What's Changed
- Cancel generation when client disconnects mid-request by @michaelandrewgamble in #117
- Pass through sampling params the engine already supports by @michaelandrewgamble in #119
- Serve /v1/models off the event loop with bounded directory scans by @michaelandrewgamble in #121
- Add prompt/KV-cache reuse for the chat completions path by @michaelandrewgamble in #123
- Update MLX audio support and bump version to 2.3.0 by @sxy-trans-n in #125
- Update packaged dependency bundles for Swama 2.3.0 by @sxy-trans-n in #126
New Contributors
- @michaelandrewgamble made their first contribution in #117
Full Changelog: v2.2.0...v2.3.0
Contributors
michaelandrewgamble and sxy-trans-n