web@inode64.com Fco. Javier Félix Lemonade is a local AI server that exposes optimized LLMs through OpenAI, Anthropic and Ollama compatible APIs, running inference on AMD NPU and GPU. The C++ server core (lemond, lemonade) is packaged; the bundled web frontend is available behind USE=webui and the Tauri desktop wrapper behind USE=tauri. Pairs with sci-ml/fastflowlm to drive the AMD Ryzen AI XDNA2 NPU. Reuse the system sci-ml/fastflowlm (/usr/bin/flm) for the NPU backend instead of letting lemond download the flm runtime at first NPU use Reuse the system sci-misc/llama-cpp (/usr/bin/llama-server) for the llama.cpp backend instead of fetching an upstream prebuilt binary at runtime Point lemond's ROCm runtime at the system ROCm in /usr (ROCM_PATH) so its ROCm backends reuse it instead of downloading AMD's TheRock runtime; pulls dev-util/hip Build the Tauri desktop app (lemonade-app), a native webkit2gtk window for the same UI as USE=webui. Fetches Rust crates over the network at build time Build the bundled React web frontend served by lemond at the server root; runs npm over the network at build time lemonade-sdk/lemonade https://github.com/lemonade-sdk/lemonade/issues https://lemonade-server.ai/docs/