fix: preserve partial LLM completion on stream abort

[amabito]

Summary

• When a user aborts (Stop) a streaming response, PromptLog.completion was left empty
• This change accumulates streamed content and persists partial completion on abort so logs/analytics remain accurate

Changes

• core/llm/streamChat.ts — accumulate chunk content during streaming; on abort, set errorPromptLog.completion before returning
• core/llm/streamChat.test.ts — 3 test cases: normal abort, immediate abort, MessagePart[] content

Details

• Covers both legacy slash-command path and standard chat path
• Handles string and MessagePart[] content types
• No new dependencies
• Existing yield/UI streaming behavior unchanged
• ~16 lines of logic added

Test plan

• streamChat.test.ts passes (3 cases)
• Existing tests unaffected
• Manual: abort mid-stream, verify PromptLog.completion is populated

🤖 Generated with Claude Code

---

Continue Tasks: ✅ 1 no changes — View all

---

Summary by cubic

Preserves partial LLM completion when a user stops a streaming response, so PromptLog.completion and analytics stay accurate. Works for both legacy slash-command and standard chat paths.

• Bug Fixes
  • Accumulates streamed chunks and persists partial text on abort (supports string and MessagePart[] content).
  • Adds Vitest unit tests for content extraction and accumulation, including mixed content.

^{Written for commit 87fb47a. Summary will update on new commits.}

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[amabito]

I have read the CLA Document and I hereby sign the CLA

[cubic-dev-ai]

No issues found across 2 files

[amabito]

I have read the CLA Document and I hereby sign the CLA

[amabito]

recheck

[amabito]

I have read the CLA Document and I hereby sign the CLA