feat(notes): add ai-powered url analyzer with opt-in toggle

[sanjanasingineedi]

Summary

Closes #42

Adds a Source URL Manager and AI-powered web content analyzer to the Notes Panel, allowing students to save reference URLs and generate structured study notes from them using a local AI model.

Type of Change

• ✨ New feature (non-breaking change that adds functionality)

Related Issues

Closes #42

Changes Made

Part 1 — Source URL Manager (always visible)

• New "Source URLs" section in the notes editor
• Paste any public http/https URL and click Add to save it
• Multiple URLs can be saved per note
• URLs are stored in the database (source_urls column) with automatic migration for existing databases
• Saved URLs are clickable (opens in default browser) with a remove button
• URLs persist when switching between notes

Part 2 — AI Web Analyzer (opt-in toggle, off by default)

• Toggle switch "AI Analyzer" — disabled by default as requested
• When enabled, "Analyze with AI" button appears
• On click, the Electron main process fetches all saved URLs (bypasses CORS), extracts readable text, and sends it to the local AI for analysis
• AI generates structured study notes shown in a two-panel modal:
  • Left panel: section headings
  • Right panel: full content of selected section (selectable text)
• Three import options:
  • Import Selected Text — highlight specific lines and import only those
  • Import Full Section — append entire section as markdown
  • Append All — import all sections at once
• Graceful error message when no AI model is available

AI Stack

• Tries embedded local LLM first (node-llama-cpp)
• Falls back to Ollama (same pattern as existing generateNoteInsights)
• Default Ollama model is phi (as configured in the project)
• Shows a clear message if neither is available
• Note: larger models (llama3, mistral) produce better structured output; smaller models like phi or tinyllama may return plain text which is auto-parsed into sections as a fallback

Files Changed

File	Change
src/main/db/database.ts	Added source_urls column + automatic migration for existing DBs
src/main/db/queries.ts	Updated insertNote and updateNote to include source_urls
src/main/handlers/ipcHandlers.ts	Added notes:fetchUrl and notes:analyzeUrls IPC handlers
src/preload.ts	Exposed window.api.notes.fetchUrl and window.api.notes.analyzeUrls via contextBridge
src/shared/ipc.ts	Added IPCUrlSection, IPCFetchUrlResult, IPCAnalyzeUrlsResult types; added source_urls to IPCNote, IPCNoteCreateInput, IPCNoteUpdateInput
src/index.d.ts	Updated IPCNotes interface with fetchUrl and analyzeUrls
src/components/NotesPanel.tsx	Full UI implementation — URL manager, toggle, analyzer modal

Testing

• Tested with multiple GFG articles (OS, DBMS topics)
• Tested with 2 URLs analyzed together
• Verified toggle on/off behavior
• Verified Import Selected Text, Import Full Section, Append All
• Verified graceful fallback when no model available
• Verified URL persistence across note switching
• Verified note saving with source URLs

Limitations (by design)

• Only public http/https URLs supported
• Pages behind login walls (ChatGPT, Gemini, Claude) not supported
• AI quality depends on locally available model — larger models
  (llama3, mistral) produce better structured output than tinyllama

Known Limitation & Proposed Improvement

When the AI returns plain text instead of structured JSON (common with smaller models like phi or tinyllama), the response is auto-parsed into sections by sentence chunking, which may produce fewer sections than expected.

Proposed follow-up improvement: Pre-section the content using HTML headings extracted during fetch (already available in the pipeline), then use AI only to summarize each section — this would give reliable sections regardless of model size and is planned as a follow-up.

Checklist

• I have followed the code style guidelines
• I have tested my changes locally
• My changes don't introduce new warnings or errors

[ManabBiswas]

@sanjanasingineedi can you give me demo video or its link here.

[ManabBiswas]

@sanjanasingineedi can you give me any screenshots?

@sanjanasingineedi can you give me demo video or its link here.