Skip to content

v1.2.0 - Native Embedding Support

Choose a tag to compare

View all tags
@zeekay zeekay released this 21 Nov 21:10
· 5929 commits to main since this release
v1.2.0
fc63fdc

🎉 Major Release: Native Embedding Support

This release introduces native embedding model support through the Hanzo Engine integration, enabling local inference without external API dependencies.

✨ New Features

  • Native Embedding Models: Integrated Hanzo Engine for local inference
  • Qwen3 Embedding Models: Added support for state-of-the-art Qwen3 models (8B, 4B, 0.6B)
  • Qwen3 Reranker Models: Implemented reranking for improved search quality
  • GGUF Format Support: Efficient model deployment with GGUF format
  • GPU Acceleration: Full GPU support for faster embeddings

🛠️ Improvements

  • CI/CD Pipeline: Added GitHub Actions workflow for automated testing
  • Makefile: Created simplified build and development commands
  • Test Fixes: Resolved compilation errors and unreachable patterns
  • Post-Quantum Ready: Infrastructure prepared for post-quantum cryptography

🔧 Technical Details

  • Default embedding model: Qwen3-8B (4096 dimensions, #1 MTEB multilingual)
  • Support for 32K context windows
  • Native inference without Ollama dependency
  • Automatic model downloading and caching

📦 Installation

# Build from source
make build

# Run locally
make run

⚠️ Known Issues

  • CI tests are still being stabilized
  • Some test framework imports need adjustment

📝 Notes

The node compiles and runs successfully locally. CI infrastructure is being improved for automated testing.

Assets 3
Loading