Release Date: May 29, 2026
AudioMuse AI v2.1.1 release introduces several improvements across the codebase, including major code cleanup and refactoring, the addition of centralized error handling, and expanded unit and integration test coverage.
For users, we also improved the index rebuild process. The new approach is more memory-efficient and provides better stability on low-power hardware.
Additionally, we tested lower values for the following parameters:
VOYAGER_EF_CONSTRUCTION=200 # previously 1024
VOYAGER_M=32 # previously 64Our tests showed virtually no impact on quality (around 1% recall difference), while significantly improving indexing performance.
On an Intel i7 14th Gen machine with a library of 170k+ songs, full index rebuild time dropped from approximately 15 minutes to just 5 minutes.
We recommend updating these values through the setup wizard, and run a 1 album analysis to recreate the index, for improved performance and lower resource usage. New deployment will automatically come with this new value.
What's Changed
- Document Atlas Cloud OpenAI-compatible setup by @SpencerJung in #573
- Fix rebuild efficency by @NeptuneHub in #586
- Code Cleaning and test improvement by @NeptuneHub in #588
- Error handler by @NeptuneHub in #589
New Contributors
- @SpencerJung made their first contribution in #573
Full Changelog: v2.1.0...v2.1.1