List LLM metrics | Respan Docs

curl -X POST https://api.respan.ai/api/dashboard/llm-metrics/ \
     -H "Authorization: Bearer sk_live_xxxxx" \
     -H "Content-Type: application/json" \
     -d '{
  "end_time": "2026-02-22T23:59:59Z",
  "start_time": "2026-02-01T00:00:00Z"
}'

[
  {
    "date_group": "2026-02-01T00:00:00Z",
    "number_of_requests": 1250,
    "total_cost": 312.75,
    "total_prompt_tokens": 450000,
    "total_completion_tokens": 350000,
    "total_tokens": 800000,
    "error_count": 15,
    "error_percentage": 1.2,
    "average_prompt_tokens": 360,
    "average_completion_tokens": 280,
    "average_tokens": 640,
    "average_cost": 0.25,
    "average_tps": 12.5,
    "average_latency": 350,
    "average_ttft": 120,
    "prompt_cache_hit_tokens": 90000,
    "reasoning_tokens": 150000,
    "cache_hit_percentage": 11.25,
    "requests_per_second": 0.0145
  }
]

Returns LLM usage metrics (requests, tokens, cost, latency, cache hit rate, etc.) bucketed by time_tick (minute / hour / day). Use metrics_to_aggregate to request a subset of metrics.

Request

This endpoint expects an object.

end_timedatetimeRequired

Inclusive end of the range, ISO 8601.

start_timedatetimeRequired

Inclusive start of the range, ISO 8601.

filtersobjectOptional

Filter criteria. See Filters API Reference for operator syntax.

metrics_to_aggregatelist of enumsOptional

Optional. If provided, the response includes only these metric fields. If omitted, all LLM metrics are returned.

time_tickenumOptional

Bucket granularity for time-series endpoints. If omitted, inferred from the range. Has no effect on /summary/ endpoints.

Allowed values:

Response

Successful response.

date_groupdatetime

Bucket start time, ISO 8601.

number_of_requestsinteger

total_costdouble

total_prompt_tokensinteger

total_completion_tokensinteger

total_tokensinteger

error_countinteger

error_percentagedouble

average_prompt_tokensinteger

average_completion_tokensinteger

average_tokensinteger

average_costdouble

average_tpsdouble

average_latencydouble

average_ttftdouble

prompt_cache_hit_tokensinteger

reasoning_tokensinteger

cache_hit_percentagedouble

requests_per_seconddouble

Headers

Request

Response