← all services
chat-agent 5 endpoints
Endpoint-level X-ray over the last 3d — 140 API calls.
Endpoints are derived from the request logs and templatized (:id).
Click one for its Neo4j fan-out, traces, and logs (fetched live).
See the raw calls:
all API calls ·
failing (5xx) →
What it costs:
its bill · namespace tilda · requests vs use · its LLM calls · AI spend by org
error budget
healthy
99.90% derived · 0.0× burn
p95 vs normal
—
no baseline yet
last deploy
—
no ArgoCD sync recorded
Neo4j footprint
145
1.0 per request · 0 slow
owner · on-call
Sambasiva Bandarupalli
backup Ahmed Aburaed · CODEOWNERS /data-science/
| Endpoint | Calls ▾ | Avg latency | Max | Err % |
|---|---|---|---|---|
| GET /api/v1/threads | 114 | 48 ms | 115 ms | 0.0% |
| POST /api/v1/chat/:id/runs/stream | 9 | — | — | 0.0% |
| GET /api/v1/chat/:id/state | 7 | 76 ms | 190 ms | 0.0% |
| POST /api/v1/threads | 7 | 48 ms | 52 ms | 0.0% |
| POST /api/v1/chat/:id/runs/:id/resume | 3 | 14.68 s | 21.77 s | 0.0% |
Per-endpoint Neo4j fan-out (and N+1 detection) is one click away on each endpoint's Queries tab — fetched live by sampling that endpoint's recent traces.