All pull requests

Pull requests

• fix: pass plaintext reasoning back to DeepSeek and DashScope models so tool-calling subagents survive their second request

  #385·xh123637 opened Nov 3, 2026·

  ·
  1

• feat(providers): add Obscurify.ai predefined endpoint (obscurify@)

  #369·tonydavis629 opened Oct 28, 2026·

  ·

• feat(openai-codex): make Claude Code /fast send the account's Fast tier

  #368·dingyifei opened Oct 28, 2026Contributor·

  ·
  2

• feat(stream): opt-in first-useful-event watchdog for admitted-but-mute upstreams

  #355·jsboige opened Oct 25, 2026Contributor·

  ·

• feat(catalog): add Cheaper Inference predefined endpoint

  #344·aiapienthusiast opened Oct 23, 2026·

  ·

• test(routing): stub Devin discovery in routing regression

  #340·blueDaydreaming725 opened Oct 22, 2026·

  ·

• fix(catalog): skip warming for native-only launches

  #339·blueDaydreaming725 opened Oct 22, 2026·

  ·

• fix(effort): replace source NUL bytes with escapes

  #338·blueDaydreaming725 opened Oct 22, 2026·

  ·

• fix: support OpenRouter preset model IDs

  #331·blueDaydreaming725 opened Oct 21, 2026·

  ·
  1

• feat(discovery): list a custom endpoint's own models in the picker

  #287·anandghegde opened Sep 30, 2026·

  ·
  1

  

• fix(runner): keep a user's settings env off the sessions claudish routes

  #282·MayCXC opened Sep 25, 2026·

  ·
  2

  

• fix(runner): stop Claude Code before exiting on a signal, and report any signal's number

  #281·MayCXC opened Sep 25, 2026·

  ·
  2

  

• feat(daemon): run Claude Code's supervisor behind a --monitor proxy

  #280·MayCXC opened Sep 25, 2026·

  ·
  2

  

• fix(monitor): forward a --monitor session as the host it stands in for

  #278·MayCXC opened Sep 25, 2026·

  ·
  2

  

• fix(test): keep the suite out of the real port-keyed token file

  #268·MayCXC opened Sep 19, 2026·

  ·
  2

  

• fix(catalog): honour CLAUDISH_CATALOG_URL on every models-index endpoint

  #267·MayCXC opened Sep 19, 2026·

  ·
  2

  

• perf(mcp): dispatch run_prompt/compare_models in-process via app.fetch

  #257·MayCXC opened Sep 15, 2026·

  ·
  2

  

• fix(cache): cached tokens in the reported prompt, and Kimi's top-level cache_control

  #256·MayCXC opened Sep 15, 2026·

  ·
  2

  

• fix(local): detect Ollama backend by probe, not exact provider name

  #255·MayCXC opened Sep 15, 2026·

  ·
  2

  

• refactor: collapse the three provider queues onto a shared RequestQueue base

  #254·MayCXC opened Sep 15, 2026·

  ·
  2

  

• feat(mcp): add files param to run_prompt and compare_models

  #253·MayCXC opened Sep 15, 2026·

  ·
  2