Bookmarks by Localforge

A browser extension for Chrome and Firefox that captures web pages as bookmarks, extracts readable content, and enables semantic search using RAG (Retrieval-Augmented Generation). All data is stored locally in your browser.

Download for Firefox or Chrome

Screenshots

Popup	Library	Search

Features

• One-click capture — Save URL, title, and full DOM HTML with a click or keyboard shortcut
• Content extraction — Converts HTML to Markdown using Mozilla's Readability
• Q&A generation — LLM generates question-answer pairs for each bookmark
• Semantic search — Search bookmarks by meaning using embeddings
• Tag organization — Flat organization with tags; click to filter, type to create
• Stumble mode — Randomly surface bookmarks you may have forgotten
• Health indicator — Visual status shows processing state; click for diagnostics
• Jobs dashboard — Monitor processing jobs with status and progress
• Bulk URL import — Import multiple bookmarks at once with validation
• Import/Export — Backup and restore bookmarks as JSON files
• WebDAV sync — Sync bookmarks across devices using your own WebDAV server
• Configurable API — Use OpenAI or any compatible endpoint (local models included)

Installation

Prerequisites

• Node.js 18+ and npm
• Chrome or Firefox browser
• OpenAI API key (or compatible API endpoint)

Build Steps

1. Install dependencies:

   npm install

2. Build the extension:

   npm run build

   This creates a dist/ folder with the built extension.

Load in Chrome

1. Open Chrome and go to chrome://extensions/
2. Enable "Developer mode" in the top right
3. Click "Load unpacked"
4. Select the dist folder

Load in Firefox

1. Open Firefox and go to about:debugging#/runtime/this-firefox
2. Click "Load Temporary Add-on"
3. Navigate to the dist folder and select manifest.json

Web Version

A standalone web version is available for testing; but it's of limited use because of the lack of generalized cross-origin requests.

npm run dev:web

Configuration

1. Click the extension icon and select "Settings"
2. Configure your API settings:
   • API Base URL: Default is https://api.openai.com/v1
   • API Key: Your OpenAI API key
   • Chat Model: Model for Q&A generation (e.g., gpt-4o-mini)
   • Embedding Model: Model for embeddings (e.g., text-embedding-3-small)
3. Click "Test Connection" to verify your settings
4. Click "Save Settings"

Using Local Models

The extension works with any OpenAI-compatible API server (vLLM, llama.cpp, etc.):

1. Set API Base URL to your local endpoint (e.g., http://localhost:1234/v1)
2. Set API Key to any non-empty string (many local servers don't require a key)
3. Set model names that your local server supports

Usage

Saving Bookmarks

Method 1: Extension Icon

1. Navigate to any web page
2. Click the extension icon
3. Click "Save This Page"

Method 2: Keyboard Shortcut

• Windows/Linux: Ctrl+Shift+B
• Mac: Cmd+Shift+B

Browsing Bookmarks

1. Click the extension icon
2. Click "Explore Bookmarks"
3. View bookmarks with their processing status:
   • Pending: Waiting to be processed
   • Processing: Currently being processed
   • Complete: Ready to search
   • Error: Processing failed (can retry)

Searching Bookmarks

1. Open the Search view
2. Enter a search query
3. Results are ranked by semantic similarity
4. Click any result to view the full bookmark details

Using Tags

• Click any tag pill to filter bookmarks by that tag
• In bookmark details, type in the tag input to add new tags
• Tags are flat (no hierarchy) and use lowercase with hyphens

Stumble Mode

1. Open the Stumble view
2. See 10 random bookmarks with a Q&A preview
3. Click "Shuffle" for a new random selection
4. Filter by tags to stumble within a topic

Bulk Importing URLs

1. Open Settings
2. Scroll to "Bulk Import URLs"
3. Paste a list of URLs (one per line)
4. Click "Import URLs"
5. Monitor progress in the Jobs dashboard

WebDAV Sync

1. Open Settings
2. Enable WebDAV sync
3. Enter your WebDAV server URL, username, and password
4. Set a sync path (default: /bookmarks)
5. Sync happens automatically when bookmarks change

Import/Export Bookmarks

Export:

1. Open Settings
2. Click "Export All Bookmarks"
3. A JSON file will be downloaded

Import:

1. Open Settings
2. Click "Choose File to Import"
3. Select a previously exported JSON file
4. Duplicate URLs are skipped automatically

How It Works

Processing Pipeline

When you save a bookmark:

1. Capture: Full page HTML is captured and saved locally
2. Extract: Readability extracts the main content and converts to Markdown
3. Generate Q&A: LLM generates 5-10 question-answer pairs about the content
4. Embed: Each Q&A pair is converted to embeddings (question, answer, and combined)
5. Index: Everything is stored in IndexedDB

Search Algorithm

When you search:

1. Your query is converted to an embedding
2. Cosine similarity is computed against all stored Q&A embeddings
3. Results are ranked by similarity score
4. Top results are grouped by bookmark and displayed

Data Storage

• All bookmarks are stored locally in your browser's IndexedDB
• Only the extracted Markdown content is sent to your configured API for processing
• No data is sent to any third-party servers (except your configured LLM API)
• WebDAV sync sends bookmark data to your own server
• Export your data anytime as JSON files

Development

# Development mode (Chrome)
npm run dev:chrome

# Development mode (Firefox)
npm run dev:firefox

# Development mode (Web)
npm run dev:web

# Run unit tests
npm run test:unit

# Run all tests
npm run test:all

Security

Send us an email with the title of this section at localforge.org. We don't have a GPG key configured, but will monitor and answer fast (as of late 2025).

Here are some attack vectors we considered from web content pages:

• XSS via bookmark title: use textContent always when displaying title
• XSS via bookmark content: use lib DOMPurify to sanitize
• URL injection: only allow http/https protocols
• Content script injection: captured HTML goes through Readability + DOMPurify
• Stored XSS via IndexedDB: unlikely, but avoid innerHTML at all cost and always try to use DOM libs, even for already stored contend, and never display captured HTML

License

MIT

Credits

Built using:

• Dexie.js for IndexedDB
• @mozilla/readability for content extraction
• Turndown for HTML to Markdown conversion
• Vite and @crxjs/vite-plugin for building