Observability¶
Metrics¶
| Metric | Description |
|---|---|
chat_message_total |
Count of messages sent and received. |
chat_error_total |
Number of failed API calls surfaced to users. |
chat_latency_ms |
End-to-end response time from send to final token. |
SLOs¶
- 99% of chat responses stream within 5s.
- API error rate remains below 1% of requests.
- Evidence retrieval yields at least one document 95% of the time.