A local AI engine in Rust: chat, vision, speech and embeddings on your own hardware, with the placement on GPU, RAM and CPU decided for you

[!WARNING] Pre-release. v0.1.0 is the first public build. Expect breaking changes in settings, APIs and file layout until the stable release.

About Cluaiz

Cluaiz is one engine process that runs local models and serves them to apps, scripts and agents:

• GGUF models through a fork of llama.cpp linked into the engine: chat, vision, audio input, embeddings.
• ONNX models through ONNX Runtime: text-to-speech (Kokoro and others), speech-to-text (Whisper), embeddings and vision encoders.
• A placement planner that measures the free GPU memory and RAM on every load and chooses where the model runs, how large the context is and which KV cache type fits. You can override any of it.
• MoE models larger than the GPU, and larger than free RAM: attention stays on the GPU, the experts sit in RAM, and whatever does not fit RAM is read from the SSD when it is used.
• APIs that existing clients already speak (OpenAI-compatible and Ollama-style routes), plus CEP, one…