Skip to content

Release v0.4.0

Choose a tag to compare

View all tags
@zelda2003 zelda2003 released this 02 May 19:16
· 15 commits to main since this release
v0.4.0
67eb9dd

Release v0.4.0 - Expanded Embedding Options & CPU Focus

This release significantly expands the embedding capabilities of db2vec, offering more flexibility while maintaining a strong focus on efficient CPU-based processing.

✨ New Features

  • Multiple Embedding Providers: You can now choose your embedding engine!
    • rustbert: Utilizes the built-in all-MiniLM-L6-v2 model (384 dimensions) directly on your CPU via the rust-bert library. No external service needed!
    • google: Connect to Google's Generative AI API (e.g., text-embedding-004) using your API key.
    • ollama: Continued support for using any model served locally via Ollama.

🚀 How to Use New Embeddings

Select your provider and model using the following flags:

  • --embedding-provider <PROVIDER>: Choose ollama, rustbert, or google.
  • --embedding-model <MODEL_NAME>: Specify the model name (e.g., nomic-embed-text for Ollama, text-embedding-004 for Google). This flag is ignored for rustbert as it uses the built-in model.
  • --dimension <DIM>: Crucial! Set this to match your chosen model's output dimension (e.g., 384 for rustbert, 768 for text-embedding-004).
  • --embedding-api-key <KEY>: Required only when using --embedding-provider google. Can also be set via SECRET env var.
  • --embedding-url <URL>: Primarily for Ollama, but can be used to specify a different Google API endpoint if needed.

Example (Rust-Bert):

./db2vec -f dump.sql -t qdrant --embedding-provider rustbert --dimension 384
Assets 8
Loading