A new project called LLMNet turns local LLMs into a private, offline search engine. It uses a PostgreSQL vector database for semantic retrieval, allowing users to index websites and wikis for instant, local answers. The stack includes Next.js, Bun, and OpenAI-compatible APIs for a self-hosted experience.

This tool addresses growing privacy concerns by keeping all queries and data on your machine. It combines local generative AI with a high-performance RAG pipeline for structured search. For developers tired of cloud dependencies, it offers a sovereign alternative for building searchable knowledge bases.

The project is available on GitHub, with a quick-start guide for setup. It requires a local LLM server, embedding service, and Postgres with the vector extension. As local AI models mature, tools like LLMNet show how offline, privacy-first applications are becoming practical for everyday use.