admin

List the lowest-scored replies

The worst replies first, for internal QA inspection (scores are never tenant-facing).

dimension is validated against the known score columns (fail-closed to overall), so it can't be an injection vector. scorer_version defaults to the current generation so the default view never mixes non-comparable versions; all lifts the filter. Cross-tenant by default; pass tenant_id to drill in.

get/v1/admin/reply-quality

Request

  • Base URL: https://api.patricia.app (the document declares no server; this is the origin it was published from)
  • URL: https://api.patricia.app/v1/admin/reply-quality
  • Auth: HTTP bearer

Query parameters

tenant_idstring uuid nullable

Drill into one workspace; omit for the fleet view.

dimensionstring

Sort worst-first by this dimension.

sincestring date-time nullable

Only replies scored at/after this time (UTC).

untilstring date-time nullable

Only replies scored before this time (UTC).

scorer_versionstring

Which grading generation to show. Defaults to the current scorer version, because scores from different versions are NOT comparable, so mixing them (old low rows dominating the worst-first list) is misleading. Pass 'all' to inspect every version.

review_statusstring nullable

Triage state. A reply nobody has reviewed counts as 'open'.

outcome_codestring nullable

Judge-classified outcome of the reply.

reply_type_codestring nullable

Judge-classified kind of reply.

max_overallnumber nullable

Only replies at or below this overall score.

limitinteger
offsetinteger

Response

Successful Response

idstring uuid required
tenant_idstring uuid required
run_idstring uuid nullable
conversation_idstring uuid nullable
correctnessinteger required
correctness_notestring required
groundednessinteger required
groundedness_notestring required
instruction_followinginteger required
instruction_following_notestring required
tone_voiceinteger required
tone_voice_notestring required
actionabilityinteger required
actionability_notestring required
overallnumber required
reply_type_codestring required
outcome_codestring required
topic_codestring nullable
tool_usedboolean required
tool_roundsinteger required
tool_familiesstring[] required
output_type_codestring required
complexity_codestring required
judge_modelstring required
effort_tierstring required
judge_cost_micro_usdinteger required
scorer_versionstring required
rubric_versionstring required
parse_statusstring required
sample_reasonstring required
created_atstring date-time required

Changes

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