Document-to-Notes

Overview

The Document-to-Notes feature takes an uploaded document (PDF, DOCX, TXT, or Markdown), extracts its text client-side, sends it to the AI, and returns a set of candidate sticky notes for you to review and import.

User flow

1. Click Upload in the bottom toolbar (or the Upload tab in the note overlay).
2. Drag and drop a file, or click the file picker.
3. The client extracts text locally — no raw file is sent to the server.
4. The extracted text is sent to the AI.
5. A preview grid of generated notes appears.
6. Select the notes you want and click Import to board.

Supported formats

Text extraction

Extraction happens client-side for responsiveness and privacy:

The extractor returns { text, truncated, charCount }.

Truncation

If the document is too large, extraction truncates to a safe maximum character count. When this happens:

• The UI shows a truncation warning.
• Only the first chunk of the document was used for generation.
• Fix: Upload smaller sections, or use a focused instruction prompt.

AI output contract

The server returns a Zod-validated JSON object:

{
  "notes": [
    {
      "title": "Decision — Use edge-side auth",
      "content": "**Rationale:** reduces latency...",
      "suggestedColor": "purple",
      "suggestedType": "Decision"
    }
  ]
}

Between 1 and 15 notes are generated per upload. The AI tries to identify distinct PM artifacts (decisions, epics, user feedback, action items) in the document.

Tips for better results

• Focus the prompt — add a brief instruction in the prompt bar: "Extract decisions and open questions only."
• Upload focused sections — a specific requirements section produces tighter notes than an entire 50-page spec.
• Review before importing — deselect notes that are too generic or don't fit the current board.

See also

• Boards and Notes — Note types and colours
• Plans and Limits — Document upload quotas
• AI Overview — AI safety principles