MOmoda-dev

Get hallucination detections

Returns hallucinations (agent replies the conversation shows were false, each naming its Problems incident in incident_id) and earlier verified claims, with evidence and aggregate counts.

get/api/v1/data/hallucinations

Request

  • Base URL
    https://moda.dev
  • Auth
    API key in header x-api-key

Query parameters

days_backinteger
limitinteger
offsetinteger
conversation_idstring
conversation_idsstring

Scope every figure to several traces: up to 50 trace IDs, comma-separated. Cannot be combined with conversation_id.

kind'contradicted' | 'verified'

Filter the rendered detection list.

group_by'message' | 'trace'

trace reads the dashboard's per-trace list: rows in traces, totals count traces (counting: per_trace). The default lists one row per scored message (counting: per_message).

searchstring

Trace ID substring. Needs group_by=trace.

status'contradicted' | 'all'

contradicted: traces with a contradiction; all: also verified-only traces. Needs group_by=trace.

rulestring

One rule_id; scopes the summary too. Needs group_by=trace.

source'rule' | 'classifier' | 'gpu' | 'judge' | 'human'

Needs group_by=trace.

sort'contradicted_count' | 'verified_count' | 'last_detected_at'

Needs group_by=trace.

order'asc' | 'desc'

Needs group_by=trace.

cohortstring

builtin:<key> or a saved cohort id: only traces of the cohort's members, summary included. Needs group_by=trace.

Response

Successful response.

object required

Hallucination summary, rule breakdown, detections (or traces with group_by=trace), pagination information, and counting, which says whether totals count messages or traces.

Changes

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