fix: eliminate double-summarization in conversations_to_string (#4655)

[beastoin]

Summary

• conversations_to_string() previously included both structured.overview AND apps_results[0].content, feeding redundant content to all 12 downstream LLM callers
• Fix: when apps_results[0] has non-empty content, use it instead of overview — never both
• Empty/whitespace app results fall back to overview (existing behavior preserved)

Changes

• backend/models/conversation.py: Prefer app_result as summary, remove duplicate "Summarization:" block
• backend/tests/unit/test_conversations_to_string.py: 7 unit tests covering selection logic
• backend/test.sh: Register new test file

Test plan

• 7 unit tests pass (overview-only, app_result replacement, empty/whitespace fallback, no-double-summary, title-always-present, multiple-app-results)
• Full backend test suite passes
• Verify downstream LLM output quality improvement (shorter, non-redundant context)

Closes #4655

🤖 Generated with Claude Code

[beastoin]

Review: APPROVED

Reviewed the diff for correctness, safety, and test coverage.

Correctness: Fix is clean — when apps_results[0].content is non-empty, it replaces overview (never both). Empty/whitespace falls back to overview. Old "Summarization:" block removed.

Safety: AppResult.content is str (non-optional), so .strip() is safe. Index [0] is guarded by the truthy + len check.

Minor nit (non-blocking): conversation.apps_results and len(conversation.apps_results) > 0 — the len() check is redundant since an empty list is already falsy. Fine to leave for readability.

Tests: 7 tests cover all branches (overview-only, app_result replacement, empty fallback, whitespace fallback, no-double-summary, title-always-present, multiple-app-results). Good coverage.

One subtle improvement: Old code omitted \n after app content; new code adds it consistently. This is actually a fix for downstream formatting.

No issues found. Ship it.

---

Reviewer: AI for @beastoin

[gemini-code-assist]

Code Review

This pull request effectively addresses the issue of double summarization in conversations_to_string. The change correctly prioritizes the apps_results summary over the structured.overview, with a proper fallback mechanism for empty or missing app summaries. The implementation is clean and is supported by a thorough new test suite that covers all relevant edge cases. The changes improve data quality for downstream consumers by removing redundant information. Overall, this is a high-quality contribution, and I did not find any high or critical severity issues.

[github-actions]

Hey @beastoin 👋

Thank you so much for taking the time to contribute to Omi! We truly appreciate you putting in the effort to submit this pull request.

After careful review, we've decided not to merge this particular PR. Please don't take this personally — we genuinely try to merge as many contributions as possible, but sometimes we have to make tough calls based on:

• Project standards — Ensuring consistency across the codebase
• User needs — Making sure changes align with what our users need
• Code best practices — Maintaining code quality and maintainability
• Project direction — Keeping aligned with our roadmap and vision

Your contribution is still valuable to us, and we'd love to see you contribute again in the future! If you'd like feedback on how to improve this PR or want to discuss alternative approaches, please don't hesitate to reach out.

Thank you for being part of the Omi community! 💜