judges

Get judge stats

Average scores over completed evaluations, per-criterion averages over the 100 most recent ones, the failure count, and token usage and cost across all of the judge's evaluations.

Required permission: judges.view.

get/v1/projects/{project_id}/judges/{judge_id}/stats

Request

  • Base URL: https://api.connic.co
  • URL: https://api.connic.co/v1/projects/{project_id}/judges/{judge_id}/stats
  • Auth: HTTP bearer

Path parameters

project_idstring uuid required

ID of the project the API key belongs to.

judge_idstring uuid required

Judge ID from GET /judges.

Query parameters

convert_costs_to'EUR' | 'USD' nullable

Convert token costs to EUR or USD at the daily ECB reference rate of each model call

Response

Successful Response

avg_scorenumber nullable

Average overall score of completed evaluations; null when none have completed.

avg_max_scorenumber nullable

Average overall maximum score of completed evaluations; null when none have completed.

total_evaluatedinteger

Number of completed evaluations.

total_failedinteger

Number of failed evaluations.

total_tokensinteger

Input, output and thinking tokens across all of the judge's evaluations.

total_input_tokensinteger

Input tokens across all evaluations, including cached ones.

total_output_tokensinteger

Output tokens across all evaluations.

total_thinking_tokensinteger

Reasoning tokens across all evaluations.

total_cached_input_tokensinteger

Input tokens served from the provider's cache; part of total_input_tokens.

total_audio_input_tokensinteger

Audio part of total_input_tokens.

total_audio_output_tokensinteger

Audio part of total_output_tokens.

total_audio_cached_input_tokensinteger

Audio part of total_cached_input_tokens.

total_audio_thinking_tokensinteger

Audio part of total_thinking_tokens.

total_audio_usage_missingboolean

True when at least one model call did not report its audio token split, so audio totals may be low.

avg_tokens_per_evalnumber

Average total_tokens per evaluation with recorded usage, rounded to one decimal; 0 when there is none.

Changes

No recorded changes to this operation across all 2 revisions of this API.