Show HN: Ved AI Voice Assistant (github.com)

Ved AI Voice Assistant has made its debut as a resource-efficient AI agent designed for Windows, Linux, and macOS, capable of operating completely offline with local models, requiring only 12 GB VRAM and 32 GB RAM. It uniquely allows fallback to cloud processing through the OpenRouter API with reduced hardware requirements of 4 GB VRAM and 16 GB RAM. Ved features a user-friendly Tkinter UI and a FastAPI server that facilitates interaction through an HTTP API. The assistant emphasizes privacy, as telemetry is recorded locally without transmitting data outside the user's network, and users can opt-out if desired. This development is significant for the AI/ML community as it showcases the growing trend of creating powerful AI solutions that maintain user privacy while enabling offline functionality. The architecture supports multiple modes, including casual chat and coding assistance, through different model configurations and threads. Key technical elements include an interruptible offline voice pipeline, per-thread retrieval-augmented generation (RAG), and a human-in-the-loop approval system for executing scripts. Moreover, Ved's deterministic model switching and local RAG storage enhance its versatility, empowering users to customize workflows effectively while mitigating reliance on cloud computing resources.