Skip to main content
GET
cURL

Authorizations

X-CEKURA-API-KEY
string
header
required

API Key Authentication. It should be included in the header of each request.

Query Parameters

progress_id
string
required

Progress ID returned from the process_feedbacks endpoint

agent_id
integer

Agent ID for permission validation

project_id
integer

Project ID for permission validation

Response

200 - application/json
improved_metric_description
string
required

Improved metric description

improved_evaluation_trigger
string
required

Improved evaluation trigger. Interpretation depends on improved_trigger_kind: 'always' (literal) → run on every example; Python source → trigger.py from the sandbox (when improved_trigger_kind='custom_code'); LLM-judge prompt otherwise.

improved_trigger_kind
enum<string>
default:always

Tells the frontend how to interpret improved_evaluation_trigger. Set to 'custom_code' when meta-harness produced an optimized Python trigger; 'always' when no trigger gating is needed.

  • always - Always
  • llm_judge - LLM Judge
  • custom_code - Custom Code
Available options:
always,
llm_judge,
custom_code
trigger_type
enum<string>

Type of evaluation trigger (llm_judge or custom_code)

  • llm_judge - LLM Judge
  • custom_code - Custom Code
Available options:
llm_judge,
custom_code
metric_description_program
any

Advanced metric configuration

metric_description_variables
any[]

Variables used in metric description

evaluation_trigger_program
any

Advanced trigger configuration

evaluation_trigger_variables
any[]

Variables used in evaluation trigger

review_results
object[]

Evaluation results for each test set

custom_code
string
default:""

Optimized Python code (meta-harness output). When present, metric should switch to custom_code type.

type
enum<string>

Suggested metric type after optimization (custom_code if meta-harness produced code)

  • basic - Basic (Deprecated in favor of LLM Judge)
  • custom_prompt - Custom Prompt ( Deprecated in favor of LLM Judge)
  • custom_code - Custom Code
  • llm_judge - LLM Judge
Available options:
basic,
custom_prompt,
custom_code,
llm_judge
feedback_edits
any[]

User-answer-driven feedback corrections applied to metric reviews during optimization

unauthorized_label_changes
any[]

Expected-value changes detected in the run that lacked explicit user approval

unresolved_samples
any[]

Samples the optimizer could not resolve, each with a verdict (noise / ambiguous label / prompt limit), the observed pass rate across repeated evaluations, and a user-facing explanation of why 100% accuracy was not reachable for it.

optimization_summary
string
default:""

Plain-language summary of what the optimized metric does, shown in the review dialog so the user can check intent without reading the code.