Personal LLM 2.0

A fully offline private AI assistant that loads open-source Large Language Models directly into your PC's memory. Features 28 pre-configured GGUF models (Phi-4, DeepSeek-R1, Llama 3.2, Qwen 3, CodeLlama, etc.), multi-turn chat with persistent JSON history, RAG document Q&A via ChromaDB + sentence-transformers, 4-layer Context Intelligence Engine (RAG → Recursive Decomposition → Self-Refine → Chain-of-Thought), cloud proxy for mixing local and cloud AI, and a hot model swap system. Full-stack monorepo spanning Python (FastAPI backend), React (Electron desktop), React Native (Expo Android app), and Next.js (marketing site). All data stays on your machine — zero telemetry, zero analytics, zero cloud.