Top users by tokens
Top users by token usage.
/analytics/llm-top-users-by-tokensTop users by token usage.
Query Parameters
Number of items to skip before returning results.
int320 <= value0Maximum number of items to return per page.
int320 <= value <= 2010Start of the time range to report on, inclusive.
date-timeEnd of the time range to report on, exclusive.
date-timeFilter to specific LLM telemetry providers.
Filter to specific model names.
Filter to specific users by id.
Filter to users in specific groups.
Response Body
application/json
application/json
curl -X GET "https://example.com/analytics/llm-top-users-by-tokens?start_time=2019-08-24T14%3A15%3A22Z&end_time=2019-08-24T14%3A15%3A22Z"{ "items": [ { "id": "497f6eca-6276-4993-bfeb-53cbbbba6f08", "name": "string", "tokens": 0 } ], "meta": { "total": 0, "offset": 0, "limit": 0 }}Top users by cost GET
Top users by API-equivalent cost (USD list price of the usage) — an estimate, not billed spend.
LLM user percentiles GET
Usage percentile per gateway-attributed user with usage in the range, ranked by total tokens. A percentile of 0.25 or above is what the team adoption views count as a seat in use. Filters are limited to provider and user_group, and each user's groups come back narrowed to the selected ones. Bounded rather than paged: the heaviest 500 users are returned in one response and no page params are accepted. Percentiles are always computed over the whole population, so meta.total is the real headcount and exceeds items.length exactly when the response is a sample.