v4.7.7

Tencent-Hunyuan/HY-Motion-1.0v4.7.7Sep 16, 2026by github-actions[bot]

Upgrading?

AI Summary

This release focuses on fixing authentication flows, chat model output limits, and specific behaviors for Claude 5 and adaptive-thinking models. It also updates build configurations for the AMD64 model server.

Key Highlights

  • Fixed OAuth link forwarding for pass-through tools
  • Respects model limits when setting output allowances
  • Sends output_config effort on Claude 5 turns after a tool call
  • Honors 'reasoning off' settings on adaptive-thinking Claude models
  • Increased model-server build volume to 60GB for AMD64

Full Release Notes

See the assets to download this version and install.

## What's Changed
* fix(auth): forward the live OAuth link to pass-through tools, not the first row (#14756) to release v4.7 by @onyx-cherry-pick[bot] in https://github.com/onyx-dot-app/onyx/pull/14796
* fix(chat): respect model limits when setting output allowances (#14605) to release v4.7 by @nmgarza5 in https://github.com/onyx-dot-app/onyx/pull/14808
* ci: raise the amd64 model-server build volume to 60gb (#14797) to release v4.7 by @onyx-cherry-pick[bot] in https://github.com/onyx-dot-app/onyx/pull/14814
* fix(chat): point the out-of-context file notice at the python tool (#14792) to release v4.7 by @onyx-cherry-pick[bot] in https://github.com/onyx-dot-app/onyx/pull/14828
* fix(llm): send output_config effort on Claude 5 turns after a tool call [mirror of #14062 for edits] (#14812) to release v4.7 by @nmgarza5 in https://github.com/onyx-dot-app/onyx/pull/14837
* fix(llm): honor reasoning off on adaptive-thinking Claude models (#14839) to release v4.7 by @nmgarza5 in https://github.com/onyx-dot-app/onyx/pull/14844


**Full Changelog**: https://github.com/onyx-dot-app/onyx/compare/v4.7.6...v4.7.7

Related Repositories