Usage analytics and exports

Usage analytics are computed from immutable request, normalized usage, reservation, and posted-ledger facts. Time ranges are half-open: [start, end). The timezone value controls bucket alignment, while timestamps on the wire remain UTC.

Safe request metadata

The request explorer exposes request ID, status, endpoint, model, project, safe key prefix, latency, normalized meters, charge, and finish reason. Prompts, completions, credentials, upstream routes, and internal costs are not stored in this surface.

Long ranges use server-side rollups. The response includes aggregation_version and last_updated_at; never add rounded chart labels to calculate a financial total.

Asynchronous exports

Request CSV or JSON from the authenticated workspace endpoint. Exports are limited to 31 days and 100,000 rows. A ready file is available through an authenticated, expiring download URL. The file contains request metadata and financial customer charges only.

curl "$INFERENCEPASS_API_BASE/v1/chat/completions" \
  -H "Authorization: Bearer $INFERENCEPASS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.6-sol","messages":[{"role":"user","content":"Explain what InferencePass is in one sentence."}],"stream":false,"max_completion_tokens":128,"temperature":0.7}'