lemonade-sdk/lemonade v8.0.3

on GitHub

Headline

• Support for large (sharded) GGUF models. Added Llama-4-Scout-17B-16E-Instruct-GGUF to server models list @danielholanda.

Additional Improvements

• Overhauled the pip install extras (no breaking change yet, old extras still work): dev, oga-cpu, and oga-hybrid are the primary extras now. See https://lemonade-server.ai/install_options.html for details (@jeremyfowers).
• Added LLM Chat button to the system tray context menu, and Model Management button now directly opens the model manager @jeremyfowers
• OGA v0.8.2 is now used for CPU and CUDA backends @ramkrishna2910
• Hugging Face Hub xet protocol is enabled @jeremyfowers
• Server testing improvements @jeremyfowers
• Add a developer setup button to the website by @jeremyfowers in #53

Fixes

• Improved compatibility with user-specified GGUF files @danielholanda
• Fix a crash on first server usage when installing from source @danielholanda
• Fixed Phi-3.5-Mini-Instruct-Hybrid/CPU support @jeremyfowers
• Removed unintentional "Running lemonade-server-dev..." printout @danielholanda
• Fixed UnicodeDecodeError that sometimes shows up for GGUF models @jeremyfowers

Full Changelog: v8.0.2...v8.0.3