Skip to content

V0.12.7

Version 0.12.7

This release introduces new features enhancing documentation and example interfaces, improves the configurability of embedding models in LanceDBDocStore, and includes several bug fixes and refinements in documentation and codebase.

New Features

  • Added a new example script demonstrating how to build an interactive chat interface using LlamaBot and Marimo, complete with a grid layout JSON file for the UI. (133f12e) (Eric Ma)
  • Introduced configurable embedding model settings in LanceDBDocStore, allowing users to specify embedding registries and models, enhancing customization. (3356016) (Eric Ma)

Bug Fixes

  • Fixed the initialization of the Vector field in LanceDBDocStore to use the correct dimensions, ensuring proper setup. (3595285) (Eric Ma)
  • Updated the embedding function in LanceDBDocStore to use 'minishlab/potion-base-8M' for improved performance and compatibility. (56532ac) (Eric Ma)

Documentation Updates

  • Removed outdated note about GPT4 authorship from the QueryBot tutorial, ensuring clarity and accuracy in documentation. (a00798a) (Eric Ma)
  • Updated the QueryBot tutorial to recommend explicit document store and chat memory management, replacing GPT-4 specific instructions with more general LLM usage and introducing LanceDBDocStore for better management. (046286c) (Eric Ma)
  • Added comprehensive documentation for the new LlamaBot documentation chat interface example, including setup and usage instructions. (133f12e) (Eric Ma)
  • Enhanced the QueryBot tutorial to demonstrate how to configure embedding model settings in LanceDBDocStore, complete with default settings and customization options. (3356016) (Eric Ma)

Refinements

  • Simplified the embedding function initialization in LanceDBDocStore by removing redundant imports and optimizing the creation process. (2db2f46) (Eric Ma)
  • Commented out unused hypothesis strategies import and related unused code in test_docstore.py, cleaning up the test suite. (bef8924) (Eric Ma)