Starstruk Logo MarkStarstruk

v0.25.0

TabbyML/tabbyv0.25.0Feb 20, 2025by github-actions[bot]

AI Summary

A significant release introducing increased parallelism defaults, a new legacy embedding type, and exposing the Answer Engine's thinking process to users.

Key Highlights

  • Default parallelism increased from 1 to 4 (may increase VRAM usage)
  • Introduce new embedding kind for legacy llama.cpp API
  • Expose Answer Engine thinking process in thread messages

Breaking Changes

  • Default parallelism increased to 4

New Features

  • Exposing Answer Engine thinking process
  • Directory file list access for Answer Engine
  • Mention symbols using '@' in Chat Sidebar
  • Repository-aware default question recommendations

Full Release Notes

## ⚠️ Notice

Significant changes have been implemented in this release; please consider adjusting them to fit your specific use case.

* The default parallelism has been increased from 1 to 4, which might increase VRAM usage. ([#3832](https://github.com/TabbyML/tabby/pull/3832))
* Introduce a new embedding kind `llama.cpp/before_b4356_embedding` for llamafile or other embedding services utilizing the legacy llama.cpp embedding API. ([#3828](https://github.com/TabbyML/tabby/pull/3828))

## 🚀 Features

* Expose thinking process of Answer Engine to the answers in thread message. ([#3785](https://github.com/TabbyML/tabby/pull/3785)) ([#3672](https://github.com/TabbyML/tabby/pull/3672))
* Enable the Answer Engine to access the repository's directory file list as needed.  ([#3796](https://github.com/TabbyML/tabby/pull/3796))
* Enable the use of `@` to mention a symbol in Chat Sidebar. ([#3778](https://github.com/TabbyML/tabby/pull/3778))
* Provide default question recommendations that are repository-aware on Answer Engine. ([#3815](https://github.com/TabbyML/tabby/pull/3815))

## 🧰 Fixed and Improvements

* Provide a configuration to truncate text content prior to dispatching it to embedding service.. ([#3816](https://github.com/TabbyML/tabby/pull/3816))
* Bump llama.cpp version to b4651. ([#3798](https://github.com/TabbyML/tabby/pull/3798))
* Automatically retry embedding when the service occasionally fails due to issues with llama.cpp. ([#3805](https://github.com/TabbyML/tabby/pull/3805))
* Enhance the user interface experience for Answer Engine. ([#3845](https://github.com/TabbyML/tabby/pull/3845)) ([#3794](https://github.com/TabbyML/tabby/pull/3794))
* Resolve the deserialization issue related to `finish_reason` in chat response from LiteLLM Proxy Server.([#3882](https://github.com/TabbyML/tabby/pull/3882))

## 💫 New Contributors

@zhanba made their first contribution in #3675
@faceCutWall made their first contribution in #3812

**Full Changelog**: https://github.com/TabbyML/tabby/compare/v0.24.0...v0.25.0