●Change8

1.3.5

📦 chromaView on GitHub →
✨ 11 features🐛 8 fixes🔧 3 symbols

Summary

Version 1.3.5 introduces several new embedding functions (Nomic, Google GenAI, transformers.js) and enhances garbage collection and schema validation. This release also includes numerous bug fixes across Python and JavaScript clients.

✨ New Features

  • Implemented garbage collection for soft deleted attached functions.
  • Added auto-loading of EF (Embedding Function) package if installed.
  • Added support for non-prefixed environment variables for EF in Python.
  • Added keepalive and max connections settings to the Python client.
  • Added Nomic embedding function support.
  • De-duplicated block path prefixes during Garbage Collection to reduce memory usage.
  • Added validation to schemas for EF.
  • Added Google GenAI embedding function support.
  • Added validation on the JS client: if a sparse source key is provided, an EF must also be provided.
  • Added transformers.js embedding function, maintaining compatibility with Python.
  • Added GCS client support to storage.

🐛 Bug Fixes

  • Fixed an issue where offset was logged even when records were empty.
  • Fixed Rust doc tests and added 'cargo test --doc' to CI.
  • Fixed an issue where the schema EF was used when trying to resolve the overall JavaScript EF, removing an exception on the default EF.
  • Fixed an issue where the TypeScript vector config source key accepted 'Key' incorrectly.
  • Fixed sparse index not sorting offset ID correctly.
  • Updated common JS-EF package for Turbopack support.
  • Fixed auto-loading of default-embed when using a schema.
  • Disallowed invalid fields in Pydantic schema.

🔧 Affected Symbols

schemadefault_embeddefault_knn_index