All pull requests

Pull requests

• feat: add The Grid AI as a gateway model provider

  #3426·CastilloLuis opened Oct 7, 2026·

  ·
  1

• fix(models): drop LiteLLM default temperature when model rejects it

  #3422·KardeniaPoyu opened Oct 6, 2026·

  ·
  1

  1

• fix(metrics): count missing conversational QAG verdicts against the score denominator

  #3420·zstring opened Oct 6, 2026·3 tasks done·

  ·
  1

  1

• fix: hash tool calls with mixed dictionary keys

  #3419·GoldArowana opened Oct 5, 2026·

  ·
  1

• fix: include images from structured retrieval contexts

  #3418·GoldArowana opened Oct 5, 2026·

  ·
  1

• feat(dataset): send the pulled dataset version with test runs

  #3417·ykocaogullar opened Oct 5, 2026Contributor·

  ·
  1

• fix(metrics): let deterministic metrics score multimodal test cases

  #3415·puiji opened Oct 4, 2026·

  ·
  1

• fix(anthropic): add claude-sonnet-5-5 to the model registry

  #3413·VanshBhardwaj24 opened Oct 4, 2026·

  ·
  1

  1

• Feat/community/eval gap metrics

  #3411·MrKyaw opened Oct 4, 2026·

  ·
  1

• fix: make test_core green on main (closes #3372)

  #3410·Raghu6245 opened Oct 4, 2026·

  ·
  2

• fix(benchmarks): load datasets by namespaced Hub repo id

  #3409·chrikrah opened Oct 3, 2026·

  ·
  1

  2

• fix(ragas): stop defaulting Ragas metrics to the deprecated gpt-3.5-turbo

  #3407·manobhisriram opened Oct 3, 2026·

  ·
  1

  2

• fix(tracing): record zero top_k / chunk_size in update_retriever_span

  #3406·Lesereingrape opened Oct 3, 2026·

  ·
  1

• fix(metrics): enforce valid verdict vocabulary in TS prompt-alignment and role-violation (#3391)

  #3405·HarshitR2004 opened Oct 2, 2026·

  ·
  1

  1

• docs: fix template links and a typo

  #3400·pratikgx opened Sep 30, 2026·

  ·
  1

• fix(tracing): restore async generator context on resume

  #3399·zmuxuny opened Sep 30, 2026·

  ·
  1

  2

• fix(models): decide schema support from the signature, not from a TypeError

  #3397·sclfcz opened Sep 30, 2026·

  ·
  1

  1

• fix(voice): let the tool-result fallback cover values json refuses to encode

  #3395·sclfcz opened Sep 30, 2026·

  ·
  1

  1

• fix(tracing): scope eval environment to the trace instead of the shared manager

  #3393·RichardoMrMu opened Sep 30, 2026·

  ·
  2

• feat: add opt-in judge provenance and replay markers

  #3392·seyi opened Sep 30, 2026·

  ·
  2

• feat(metrics): add deterministic production metrics to community

  #3390·MrKyaw opened Sep 28, 2026·

  ·
  4

• fix(openai): trace streams through consumption

  #3389·dafyy321-pixel opened Sep 28, 2026·

  ·
  1

  1

• fix(cli): validate OpenAI model before switching providers

  #3387·dafyy321-pixel opened Sep 28, 2026·

  ·
  1

  1

• fix(evaluate): apply hyperparameters before cache access

  #3385·dafyy321-pixel opened Sep 28, 2026·

  ·
  1

  3

• fix: decouple legacy Dbias/Detoxify backends from install time

  #3383·milangeorge2000 opened Sep 28, 2026·

  ·
  1

  1