VEverdun
training

Rerun Training Sample

Synchronous replay through ii (10-60s). The response is the persisted rerun row — status 'failed' rows come back 200 with the error inline so the network prerequisite (api-backend -> ii security group) degrades visibly, not silently.

post/training/samples/{sample_id}/rerun

Request

  • Base URL
    https://api.verdun.ai (the document declares no server; this is the origin it was published from)
  • URL
    https://api.verdun.ai/training/samples/{sample_id}/rerun
  • Auth
    HTTP bearer

Path parameters

sample_idstring required

Request body

prompt_profile'production' | 'demo' required

Which prompt to replay against: 'production' is the live prompt; 'demo' is the SSM-backed testing-ground prompt (iterable without a redeploy) — the selector makes re-runs a prompt-A/B surface

Response

Successful Response

idstring uuid required

UUID of the rerun row

sample_idstring uuid required

Training sample this rerun replayed

created_atstring date-time required

When the rerun was executed

requested_bystring uuid nullable

Owner who requested the rerun

prompt_profilestring required

Prompt profile the clip was replayed against

statusstring required

Whether the AI pipeline returned a result; failures are persisted too

errorstring nullable

Failure detail when status is 'failed'

rerun_incident_typestring nullable

Incident type from the rerun result

rerun_severitystring nullable

Severity from the rerun result

rerun_confidence_bandstring nullable

Confidence band from the rerun result

rerun_recommended_actionstring nullable

Operator-aligned recommendation from the rerun result

rerun_rationalestring nullable

Model rationale from the rerun result

llm_providerstring nullable

Provider that served the rerun

llm_modelstring nullable

Model that served the rerun

llm_prompt_versionstring nullable

Prompt version the rerun ran under

pipeline_versionstring nullable

Pipeline version the rerun ran under

latency_msinteger nullable

Wall-clock rerun latency in milliseconds

cost_usdstring nullable

Inference cost reported by the pipeline

matches_originalboolean nullable

Rerun recommendation vs the sample's original AI recommendation; null when either side carries none

matches_operatorboolean nullable

Rerun recommendation vs the human's crosswalked expected action; null when either side carries none

Example response

{
  "prompt_profile": "production",
  "status": "succeeded"
}

Changes

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