maid

Why we included this project

Chatting with a large language model on Android doesn't have to mean a subscription. Maid runs GGUF models locally through llama.cpp, so conversations stay on your device and offline, and you can also connect your own API keys for Anthropic, OpenAI, DeepSeek, Ollama, Mistral, or Novita when you'd rather use a cloud model. It includes a one-tap downloader for curated Hugging Face models like Qwen and Phi, accepts any GGUF file you have, and lets you adjust generation parameters and set per-conversation system prompts. A companion app adds text-to-speech. Licensed under MIT, with no telemetry and no ads, it's a solid base for building a private mobile chat client or for seeing how a React Native app talks to llama.cpp.