Universal AI Bridge for Chrome

Transform Chrome's AI APIs to work with ANY language model - from local LLMs to Chrome's built-in Gemini Nano

🚀 Install Extension (Coming Soon) 📦 View on GitHub

🚀 The Problem We Solve

Chrome's new AI APIs are powerful but limited to Chrome's built-in models. Meanwhile, developers want to use their preferred local LLMs (Llama, Mistral, Gemma) or AI providers without rewriting their applications.

loooopback bridges this gap - it's a universal translator that makes ANY language model work seamlessly with Chrome's AI API specification. Write once, run with any model.

🔌
Universal Compatibility
  • 100% Chrome AI API compatible
  • Supports old & new API formats
  • Zero code changes needed
  • Seamless provider fallback
🤖
Multiple AI Providers
  • Chrome's Gemini Nano
  • Ollama (100+ models)
  • LM Studio
  • Custom endpoints
🎯
Smart Token Management
  • Real-time token counting
  • Dynamic context detection
  • Accurate usage tracking
  • Overflow prevention
⚡
Performance Optimized
  • Model keep-alive
  • Efficient streaming
  • Smart caching
  • Parallel processing
🔒
Privacy First
  • 100% local processing
  • No telemetry or tracking
  • No account required
  • Open source code
🛠️
Developer Friendly
  • Debug console
  • Performance metrics
  • Test suite included
  • Detailed error messages

📋 How It Works

1

Install the extension

2

Choose your AI provider

3

Select your model

4

Use Chrome AI APIs anywhere

That's it! No configuration files, no API keys, no complex setup.

🎮 Supported Providers

Chrome Gemini Nano
  • No setup required
  • Works offline
  • Fast responses
  • Token tracking
Ollama
  • 100+ models supported
  • Run multiple models
  • Full control
  • Extensive customization
LM Studio
  • User-friendly GUI
  • One-click downloads
  • Model browser
  • Auto optimization
Custom Endpoints
  • OpenAI-compatible
  • Local or remote
  • Custom headers
  • Full control

💻 Full API Support

Chrome AI Methods

✅ ai.languageModel.capabilities()
✅ ai.languageModel.create()
✅ session.prompt()
✅ session.promptStreaming()
✅ session.clone()
✅ session.destroy()

📊 Real-World Performance

~50ms
Gemini Nano Latency
~200ms
Local LLM Latency
50-100
Tokens/Second
~5MB
Memory Footprint

💡 Why Choose loooopback?

✓ ONE EXTENSION, ALL MODELS
✓ ZERO CONFIGURATION
✓ TRULY LOCAL
✓ ACTIVELY MAINTAINED
✓ OPEN SOURCE
✓ NO VENDOR LOCK-IN
✓ PRODUCTION READY

Stop being limited by single providers. With loooopback, you have the freedom to use any AI model you want, how you want, where you want.