timmyy123/LLM-Hub

Enables on-device multi-modal AI with chat, image generation, code completion, and specialized tools (transcription, translation, OCR-based scam detection) using optimized model formats (LiteRT Task, GGUF, ONNX) that leverage GPU/NPU acceleration across multiple model families. Built with Kotlin/Jetpack Compose, it supports flexible model imports from HuggingFace and includes Kid Mode with model-level safety guardrails, RAG memory, and web search integration—all with zero data collection and 100% local inference.

196 stars.