All pull requests

Pull requests

• feat: add Cheaper Inference as an OpenAI-compatible provider

  #146·aiapienthusiast opened Sep 30, 2026·

  ·

• fix: weight Wilson CI successes by effective samples

  #143·rudycelekli opened Sep 28, 2026·

  ·

• fix: verify schema v2 manifest run IDs

  #142·rudycelekli opened Sep 28, 2026·

  ·

• Reject malformed governance fields in legacy fixtures

  #141·rudycelekli opened Sep 28, 2026·

  ·

• Report malformed fixture files as validation errors

  #140·rudycelekli opened Sep 28, 2026·

  ·

• fix: fail layout validation for misnamed inspection folders

  #139·rudycelekli opened Sep 28, 2026·

  ·

• fix: reject incomplete or extra ground-truth verdicts

  #138·rudycelekli opened Sep 28, 2026·

  ·

• fix: reject string verdicts that inflate atomic grounding scores

  #137·rudycelekli opened Sep 28, 2026·

  ·

• fix: make CI example fail partial and unscored runs

  #136·rudycelekli opened Sep 28, 2026·

  ·
  3

• fix: enforce judge call budget under concurrent evaluations

  #135·rudycelekli opened Sep 28, 2026·

  ·

• fix: show unscored artifact results without false grades or verdicts

  #134·rudycelekli opened Sep 28, 2026·

  ·

• fix: fail inspections whose consistency cap drops below threshold

  #133·rudycelekli opened Sep 28, 2026·

  ·

• fix: compare missing and unscored inspections without false gaps

  #132·rudycelekli opened Sep 28, 2026·

  ·

• fix: avoid granting destructive tools to non-admin roles

  #131·rudycelekli opened Sep 28, 2026·

  ·

• fix: reject inconsistent checkpoint entries during resume

  #130·rudycelekli opened Sep 28, 2026·

  ·

• fix: generate authorization cases using provider role IDs

  #129·rudycelekli opened Sep 28, 2026·

  ·

• fix: preserve audit inputs when saving generated fixtures

  #128·rudycelekli opened Sep 28, 2026·

  ·

• fix: stabilize generated authorization case IDs across runs

  #127·rudycelekli opened Sep 28, 2026·

  ·

• fix: compare scorecards with unscored overall results

  #126·rudycelekli opened Sep 28, 2026·

  ·

• fix: handle malformed HTTP success payloads as provider errors

  #125·rudycelekli opened Sep 28, 2026·

  ·

• fix: prevent stale HTTP sources from contaminating retrieval evidence

  #124·rudycelekli opened Sep 28, 2026·

  ·

• feat: add API Route provider

  #123·DennyHo0917 opened Sep 27, 2026·

  ·

• fix: judge calls ran twice

  #121·Sebabaian opened Sep 24, 2026Collaborator·

  ·