Problem

Users frequently want to add the contents of a screenshot (e.g. a slide they saw, an error message, a printed page) into NotebookLM as a source. Today they have to type it manually or use an external OCR tool.

Proposed approach

1. Add a "Screenshot → Source" button in the title bar (or accept a paste of an image from the clipboard).
2. Run the image through tesseract.js (~3 MB dep) to extract text in the main process.
3. Inject the extracted text as a new note via the existing quick-clip-paste mechanism in src/webview-preload.js.

Files to touch

• src/main.js — IPC handler for OCR
• src/webview-preload.js — paste extracted text
• New runtime dep: tesseract.js

Acceptance criteria

• Pasting an image (Ctrl+V on the main window) triggers OCR.
• Extracted text appears in the active NotebookLM pane within 5s for a 1-page screenshot.
• Falls back to "OCR failed" notification if image is unreadable.