10 Commits

• feat(thinking): enable thinking toggle for qwen3 and deepseek models

  Fix #1245

  hkfires · 2026-01-28 09:54:05 +08:00
  c8c27325dc
• feat(claude): clamp thinking budget to max_tokens constraints
  hkfires · 2026-01-19 16:32:20 +08:00
  239a28793c
• refactor(claude): move max_tokens constraint enforcement to Apply method
  hkfires · 2026-01-19 15:50:35 +08:00
  c421d653e7
• refactor(thinking): extract antigravity logic into a dedicated provider
  hkfires · 2026-01-15 19:08:22 +08:00
  4ad6189487
• test(thinking): remove legacy unit and integration tests
  hkfires · 2026-01-15 13:06:40 +08:00
  33d66959e9
• fix(thinking): improve model lookup and validation
  hkfires · 2026-01-15 13:06:40 +08:00
  7f1b2b3f6e
• fix(thinking): use static lookup to avoid alias issues
  hkfires · 2026-01-15 13:06:40 +08:00
  40ee065eff
• fix(executor): properly handle thinking application errors
  hkfires · 2026-01-15 13:06:39 +08:00
  72f2125668
• fix(thinking): fix auth matching for thinking suffix and json field conflicts
  hkfires · 2026-01-15 13:06:39 +08:00
  e8f5888d8e
• refactor: improve thinking logic
  hkfires · 2026-01-15 13:06:39 +08:00
  0b06d637e7