Skip to content

fix(providers): detect rejected reasoning history - #231

Merged
co-l merged 1 commit into
co-l:developfrom
SulimanAbdulrazzaq:fix/mistral-reasoning-probe
Aug 12, 2026
Merged

fix(providers): detect rejected reasoning history#231
co-l merged 1 commit into
co-l:developfrom
SulimanAbdulrazzaq:fix/mistral-reasoning-probe

Conversation

@SulimanAbdulrazzaq

Copy link
Copy Markdown
Contributor

Summary

Closes #220.

Auto-config now probes whether a thinking-capable provider accepts an assistant history message containing reasoning. A real non-2xx response sets sendReasoningInMessages=false; transport failures remain inconclusive and preserve the default. ProviderModal applies the explicit rejection while leaving reasoning_effort and other thinking settings unchanged.

Validation:

  • src/server/providers/auto-config.test.ts: 17 passed
  • ProviderModal.test.tsx: 13 passed with a dependency-safe targeted configuration
  • Server and E2E TypeScript checks passed
  • ESLint and Prettier passed
  • git diff --check
  • Full suite not run; the standard browser setup is blocked by an existing junctioned dependency path

AI-Enhanced Development

  • AI Models: GPT-5.6 (Codex)

Cache Impact

  • No

Probe assistant reasoning compatibility after thinking support detection and disable history fields only when the provider rejects them.

Closes co-l#220
@co-l
co-l merged commit 9fb608e into co-l:develop Aug 12, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants