Policy

Evaluate a card against an agent's historical traces

Replay an agent's recent AIP traces through the policy derived from a card to identify what would have been blocked, escalated, or flagged. By default the agent's current canonical alignment card is used; supply card_json to dry-run a hypothetical card instead. Useful for understanding the impact of a card/policy change before rolling it out. No standalone policy document is accepted (UC-8).

post/policies/evaluate/historical

Request

  • Base URL: https://api.mnemom.ai/v1
  • URL: https://api.mnemom.ai/v1/policies/evaluate/historical
  • Auth: one of:
    • HTTP bearer
    • API key in header X-Mnemom-Api-Key

Request body

agent_idstring required

Agent whose historical traces will be replayed. The caller must be an owner, admin or member of the agent's org; otherwise the response is the same 404 as for an unknown agent.

card_jsonobject

Optional hypothetical alignment card to evaluate against. If omitted, the agent's current canonical card is used (404 if the agent has none and no card_json is supplied).

limitinteger

Maximum number of recent traces to evaluate (capped at 200).

Response

Historical evaluation result

agent_idstring required
card_idstring

Card the traces were evaluated against.

total_tracesinteger

Number of traces returned by the evaluation.

messagestring

Present only when the agent has no historical traces.

Changes

No recorded changes to this operation across all 1 revision of this API.