docs(lark-drive): add answer-oriented search refinement

[BytedanceSearch]

Summary

Adds answer-oriented search refinement guidance to drive +search.

The change is intentionally scoped to one search strategy doc:

• Preserve hard constraints such as business entities, time, product names, and numeric terms.
• Add query expansions by question type, such as case, policy, metric, and integration questions.
• Fetch strong candidates instead of stopping after one raw query.
• Keep suspected eval/process artifacts observable rather than hidden.
• Require final answers to fill the key slot for the question type.

Eval Search Run

• Run ID: 2026-05-07T11-35
• Cloud report: https://www.feishu.cn/docx/JM1kd1QY7ocX2wxxCDScTDFbn5c
• Cases searched: 14
• Cases scored: 14
• Score: 62 / 210 (29.5%)
• Recall avg: 0.93
• Accuracy avg: 1.93
• Completeness avg: 1.57

Pollution handling:

• Tainted search results observed: 43
• Non-evidential results excluded: 43
• Tainted fetches: 0
• Contamination penalty cases: 0

Findings Addressed

• F-001 / search_strategy / high: add type-aware blind refinement for short queries, metrics, launch summaries, customer cases, and policy searches.
• F-003 / skill_prompts / high: require slot-complete final answers instead of long source excerpts.

Unhandled:

• F-002 / tool_capability / medium: message applinks, external webpages, and file/PDF fetch support remain collector/tool capability gaps.

Checks

• git diff --check
• node --experimental-strip-types tests/eval-search/eval-search-run.ts --help
• node --experimental-strip-types tests/eval-search/eval-search-collect-search.ts --help
• JSON parse for generated summary.json, verdicts.json, and executor_search.json

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 40f304da-6677-4d52-9c7d-049c1cc03f58

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch eval-search-auto-pr-2026-05-07T11-35

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 65.40%. Comparing base (7c68639) to head (75f2aec).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #773   +/-   ##
=======================================
  Coverage   65.40%   65.40%           
=======================================
  Files         508      508           
  Lines       46795    46795           
=======================================
  Hits        30605    30605           
  Misses      13548    13548           
  Partials     2642     2642           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.