feat: feishu_doc Markdown 富文本写入 - #122

Claude finished @lishuceo's task in 2m 2s —— View job

PR Review: feat: feishu_doc Markdown 富文本写入

• Check for previous review comments
• Read changed files and understand context
• Review markdown-to-blocks.ts (new, 355 lines)
• Review doc.ts changes
• Review test files
• Post review summary

✅ Approved — clean implementation with minor suggestions

Overall assessment: This is a well-structured PR. The markdown-to-blocks converter is thorough, handles edge cases sensibly (tables → code block fallback, blockquotes → text fallback), and the test coverage with 39 new tests is solid. The integration into doc.ts is clean and minimal — only the write/append paths changed, replacing raw text insertion with the new block conversion + batching.

2 minor suggestions posted as inline comments:

• 🟡 Unclosed code fence (confidence: 82) — An opening ``` with no closing fence silently consumes all remaining lines. The i++ on line 197 increments past lines.length, which is harmless but means truncated LLM output could lose trailing content. Suggested a guard: only skip closing fence if found.

• 🟡 encodeURI double-encoding (confidence: 78) — encodeURI(rawUrl) on line 117 will corrupt pre-encoded URLs (e.g. %20 → %2520). Since markdown links are typically well-formed, passing the URL through directly or using new URL().href would be safer.

Things done well:

• Correct batch splitting (50 blocks per request) to respect Feishu API limits
• Todo checkbox parsing ordered before bullet list parsing (avoids - [ ] being captured as a bullet)
• Long content splitting in splitLongContent with line-boundary awareness
• Comprehensive language map for code blocks
• Tests cover inline formatting, all block types, mixed content, and edge cases (empty input, whitespace-only)

Claude finished @lishuceo's task in 1m 47s —— View job

PR Review: feat: feishu_doc Markdown 富文本写入 (re-review after sync)

• Check previous review comments and resolve fixed issues
• Read current changed files
• Review for new issues
• Post summary

⚠️ Issues Found

Previous issues (2 still open):

• 🟡 Unclosed code fence (markdown-to-blocks.ts:197) — still present, not addressed in new commit
• 🟡 encodeURI double-encoding (markdown-to-blocks.ts:117) — still present, not addressed in new commit

New issue (1):

• 🔴 insert_blocks batch ordering bug (doc.ts:185-195, confidence: 90) — When content exceeds 50 blocks, each batch is inserted at the same index, resulting in batches appearing in reverse order. Fix: increment index by batch.length after each insertion.

New code in commit 4bfb88a (block-level edit operations) is otherwise clean:

• update_block correctly uses parseInlineMarkdown for single-block text updates
• delete_blocks properly resolves parent → child index before calling batchDelete
• Input validation with validateToken for block_id is good
• Test coverage for the 3 new actions is thorough (7 new test cases)

Claude finished @lishuceo's task in 1m 25s —— View job

PR Review: feat: feishu_doc Markdown 富文本写入 (re-review after sync)

• Check previous review comments and resolve fixed issues
• Read current changed files
• Review for new/remaining issues
• Post summary

⚠️ Issues Found — 1 fixed, 2 still open

Previous issues status:

• ✅ insert_blocks batch ordering bug — Fixed in 8fda34a. Now correctly increments currentIndex += batch.length after each batch. Thread resolved.
• 🟡 Unclosed code fence (markdown-to-blocks.ts:197, confidence: 82) — Still present. An opening ``` with no closing fence silently consumes all remaining lines.
• 🟡 encodeURI double-encoding (markdown-to-blocks.ts:117, confidence: 78) — Still present. encodeURI(rawUrl) corrupts pre-encoded URLs (%20 → %2520).

Latest commit 8fda34a improvements (all good):

• Added API response validation (code !== 0 checks) to all previously unchecked API calls in write/append/update_block/delete_blocks
• Added z.number().int().min(0) validation on the index parameter
• Improved update_block tool description to clarify inline-only Markdown support

Overall: The critical bug (batch ordering) is fixed. The two remaining warnings are low-severity edge cases — the unclosed fence issue only affects truncated LLM output, and the encodeURI issue only affects pre-encoded URLs. Both are safe to address in a follow-up.

Claude finished @lishuceo's task in 55s —— View job

PR Review: feat: feishu_doc Markdown 富文本写入 (re-review after fixes)

• Check previous review comments and verify fixes
• Read current changed files
• Review for any remaining issues
• Resolve fixed comment threads
• Post summary

✅ Approved — all previous issues fixed

Previous issues (3 resolved):

• ✅ Unclosed code fence — Fixed in 52fd448. Added if (i < lines.length) i++ guard so unclosed fences no longer skip past array bounds.
• ✅ encodeURI double-encoding — Fixed in 52fd448. Removed encodeURI(), raw URL is passed directly to the Feishu API.
• ✅ insert_blocks batch ordering — Fixed in 8fda34a. currentIndex += batch.length correctly maintains insertion order.

No new issues found. The latest commit cleanly addresses both remaining warnings. Code looks good to merge.