All pull requests

Pull requests

• Require explicit function call IDs in Responses input

  #315·julianvanceblackwood opened Sep 24, 2026·

  ·
  1

  3

• Run the python tool's blocking backends off the event loop

  #312·Linxiushen opened Sep 21, 2026·

  ·

• fix: import get_tokenizer in reference-system-prompt.py

  #311·MiamiBeachBroker opened Sep 9, 2026·

  ·

• princess

  #310·princess38827 opened Sep 9, 2026·

  ·

• Reject duplicate function call ids in /v1/responses input

  #309·AUTHENSOR opened Sep 5, 2026·

  ·
  1

• Fix duplicate token append in Metal generator

  #306·natan4903 opened Aug 27, 2026·

  ·
  2

• Add OrcaRouter as a gpt-oss inference partner in awesome-gpt-oss.md

  #305·kuswardhanietidims-svg opened Aug 26, 2026·

  ·

• Fix HealthBench rubric grade ordering

  #304·Excelius-Wang opened Aug 25, 2026·

  ·

• Fix TypeError in gpt_oss.generate with the default --limit

  #303·AryantJadhav opened Aug 24, 2026·

  ·
  2

• Add missing pytest-asyncio to the test extra

  #302·AryantJadhav opened Aug 24, 2026·

  ·
  1

• Route functions.search-style calls to the browser tool

  #301·AryantJadhav opened Aug 24, 2026·

  ·
  4

• Fix four crashes in the Responses API server

  #300·AryantJadhav opened Aug 24, 2026·

  ·
  4

• Restore html2text globals safely after browser conversion

  #297·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  2

• Handle zero-example eval subsets and empty progress mapping

  #295·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  2

  5

• Honor explicit eval sample counts and repeat selection

  #293·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  3

  10

• Synchronize distributed generation and execute chat tools on rank zero

  #291·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  2

  5

• Keep Harmony special token IDs unique

  #287·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  2

• Normalize nullable Responses request fields before server use

  #284·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  1

• Reset Responses stub output for each new request

  #282·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  1

• Stop injecting placeholder tokens in Responses Ollama inference

  #280·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  4

• Expand Responses API checkpoint paths before backend setup

  #278·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  1

• Parse Responses vLLM tensor parallel size as an integer

  #276·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  4

• Use local rank for Responses Triton CUDA binding

  #274·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  1

• Allocate Responses Triton KV caches on the model device

  #272·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1

  2

• Preserve unlimited token sentinel in text generation CLI

  #269·sylvesterkaczmarek opened Aug 16, 2026·

  ·
  1