Observability
Observability
Where to look when you want to know what happened, what it cost, and how the gateway behaved.
Observability
Odock records what your applications do through the gateway and shows it back to you in two places:
- Usage Records — the audit trail of individual requests. One row per call, with status, latency, model, tokens, cost, and the routing trace.
- Traffic Analytics page — the analytics dashboard built on top of those records. Trends, percentiles, and tabs for traffic, latency, usage, MCP, plugins, and traces.
- LGTM stack — the full Grafana, Loki, Tempo, Prometheus, and Alertmanager workspace available when your company self-hosts Odock or runs the enterprise edition on its own infrastructure. See LGTM Stack.
Pick The Right Surface
| You want to know... | Open |
|---|---|
| What did this specific request cost? Why did it fail? | Usage Records. |
| How are requests trending for an API key, team, or model? | Traffic Analytics page. |
| Is one MCP tool or plugin slow? | Traffic Analytics page → MCP or Plugins tab. |
| Did a routing fallback fire and why? | The Routing section on a usage record. |
| Is the gateway itself healthy? | LGTM Stack (platform operator). |
How Data Reaches The UI
Recording happens after the response is returned to the client, so it never blocks your application. A new request can take a few seconds to appear in the table while the gateway finalises and writes it.
What is captured is request shape and accounting — status, latency, token counts, cost, routing attempts, plugin and safety outcomes. Prompts, completions, API keys, and provider credentials are never written into a usage record.
Pages In This Section
- Usage Records: per-request audit trail, analytics card, table, filters, and the record detail page.
- Traffic Analytics page: the organisation dashboard with traffic, latency, usage, MCP, plugins, and traces tabs.
- LGTM Stack: the platform-level monitoring stack used when self-hosting.