llm-chat

confidential LLM inference via wasi-nn, inside a wasm sandbox model target
Answer quality tracks model size - small models are quick and frequently wrong. Chat history stays in this browser tab.