Analytics — Observability
Navigate to: Dashboard → Observability
What it shows: Real-time and historical gateway traffic — request volume, token usage, latency percentiles, error rates, and DLP violation counts across all LLM providers and tenants.
Key panels:
| Panel | Description |
|---|---|
| Requests / min | Live request rate across all routes |
| Token Usage | Input and output tokens over time |
| Cost Trend | Spend per hour/day aggregated by provider |
| Error Rate | Failed requests vs. total, broken down by error type |
| DLP Events | Sensitive data detections and blocks per interval |
| Top Models | Most-used models ranked by request count and token consumption |
Step-by-step:
- Click Observability in the sidebar.
- Use the date-range picker (top-right) to set the window — last hour, last 24 h, last 7 days, or custom.
- Hover over any chart point to see the exact value at that timestamp.
- Use the Provider filter to isolate traffic for a single LLM provider.
- Click Export to download the current view as CSV.
How to test: Send a few chat completions via POST /v1/chat/completions and refresh Observability — request count and token usage should increment within 30 seconds.
Cross-reference: See section 4.6 for feature overview. For per-route traffic, use Settings → Gateway Analytics (section 12.19).