Headline
- Lemonade Server has been completely rewritten in C++ with significant gains for performance and user experience.
- The new C++ server is interface compatible with the old Python server.
- .deb and .exe installers are available on the Release page for Ubuntu and Windows, respectively.
- The old Python implementation is deprecated, but is still available on PyPI for a limited time.
What's Changed
- C++: custom llamacpp args; env vars to override cli defaults by @jeremyfowers in #547
- C++: Users can change the llamacpp build without rebuilding lemonade by @jeremyfowers in #548
- C++: clean up the --host arg by @jeremyfowers in #551
- Fix: use the model's commit hash in cache path by @jeremyfowers in #557
- Support enable_thinking arg in chat/completions by @jeremyfowers in #558
- C++: General Availability by @jeremyfowers in #549
Full Changelog: v8.2.2...v9.0.2