSQL max_connections
3. Look for connection leaks (unclosed sessions)
4. Increase pool size or reduce instances