Platform API

Analytics

Use analytics endpoints to inspect usage, latency, errors, token counts, tool calls, and individual request sequences for an organization. These are the same usage analytics available in the console.

All timestamps in analytics query parameters are Unix milliseconds.

Get Analytics Overview

GET /organizations/{org_id}/analytics/overview

Returns summary, timeseries, breakdown, status, and tool-call analytics.

Permission: api-read

Query parameters:

Parameter	Type	Required	Description
`range_start`	integer	Yes	Start timestamp in Unix milliseconds.
`range_end`	integer	Yes	End timestamp in Unix milliseconds.
`group_by`	string	Yes	Breakdown dimension.
`metric`	string	Yes	Breakdown metric.
`buckets`	integer	No	Requested timeseries bucket count. Clamped to `1` through `1000`.
`f`	string	No	Repeated filter.

Metrics:

Metric	Measures
`requests`	Request count.
`errors`	Error count.
`in_tokens`	Input tokens.
`cached_in_tokens`	Cached input tokens.
`out_tokens`	Output tokens.
`duration`	Total sequence duration.
`p50_latency`	Median gateway latency.
`p95_latency`	95th percentile gateway latency.
`error_rate`	Errors divided by requests.

Response:

{
  "data": {
    "range": {
      "start": 1767225600000,
      "end": 1767312000000
    },
    "summary": {
      "requests": 100,
      "errors": 2,
      "in_tokens": 50000,
      "cached_in_tokens": 10000,
      "out_tokens": 25000,
      "duration_ms": 123456,
      "p50_latency_ms": 220,
      "p95_latency_ms": 900,
      "error_rate": 0.02
    },
    "timeseries": {
      "bucket_ms": 3600000,
      "buckets": []
    },
    "breakdown": {
      "dim": "model",
      "metric": "requests",
      "rows": [
        { "key": "qwen/qwen3.5-4b-free", "value": 80, "count": 80 }
      ]
    },
    "status": {
      "rows": [
        { "key": "success", "count": 98 }
      ]
    },
    "tool_calls": {
      "rows": []
    }
  }
}

Each timeseries bucket contains t0, t1, requests, errors, in_tokens, cached_in_tokens, out_tokens, duration_ms, p50_latency_ms, and p95_latency_ms.

List Analytics Sequences

GET /organizations/{org_id}/analytics/sequences

Lists request sequences.

Permission: api-read

Query parameters:

Parameter	Type	Required	Description
`range_start`	integer	Yes	Start timestamp in Unix milliseconds.
`range_end`	integer	Yes	End timestamp in Unix milliseconds.
`f`	string	No	Repeated filter.
`after`	string	No	Cursor from the previous response.
`limit`	integer	No	Defaults to `50`; maximum is `500`.
`include_total`	boolean	No	Include a total row count.

Response:

{
  "data": {
    "sequences": [
      {
        "seq_id": "seq_...",
        "timestamp": 1767225600000,
        "model": "qwen/qwen3.5-4b-free",
        "api_key_id": "ak_...",
        "call_source": "chat_completions",
        "client_ip": "203.0.113.10",
        "client_ua": "curl/8.0",
        "input_tokens": 100,
        "cached_input_tokens": 0,
        "output_tokens": 40,
        "seq_duration_ms": 1200,
        "gateway_latency_ms": 50,
        "result": "success",
        "error_code": null,
        "error_message": null,
        "tool_calls": []
      }
    ],
    "next_cursor": null,
    "total": 1
  }
}

api_key_id, client_ip, client_ua, error_code, and error_message may be null. total is null unless include_total=true.

Get Analytics Sequence

GET /organizations/{org_id}/analytics/sequences/{seq_id}

Returns one request sequence.

Permission: api-read

Query parameters:

Parameter	Type	Required	Description
`range_start`	integer	No	Optional start timestamp in Unix milliseconds.
`range_end`	integer	No	Optional end timestamp in Unix milliseconds.

Response:

{
  "data": {
    "sequence": {
      "seq_id": "seq_...",
      "timestamp": 1767225600000,
      "model": "qwen/qwen3.5-4b-free",
      "api_key_id": "ak_...",
      "call_source": "chat_completions",
      "client_ip": "203.0.113.10",
      "client_ua": "curl/8.0",
      "input_tokens": 100,
      "cached_input_tokens": 0,
      "output_tokens": 40,
      "seq_duration_ms": 1200,
      "gateway_latency_ms": 50,
      "result": "success",
      "error_code": null,
      "error_message": null,
      "tool_calls": []
    }
  }
}

Filters

Analytics endpoints accept repeated f query parameters:

f=dimension:value
f=!dimension:value

The ! prefix negates the filter.

Dimension	Filters by
`model`	Model SKU.
`api_key_id`	API key id.
`call_source`	Request source, such as chat completions.
`client_ip`	Client IP address.
`client_ua`	Client user agent.
`result`	Request result.
`error_code`	Error code for failed requests.
`tool_calls`	Tool name used during the request.

# Shows successful requests for one model, excluding requests from a specific API key.
?f=model:qwen/qwen3.5-4b-free&f=result:success&f=!api_key_id:ak_...

Errors

All endpoints can return:

{
  "error": {
    "message": "resource not found"
  }
}

Common statuses:

Status	Code	When
`400`	none	Missing range, unknown `group_by`, unknown `metric`, invalid filter, invalid integer, or invalid boolean.
`401`	none	Authentication failed.
`403`	`insufficient_api_key_permissions`	API key does not have `api-read`.
`404`	none	Organization, sequence, or API-key organization scope was not found.
`503`	none	Analytics is not configured on the server.
`500`	none	Internal error.

API Key Management Audit Log