Skip to main content

Ingest AI Token Usage

POST 
/api/v1/stats/ai

Ingest AI token consumption records from the Chat Proxy. Records are batched per-user and flushed periodically (every 60 seconds or 100 records, whichever comes first).

All records in a single request must belong to the same user. The user_id in the records must match the authenticated user from the JWT.

Request Body​

{
  "records": [
    {
      "user_id": 42,
      "usergroup_id": 100,
      "input_tokens": 1240,
      "output_tokens": 380,
      "model": "claude-sonnet-4-20250514",
      "tool_calls": 2,
      "timestamp": "2026-04-04T10:00:00Z"
    }
  ]
}

Validation​

  • The JWT must be valid and the user_id in the records must match the authenticated user. Returns 403 if mismatched.
  • If usergroup_id doesn't match the JWT, the backend uses the group from the JWT (authoritative source).
  • A batch may contain between 1 and 100 records.

Request​

Responses​

Records accepted