# lilbee: Run Local AI and Search Your Files

This repository profile is provided by osrepos.com, an open source repository discovery platform.

Source: osrepos.com
Repository profile: https://osrepos.com/repo/tobocop2-lilbee
Generated for open source discovery and AI-assisted research.

lilbee is a local AI model manager and search engine for files, code, and crawled websites. It offers cited answers through a terminal app, CLI, MCP server, REST API, and Python library.

GitHub: https://github.com/tobocop2/lilbee
OSRepos URL: https://osrepos.com/repo/tobocop2-lilbee

## Summary

lilbee is a local AI model manager and search engine for files, code, and crawled websites. It offers cited answers through a terminal app, CLI, MCP server, REST API, and Python library.

## Topics

- python
- local-first
- rag
- mcp-server
- local-llm
- semantic-search

## Repository Information

Last analyzed by OSRepos: Tue Oct 06 2026 21:42:21 GMT+0100 (Western European Summer Time)
Detail views: 2
GitHub clicks: 0

## Safety Notice

OSRepos shares public repositories for knowledge and discovery only. Review source code, dependencies, licenses, and security implications before running or installing anything.

## Content

## Overview

lilbee combines a local model runner with a search engine for your files, notes, code, and crawled web pages. It indexes content and uses models to answer questions with citations that point back to source files and lines.

It is aimed at people who want a self-hosted alternative to cloud-first search and retrieval, and at coding agents that need to look up project material. Models and indexes can stay on your machine; a cloud provider is optional. You can use lilbee through a terminal interface, CLI, MCP server, REST API, or Python library.

## Key Features

- Manages chat, embedding, vision, and reranking models, with support for local GPU backends and multi-GPU placement.
- Indexes documents and code, with citations linking answers to their sources.
- Crawls websites and adds their pages to a searchable local library.
- Exposes search and indexing to MCP-aware coding agents.
- Provides a terminal UI, CLI, REST API, and Python library.
- Can use existing Ollama or LM Studio setups, as well as optional hosted models.

## Use Cases

- Researchers can index papers, manuals, and notes, then ask questions that link back to source passages.
- Developers can index a codebase and its documentation so a coding agent can retrieve relevant code with file and line citations.
- Teams or individuals can crawl documentation sites for a local, searchable copy that remains useful offline.
- Privacy-conscious users can run models and search personal files locally without relying on a cloud provider.

## Project Facts

- Language: Python
- License: MIT
- Stars: 62
- Forks: 7
- Topics: ai-search-engine, chat-with-documents, gguf, llama-cpp, lm-studio, local-ai, local-first, local-llm, local-rag, mcp-server, model-manager, multi-gpu, notebooklm-alternative, ollama, privacy, private-ai, rag, self-hosted, semantic-search, tui
- Archived: No

## Getting Started

The README recommends installing a bundled build, then running:

```bash
lilbee self-check
lilbee
```

See the [README](https://github.com/tobocop2/lilbee#install) for installation options, hardware-specific builds, and configuration. The [project site](https://lilbee.sh/) also links to tutorials and the [REST API reference](https://lilbee.sh/api/).

## Alternatives

- [SurfSense](https://osrepos.com/repo/modsetter-surfsense): SurfSense focuses on desktop document research and study materials, while lilbee also manages local models and searches code and crawled websites through multiple interfaces.

## Considerations

- The project describes itself as active beta software. Its interfaces, command names, and on-disk formats may change between pre-releases.
- A GPU is not required, but model performance and which models fit depend on available hardware and memory.
- Bundled builds are large, and the first launch has a one-time unpacking step. Developer installs through pip or uv require choosing and installing the separate model-engine extra for the relevant hardware.
- Local operation is the default, but using a hosted model sends questions and relevant retrieved snippets to the provider you configure.