v1.0.0 - Stable Oasis
🎉 First Stable Release - Production Ready!
✨ What's New
This is the first stable release of vLLM HuggingFace Bridge, a fork of the original HuggingFace VS Code extension optimized for enterprise and air-gapped deployments.
🚀 Key Features
- Air-Gapped Operation: Complete offline functionality with local vLLM/TGI servers
- Dual Mode Support: Seamlessly switch between local and HuggingFace cloud models
- Enterprise Ready: Production-tested in secure environments
- Optimized Package: 92KB (91% smaller than original)
- Smart Token Management: Automatic allocation for small context models (2048+)
- Enhanced Observability: Clear logging showing which model handles each request
📦 Installation
Quick Install (Recommended)
- Download the VSIX file below
- In VS Code:
Ctrl+Shift+P→Extensions: Install from VSIX... - Select the downloaded file and restart VS Code
Command Line
```bash
wget https://github.com/dzivkovi/vllm-huggingface-bridge/releases/download/v1.0.0/vllm-huggingface-bridge-1.0.0.vsix
code --install-extension vllm-huggingface-bridge-1.0.0.vsix
```
🔧 Configuration
For Local vLLM/TGI Servers
```json
{
"huggingface.localEndpoint": "http://localhost:8000"
}
```
For HuggingFace Cloud
- Open GitHub Copilot Chat
- Click model picker → "Manage Models..."
- Select "Hugging Face" provider
- Add your HF token
🧪 Tested With
- VS Code 1.104+
- vLLM 0.5+
- TGI 2.0+
- Various GPUs: RTX 4060, A100, H100
- Models: DeepSeek-Coder, Qwen, Mistral, Llama
📊 Technical Improvements
- All 42 tests passing
- ESLint configured for pragmatic development
- Package size reduced from 954KB to 92KB
- Enhanced logging for debugging
- Fixed VS Code API compatibility issues
🙏 Credits
Built on the excellent foundation of HuggingFace VS Code Chat.
📝 License
MIT License - See LICENSE.txt for details