AI Action Item & Task Extractor from Notebook Pages

[allenporter]

Enhance background note processor to parse Gemini OCR output for checkboxes [ ], [x], TODO:, and star marks, saving to ScheduleService.

[allenporter]

Detailed Technical Implementation Plan

1. User Journeys & Example Use Cases

• Journey A: Handwritten Meeting Action Items
  • Scenario: User writes handwritten notes during a team meeting on their Supernote device. On Page 2 of 2026-08-02-TeamSync.note, they write:
    • [ ] Review API security proposal by Friday
    • ⭐ Draft Q3 product roadmap
  • Workflow: Upon device sync, the server's background processor extracts page text via Gemini OCR, identifies the checkbox and star patterns, and automatically populates two new tasks into the user's Inbox schedule group.
  • Outcome: Tasks immediately appear in the Supernote native Task app, Web UI, and synced iCal feeds.

---

2. Requirements & Technical Specifications

A. Extractor Module (supernote/server/services/processor_modules/task_extractor.py)

• Module Signature: Implement standard ProcessorModule interface:

  class TaskExtractorModule(ProcessorModule):
      async def process(self, context: ProcessingContext) -> None: ...

• Pattern Recognition Rules:
  1. Checkbox patterns: - [ ] <title>, [ ] <title>, TODO: <title>
  2. Star mark items: Pages flagged with Supernote star symbols or ⭐ <title>
  3. Due Date Parsing: Regex matching due: YYYY-MM-DD or relative phrases (due: tomorrow, by Friday) mapped to epoch milliseconds.

B. Deduplication & Origin Tracking

• Store origin metadata in task detail: Source: Note/Daily/2026-08-02.note (Page 2).
• Generate deterministic task deduplication key: hash(user_id + notebook_path + page_index + title) stored in ScheduleTaskDO or KV store to prevent duplicate task creation upon re-syncing unchanged pages.

C. Database & Service Integration

• Retrieve or auto-create the default Inbox task group via ScheduleService.get_or_create_inbox(user_id).
• Persist extracted tasks via ScheduleService.create_task().

---

3. Verification & Testing Plan

• Automated Unit Tests (tests/server/test_task_extractor.py):
  • Test extraction against varied OCR outputs (single checkbox, multiple tasks, starred pages, due date strings).
  • Test deduplication: Processing the same page twice must not create duplicate tasks.
  • Test error resilience: Invalid/malformed OCR text handled gracefully without breaking processing pipeline.