Skip to content

Release v0.6.4

Latest
Latest

Choose a tag to compare

View all tags
@github-actions github-actions released this 02 Sep 22:22
· 4 commits to main since this release
Immutable release. Only release title and notes can be modified.
v0.6.4
478ec00
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v0.6.4 Release Notes

This release focuses on fixing critical streaming issues with Responses API models, improving model routing for new model families, enhancing cache performance and request latency, and updating CI/CD dependencies. The proxy now correctly handles function calling in streaming responses for models like muse-spark, grok, and gpt-5 variants, and uses prefix-based model classification to automatically route current and future models to their correct endpoints.

Bug Fixes

  • Fix function_call handling in Responses API streaming for models like muse-spark-1.2-contributor, grok-4.6, and gpt-5.6-luna — streaming requests now properly emit tool_use blocks instead of dropping them and returning text + end_turn prematurely
  • Switch model endpoint classification from hardcoded model IDs to prefix matching for Responses and Gemini families — newer models (grok-, muse-spark-, gemini-3.7-flash, gpt-5.6-*) now route to correct endpoints automatically
  • Accept all known providers (including openrouter, aws-bedrock) in model_overrides and model_family_overrides config — validation now uses canonical provider list from internal/config to prevent drift
  • Fix update command to stage downloads in install directory — prevents cross-filesystem rename failures (EXDEV) on tmpfs mounts and surfaces permission errors before downloading

Improvements

  • Improve cache hit rates and reduce request latency with bounded latency and TTFT (Time To First Token) metrics
  • Preserve legacy tool result output format for backward compatibility

Documentation

  • Remove stale ui command reference; document actual start and start -b commands with dashboard URL (http://127.0.0.1:3445)
  • Clarify endpoint family classification in howto-add-model.md — note that adding models may require code changes for new families

Chores

  • Bump docker/metadata-action from 5.10.0 to 6.2.0
  • Bump modernc.org/sqlite from 1.56.0 to 1.57.0
  • Bump docker/login-action from 3.7.0 to 4.6.0
  • Bump actions/checkout from v4 to v7
  • Bump signpath/github-action-submit-signing-request from 1.3 to 2.3
  • Bump docker/setup-qemu-action from 3.7.0 to 4.2.0
  • Bump actions/upload-artifact from 4.6.2 to 7.0.1
  • Bump actions/download-artifact from 4.3.0 to 8.0.1
Assets 13
Loading