Skip to content
All projects
AI & Machine LearningOpen source

WaifuChat (Local AI)

A personal AI assistant that uses local LLM models (DeepSeek), requires no internet connection, and runs entirely on-device.

Source code

Tech stack

Languages
Frameworks & libraries
Tools

Performance & metrics

  • Hardware-accelerated local inference
  • Zero-latency IPC between Rust backend and UI
  • Minimal RAM footprint vs. Electron

Why I built it

Cloud-based AI models like OpenAI or Claude constantly collect user data, monitor conversations, and impose heavy censorship. I wanted a personal, 100% private AI companion that runs strictly on my own hardware, accessible even without an internet connection.

Architecture & decisions

Built on top of Ollama to efficiently run DeepSeek-R1 models locally. The backend logic is powered by Rust to ensure zero-latency communication with the local AI server. The frontend is a cross-platform Tauri application, ensuring it uses minimal RAM compared to typical web wrappers.

Key features

  1. 01100% Offline execution (Zero cloud dependency)
  2. 02Total privacy with zero data telemetry
  3. 03Seamless integration with DeepSeek-R1
  4. 04Hardware-accelerated local inference
  5. 05Customizable system prompts and personas