Smart Notes MVP

A modular, offline-first RAG system for local notes powered by Ollama. Chat with your notes using AI - everything runs locally.

Features

• Vault-based Storage: Store notes as .md files in the vault/ folder
• Rich Text Editor: TipTap-powered editor with formatting toolbar
• Auto-Indexing: Automatic re-indexing when files change (chokidar)
• RAG Pipeline: Keyword retrieval + Ollama AI generation
• Modular Architecture: Clean separation for easy extension
• Local-first: No external APIs - runs entirely on your machine

Project Structure

Smart notes/
├── index.js              # Main entry point
├── package.json          # Dependencies
├── build-editor.js       # TipTap bundle builder
├── db.json              # Indexed chunks (auto-generated)
├── src/
│   ├── storage/         # Vault operations (CRUD)
│   ├── indexing/        # Note chunking (150-200 words)
│   ├── retrieval/       # Keyword-based retrieval
│   ├── ai/             # Ollama integration
│   ├── watcher/        # File system watcher (chokidar)
│   └── server/          # Express API server
├── public/
│   ├── index.html      # Web interface
│   └── vendor/
│       └── editor.js   # Bundled TipTap editor
└── vault/              # Your notes (.md files)

Quick Start

1. Install Dependencies

npm install

2. Build the Editor (first time only)

npm run build-editor

Or: node build-editor.js

3. Start Ollama (if not running)

ollama serve

4. Start the Server

npm start

5. Open Browser

Navigate to: http://localhost:3000

Usage

Creating Notes

1. Click + New Note in the Notes panel
2. Enter a filename (e.g., my-notes.md)
3. Use the toolbar to format text (Bold, Italic, Headings, Lists, etc.)
4. Click Save

Asking Questions

1. Type your question in the chat input
2. Press Enter or click Send
3. The AI answers based on your notes
4. Sources are shown below the answer

Editing/Deleting Notes

• Click Edit on any note to modify it
• Click Del to delete a note
• Auto-indexing updates the search index automatically

API Endpoints

Chat & Retrieval

Method	Endpoint	Description
POST	/api/ask	Ask a question about your notes
GET	/api/health	Server status and health check

Note Management

Method	Endpoint	Description
GET	/api/notes	List all notes
GET	/api/notes/:filename	Read a specific note
POST	/api/notes	Create a new note
PUT	/api/notes/:filename	Update a note
DELETE	/api/notes/:filename	Delete a note

Indexing

Method	Endpoint	Description
GET	/api/sources	List all note sources
GET	/api/chunks	View all indexed chunks
POST	/api/reindex	Manually trigger re-indexing

How It Works

┌─────────────────────────────────────────────────────────────┐
│                         User Query                          │
│                     "What is Node.js?"                      │
└─────────────────────────────┬───────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                      RETRIEVAL LAYER                        │
│  1. Tokenize query into keywords                           │
│  2. Score each chunk by keyword overlap                     │
│  3. Return top-3 most relevant chunks                       │
└─────────────────────────────┬───────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                       AI LAYER (Ollama)                    │
│  Build prompt with retrieved context                        │
│  Send to qwen2.5-coder:1.5b model                          │
│  Return generated answer                                    │
└─────────────────────────────┬───────────────────────────────┘
                              │
                              ▼
                    Answer + Sources

Architecture

Storage Module (src/storage/)

• Purpose: CRUD operations for vault files
• Methods: listNotes(), readNote(), writeNote(), deleteNote()
• Location: vault/ directory

Indexing Module (src/indexing/)

• Purpose: Split notes into searchable chunks
• Config: 150-200 words per chunk
• Output: Stored in db.json

Retrieval Module (src/retrieval/)

• Purpose: Find relevant chunks for queries
• Method: Keyword overlap scoring
• Output: Top-3 matching chunks

AI Module (src/ai/)

• Purpose: Generate responses using Ollama
• Model: qwen2.5-coder:1.5b
• Config: Temperature=0.3, top_p=0.9

Watcher Module (src/watcher/)

• Purpose: Monitor vault for changes
• Library: chokidar
• Trigger: Auto re-indexing on add/change/delete

Server Module (src/server/)

• Purpose: Express REST API
• Endpoints: All API routes defined here

Editor (src/editor/)

• Purpose: TipTap rich text editor
• Build: Bundled with esbuild to public/vendor/editor.js

Editor Features

The TipTap editor supports:

• Bold, Italic, Strikethrough
• Headings (H1, H2, H3)
• Bullet Lists & Numbered Lists
• Blockquotes
• Inline Code & Code Blocks
• Undo/Redo

Future Extensions

This modular structure makes it easy to add:

1. Vector Search: Replace keyword retrieval with embeddings
2. FAISS Integration: Add similarity search at scale
3. Multiple Models: Support for different Ollama models
4. Better Chunking: Overlapping chunks, semantic splitting
5. Caching: Cache frequent queries
6. Sync: Cloud sync for vault files

Troubleshooting

Ollama not running:

ollama serve

Port already in use:

# Find and kill the process on port 3000
netstat -ano | findstr :3000
taskkill /PID <pid> /F

No notes found:

• Add .md or .txt files to the vault/ directory
• Call POST /api/reindex to trigger indexing

License

MIT