fix(openaiShim): unconditionally set reasoning_content for DeepSeek V…

[skeeminator]

Summary

• DeepSeek V4 (and similar reasoning providers) require reasoning_content
  present on every assistant message in the request history when thinking
  mode is active. The API accepts empty string — the field just needs to
  exist.
• Prior guard only set reasoning_content when thinkingText.trim().length > 0,
  so tool-call turns and edge cases (synthetic interrupts, non-array content)
  sent assistant messages without the field → 400.
• Fix: always set reasoning_content when preserveReasoningContent is true,
  falling back to "" when thinking text is missing. Two additional sites
  (coalescing interrupt, string-content else branch) now also emit the field.

Impact

• user-facing: fixes API Error: 400 {"error":{"message":"The reasoning_content in the thinking mode must be passed back to the API."}}
  for DeepSeek V4 models (deepseek-v4-pro, deepseek-v4-flash) against
  api.deepseek.com. Tool-call rounds and plain multi-turn now work.
• developer/maintainer: gated behind existing preserveReasoningContent
  flag. Non-DeepSeek providers (OpenAI, Gemini, Mistral, etc.) unaffected.

Testing

• bun run build
• bun run smoke
• focused tests: bun test src/services/api/openaiShim.test.ts
• live verification: deepseek-v4-pro against api.deepseek.com —
  multi-turn tool-call round completed without 400

Notes

• provider/model path tested: DeepSeek (api.deepseek.com) with
  deepseek-v4-pro
• follow-up work or known limitations: none

[jatmn]

Will be impacted by #910

[kevincodex1]

lets wait for #910

[gnanam1990]

Tight, surgical fix. Gating behind the existing preserveReasoningContent flag keeps non-DeepSeek providers unaffected, and the three-site coverage in convertMessages matches the failure modes described. Clear repro against api.deepseek.com is appreciated. LGTM.

Optional follow-up: a unit test for the preserveReasoningContent && !thinkingText path would harden this against future regressions.

[AxDSan]

Omgosh, deepseek is unusable right now in OpenClaude because of this :( Will keep following until there is a merge! thanks!

[jatmn]

#910 merged, please rebase and fix conflicts.

[kevincodex1]

hello @skeeminator kindly rebase and fix conflicts this is good to merge .