CEcekura
observability

Reproduce a failure mode by running it against an agent

Changed on

Draft a scenario from one failure mode and run it against the agent on the same channel the cited calls used, retrying until a run reproduces the failure or max_attempts is reached. Choose the failure mode with failure_mode_index and the target with agent. Returns a reproduction_session_id; poll the reproduction session for status. ⚠ Executes real evaluator runs against the agent.

post/observability/v1/metric-failure-mode-insights/{id}/reproduce/

Request

  • Base URL
    https://api.cekura.ai
  • URL
    https://api.cekura.ai/observability/v1/metric-failure-mode-insights/{id}/reproduce/
  • Auth
    • API key in header X-CEKURA-API-KEY
    • or HTTP bearer
    • or HTTP bearer
    • or HTTP bearer

Path parameters

idinteger required

A unique integer value identifying this metric failure mode insight.

Request body

failure_mode_indexinteger required

Zero-based index of the failure mode within the insight's failure_modes list.

agentinteger required

ID of the agent to reproduce the failure against (resolved from the failure mode's example calls on the frontend).

max_attemptsinteger

How many times to run the scenario before giving up. The loop stops early as soon as one run reproduces the failure. Defaults to 10.

redact_transcriptsboolean

If true, strip personal identifiers from the example call transcripts before they are used to draft the reproducing scenario. Semantic context is preserved. Defaults to false.

Response

failure_mode_indexinteger required

Zero-based index of the failure mode within the insight's failure_modes list.

agentinteger required

ID of the agent to reproduce the failure against (resolved from the failure mode's example calls on the frontend).

max_attemptsinteger

How many times to run the scenario before giving up. The loop stops early as soon as one run reproduces the failure. Defaults to 10.

redact_transcriptsboolean

If true, strip personal identifiers from the example call transcripts before they are used to draft the reproducing scenario. Semantic context is preserved. Defaults to false.

Changes