Download
OpenJarvis is an open-source framework designed to build personal AI agents that run primarily on local devices rather than relying on cloud infrastructure. Developed as part of the Intelligence Per Watt research initiative, it focuses on improving the efficiency and practicality of on-device AI systems. The framework provides shared primitives for building local-first agents, along with evaluation tools that measure performance using metrics such as energy consumption, latency, cost, and accuracy. OpenJarvis integrates with local inference engines like Ollama, vLLM, SGLang, and llama.cpp to run language models directly on personal hardware. It also includes a learning loop that allows models to improve over time using locally generated interaction traces. By prioritizing local execution and efficiency, OpenJarvis aims to provide a foundation for privacy-preserving personal AI assistants.
Features
- Local-first framework for building personal AI agents that run primarily on-device.
- Integrates with local inference backends such as Ollama, vLLM, SGLang, and llama.cpp.
- Evaluates AI performance using metrics like energy usage, latency, cost, and accuracy.
- Includes a learning loop that improves models using locally collected interaction data.
- Provides shared primitives and tools for building and managing local AI agent workflows.
- Supports optional server components and hardware detection for optimized deployment.
