[Week7 - hjy] Task2: Extend extraction with priority, deadline, and category

[HyungjoonYang]

Summary

• Rewrote extract_action_items() to return list[ExtractedItem] dataclass instead of list[str]
• ExtractedItem fields: text, priority (high/medium/low), deadline (str|None), category (bug/review/feature/task)
• Expanded keyword matching: TODO, ACTION, FIXME, BUG, HACK, TASK, FOLLOW-UP, REMINDER (kept "ends with !" rule)
• Added priority detection via regex (urgent/critical/asap → high, nice to have/eventually → low)
• Added deadline parsing (by/due/before/deadline patterns)
• Added category detection (bug/fixme → bug, review/PR → review, feature/implement/add → feature)
• Added ExtractedItemRead Pydantic schema
• Added POST /notes/{note_id}/extract endpoint

Test plan

• Keyword matching: todo, action, fixme, bug, hack, task, follow-up, reminder (10 tests)
• Priority: high (urgent, critical, asap, [P0]), low (nice to have, eventually, [P2]), default medium (8 tests)
• Deadline: by/due/before patterns, none case (4 tests)
• Category: bug, fixme→bug, review, feature, default task (5 tests)
• Return type, empty text, multiline, combined extraction (4 tests)
• Endpoint: extract from note, 404 for nonexistent (2 tests)
• All 35 tests pass

Notable tradeoffs

• Used dataclass for ExtractedItem (lightweight, no Pydantic overhead for internal logic)
• Deadline regex captures to end of line which may over-capture in edge cases
• Priority/category detection is keyword-based regex; NLP would improve accuracy but adds complexity

🤖 Generated with Claude Code