All pull requests

Pull requests

• feat(record2dataset): add the Coding-Agent based (AutoEnvScaling) record2dataset session generator and its task checks

  area: docs
  #791·simonucl opened Oct 13, 2026Collaborator·9 tasks done·

  ·

• fix(record2dataset): train the generated-task recipe on multi-turn agent episodes

  area: docs
  #790·simonucl opened Oct 13, 2026Collaborator·9 tasks done·

  ·

• refactor(observability): route experiment events through storage observers

  area: corearea: servicearea: training
  #776·NyanNyanNyanpasu opened Oct 13, 2026·9 tasks done·

  ·
  1

  3

• feat(opd): math example distilling Qwen3-32B into a Qwen3-8B SFT student

  area: docsarea: examplesarea: servicearea: trainingdependencies
  #775·MikeZheng777 opened Oct 13, 2026Contributor·8 of 9 tasks·

  ·

• [Docker] Add the vLLM training image (Slime on torch 2.13 plus vLLM)

  area: ciarea: docsarea: packagingdependencies
  #774·ruizhang0101 opened Oct 13, 2026Collaborator·9 tasks done·

  ·
  2

• fix(service): translate OpenRouter web search items for Codex streams

  area: service
  #769·akushonkamen opened Oct 12, 2026Contributor·9 tasks done·

  ·
  1

• feat(service): host the SR/Envoy routing stack behind a validated manifest

  area: docsarea: service
  #768·akushonkamen opened Oct 12, 2026Contributor·9 tasks done·

  ·
  1

• feat(deploy): add independently managed provider-mode components

  area: docsarea: packagingarea: service
  #759·akramcodez opened Oct 12, 2026·7 of 9 tasks·

  ·
  1

• feat: add AgentCL coding training for SDFT, SDPO and OPD

  area: docsarea: servicearea: trainingdependencies
  #740·boyuan-xai opened Oct 9, 2026·

  ·

• fix(tasks): exclude descendants of excluded parents

  area: corearea: docsarea: service
  #727·Frankie-Xu opened Oct 7, 2026·8 of 9 tasks·

  ·

• feat(reefine): require independent evaluation before publication

  area: ciarea: corearea: docsarea: harnessarea: packagingarea: servicearea: trainingdependencies
  #718·YihongT opened Oct 5, 2026Collaborator·9 tasks done·

  ·
  1

  4

• fix(mlx): restore recovered adapters and synchronize published runtime IDs

  area: harnessarea: servicearea: training
  #711·erniesg opened Oct 3, 2026·8 of 11 tasks·

  ·
  1

• feat(harness): agent teams on the native runtime, run on Harbor tasks

  area: docsarea: harnessarea: servicearea: training
  #702·Benjamin-eecs opened Oct 1, 2026Collaborator·5 tasks done·

  ·

• feat(evaluation): keep eval tasks out of training and stop invalid evaluations from approving publication

  area: corearea: docsarea: harnessarea: servicearea: training
  #700·Benjamin-eecs opened Oct 1, 2026Collaborator·5 tasks done·

  ·

• feat(train): let recipes choose the training method for each job

  area: docsarea: harnessarea: servicearea: training
  #695·BobbyZhouZijian opened Sep 30, 2026Contributor·9 tasks done·

  ·
  1

  2

• refactor: centralize batch IDs and training request validation

  area: corearea: docsarea: harnessarea: servicearea: training
  #675·BobbyZhouZijian opened Sep 28, 2026Contributor·8 of 9 tasks·

  ·

• feat(eval): compare retained harness releases

  area: corearea: docsarea: service
  #515·wbx398 opened Sep 16, 2026Contributor·4 of 6 tasks·

  ·
  5

  

• feat: SPADE roles, the Designer's training by weights or by its prompt, and rounds

  area: docsarea: harnessarea: servicearea: training
  #506·Benjamin-eecs opened Sep 15, 2026Collaborator·8 tasks done·

  ·
  1

  2

  

• feat: MLX training backend for Apple Silicon + OpenClaw-RL end to end

  area: corearea: docsarea: examplesarea: harnessarea: packagingarea: servicearea: trainingdependencies
  #325·BobbyZhouZijian opened Sep 7, 2026Contributor·

  ·
  1