SideLlama 作者: Bence Zoltan

Release Notes

Version 1.0.0 - Initial Release

Release Date: February 2026

Overview

FireLlama is a privacy-focused browser extension that connects to local and remote LLM servers, putting AI assistance directly in your browser sidebar. No data leaves your device when using local servers.

Key Features

• Local-First Privacy: Connect to local LLM servers (llama.cpp, LM Studio, Ollama, vLLM) - your data never leaves your machine
• Persistent Sidebar Chat: Always-available chat interface for asking questions about the current page
• Page Summarization: One-click summarization of any webpage
• Page Context Awareness: The AI can read and understand the content you're viewing
• Multi-Provider Support: Works with OpenAI-compatible, Anthropic, and Google Gemini APIs
• Form Filling Assistance: AI-powered help filling out web forms
• Selected Text Queries: Highlight text on any page to ask questions about it
• 24 Languages: Full UI translation support for all EU official languages plus Brazilian Portuguese
• Beautiful Themes: Light, Dark, Nord, Dracula, and Monokai color schemes

Technical Highlights

• Zero Data Collection: This extension does not collect, transmit, or store any user data
• Manifest V3: Built with modern extension standards for security and performance
• Cross-Browser: Available for Firefox and Chrome (as SideLlama)
• Open Source: Fully transparent codebase auditable by anyone

Requirements

• Firefox: 142.0 or later (also compatible with Firefox for Android)
• Local LLM Server: Any OpenAI-compatible server running on localhost (port 8080 default) OR
• Remote API: Account with supported provider (OpenAI, Anthropic, Google Gemini)

Getting Started

1. Install FireLlama from the Firefox Add-on Store
2. Open the extension settings via the toolbar icon
3. Configure your LLM provider:
4. Local: Enter your server URL (e.g., http://localhost:8080)
5. Remote: Enter your API endpoint and key
6. Start chatting from the sidebar!

Privacy & Security

• No telemetry: No usage analytics or tracking
• No account required: Works entirely offline with local models
• Sanitized rendering: All HTML content is sanitized using DOMPurify to prevent XSS attacks
• Host permissions: Only communicates with explicitly configured API endpoints

Known Limitations

• Streaming responses require a compatible API endpoint
• Very long pages (>20,000 characters) are truncated for context
• Local servers must be running before use (extension does not start servers)

Credits

• Built with vanilla JavaScript - no frameworks, no bloat
• Uses DOMPurify for safe HTML rendering
• Icon design inspired by llama mascot

Support

• Issues: Report bugs via the Firefox Add-on store or GitHub
• Source Code: Available at [your repository URL]

Full Changelog

Since this is the initial release, all features are new. Thank you for trying FireLlama!