Lemonade: Refreshingly Simple Local Chat

2026-05-15 · source · source ↗ · raw capture

Lemonade is a local AI server providing cloud-API-like capabilities (chat, vision, image generation, speech-to-text, text-to-speech) running entirely on-device. Open source (Apache 2.0), cross-platform (Windows, macOS, Linux, Docker, Snap). Two distributions: Lemonade Server (background service) and Embeddable Lemonade (portable binary for bundling into third-party apps). Plugin-style inference engine architecture supporting llama.cpp, whisper.cpp, stable-diffusion.cpp, Kokoro, FLM (XDNA2 NPU), ryzenai-llm, and experimental vLLM. OpenAI, Anthropic, and Ollama API-compatible endpoints at localhost:13305.

Four pillars: Agile (auto-deploys latest models, optimized for Ryzen AI/Radeon/Strix Halo), Portable (<10MB binary), Omni Modal (standard endpoints for all modalities), Free & Private (open source, no telemetry).

GitHub: github.com/lemonade-sdk/lemonade