LLocalSearch: Local Search Aggregator with LLM Agents for Private Information Retrieval

Summary

LLocalSearch is a unique, completely local search aggregator that leverages LLM Agents to find answers to user questions without needing external API keys. It offers enhanced privacy and runs efficiently on low-end hardware, providing live logs for transparency. Users can observe the LLM agents' progress towards a comprehensive, locally-generated answer.

Use at your own risk

OSRepos shares public repositories for knowledge and discovery only. Any installation, execution, configuration, or use of code from these repositories is the user's own responsibility. Always review the repository, source code, dependencies, licenses, and security implications before running or installing anything. OSRepos is not responsible for issues, damages, or losses resulting from third-party repositories.

Introduction

LLocalSearch is an innovative, completely locally running search aggregator that utilizes Large Language Model (LLM) Agents. This project allows users to ask questions, and the system employs a chain of LLMs to find answers by choosing from a set of tools to search the internet for current information. The process is recursive, meaning the LLM can freely use tools multiple times based on the information it gathers. A key advantage is that no OpenAI or Google API keys are needed, ensuring a privacy-respecting experience.

Important Note: As per the project's README, this version has not been under active development for over a year. The creator is working on a rewrite/relaunch within a private beta. Interested users are encouraged to contact the owner to join the beta.

Installation

To get LLocalSearch up and running using Docker, follow these steps:

Clone the GitHub Repository:

git clone https://github.com/nilsherzig/LLocalSearch.git
cd LLocalSearch

Create and Edit an .env file (Optional):

This step is only necessary if you need to change default settings, such as if Ollama is running on a different device.
```
touch .env
# Open with your preferred editor, e.g., code .env or nvim .env
```
Refer to the project's Ollama Setup Guide if you encounter issues with the Ollama connection.
Run the Containers:
```
docker-compose up -d
```

Examples

LLocalSearch operates by allowing users to ask a question, after which the system's LLM agents begin their search process. The project includes a demo video (demo.webm) that illustrates this in action. Users can observe live logs and links, gaining insight into what the agent is doing and the information sources for its answer. This transparency provides an excellent starting point for deeper research.

Why Use LLocalSearch?

LLocalSearch offers several compelling reasons for its use, especially for those concerned with privacy and control over their information sources:

Completely Local and Privacy-Respecting: It runs entirely on your local machine, eliminating the need for external API keys and significantly enhancing user privacy.
Runs on "Low-End" Hardware: The system is designed to be efficient, capable of running on modest hardware, with the demo video showcasing it on a 300€ GPU.
Transparent Agent Activity: Live logs and links in the answer provide a clear understanding of the agent's thought process and the basis of its findings, aiding in further research.
Supports Follow-Up Questions: The system is designed to handle follow-up inquiries, allowing for more dynamic and in-depth conversations.
User-Friendly Design: It features a mobile-friendly design and supports both dark and light modes for a comfortable user experience.
Avoids Algorithmic Manipulation: In contrast to models that might prioritize content from paying partners, LLocalSearch offers a less discriminatory alternative, allowing users to avoid potential manipulation by highest bidders.