nilsherzig/llocalsearch

LLocalSearch is a locally-running search tool powered by small AI language models. It lets you ask a question in plain English, and the system uses AI agents to search the internet and compile an answer, all without sending data to OpenAI, Google, or any external API service. No API keys are required. The core idea is that a language model runs on your own computer and is given access to tools, including web search. The model reads your question, decides which tools to call, searches the web, reviews the results, and may search again based on what it finds. This cycle repeats until the model has enough information to answer. The interface shows each step as it happens, including the links and sources the agent consulted, so you can trace how the answer was formed. The project was motivated by concerns about commercial AI search products being shaped by paid partnerships with publishers. LLocalSearch runs on local hardware with no external service involved, so no third party can influence which sources appear or how they are ranked. Installation uses Docker. The system connects to Ollama, a tool for running language models on your own machine. The README notes the demo was recorded on hardware with a roughly 300 euro GPU, which the author frames as accessible to mid-range hardware rather than requiring a high-end setup. As of the most recent README update, active development on this version had paused for over a year. The author is working on a rewrite in a private beta and invited interested users to reach out to participate.