Our take
Activeregular commits and releases- Activity
- 90
- Community
- 15
- Issues
- 84
- Pull requests
- 78
+6 stars in the last 30 days
- Releases more often than 93% of the projects we track
- Merges more pull requests than 83% of the projects we track
- Fewer contributors than 82% of the projects we track
lilbee offers an unusually broad local AI stack in one install, but its active development and highly concentrated contribution history make it worth evaluating before relying on it for critical workflows.
Good fit if
- You want local search with cited answers alongside model management, rather than a standalone model runner.
- You need MCP access to indexed files or code for a coding agent, and can use Python 3.11+ for a developer installation.
- You value an MIT-licensed project that is actively maintained, with frequent commits and releases and rapid issue and pull request handling.
Look elsewhere if
- Your team requires a project with a broader contributor base: most recorded contributions come from one person.
- You need a mature, stable interface and are not prepared to account for the README's beta warning and possible changes to interfaces or data formats.
- You need a hosted, multi-user service rather than a locally operated tool, or do not want to manage local models and their hardware needs.
All health signals
| Last commit | 2026-11-03 (today) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 100+ (latest v0.6.90b691, 2026-10-28) |
| Contributors | 3 (top contributor: 98% of commits) |
| Issues closed, last 90 days | 44+ (typically closed in 1 day) |
| Pull requests merged, last 90 days | 74 (typically merged in 0 days) |
| Project age | 8 months |
Checked on 2026-11-03 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
72Forks
9Change since Nov 21, 2026: 0 stars·0 forks
Overview
lilbee is a local AI model manager and conversational search engine built on llama.cpp. It runs models across available GPUs and searches files, code, notes, and crawled websites, returning answers with source citations. Its stated goal is to keep your data, index, and inference on your machine unless you choose a cloud model.
It targets macOS, Linux, and Windows, and exposes the same system through a terminal interface, CLI, MCP server, REST API, and Python library. Rather than requiring a separate model server or vector database, lilbee bundles the model runner and search stack; it can also use models served by Ollama or LM Studio.
Key Features
- Manage chat, embedding, vision, and reranking models, with placement across multiple GPUs.
- Index documents and source code, then search or ask questions with citations to source files and lines.
- Crawl websites into a local library for later search, including offline use of the indexed copy.
- Connect coding agents through MCP tools that return cited snippets from indexed material.
- Use a TUI, CLI, REST API, and Python library, with an Obsidian community plugin also available.
- Run local models through llama.cpp on Metal, Vulkan, CUDA, or ROCm builds, or connect existing Ollama and LM Studio setups.
- Process a range of document formats and languages, with OCR options for scanned pages and images.
Use Cases
- A researcher can index papers and notes, then ask questions and follow citations back to the source.
- A developer can index a repository and its documentation so an MCP-connected coding agent can retrieve relevant code with file and line references.
- A team can crawl product documentation or an API reference and search a local copy when the website is unavailable or has changed.
- A privacy-conscious individual can build a searchable library of personal documents and use local models without sending files to a cloud service.
What you need
Detected in the repository
- Python >=3.11 (from pyproject.toml)
- A test suite and automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
The README recommends installing a standalone build, then running:
lilbee self-check
lilbee
The first command checks inference and embedding; the second starts the terminal app. See the README for platform-specific installation, GPU builds, and setup details.
Alternatives
- ragtime: Ragtime focuses on a self-hosted RAG API and chat UI, while lilbee also manages local models and offers CLI, TUI, and Python interfaces.
- SocratiCode: SocratiCode specializes in codebase search and dependency analysis, while lilbee searches files, code, and crawled websites with cited answers.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| lilbee | Python | MIT | 68 | Active |
| ragtime | Python | MIT | 34 | Active |
| SocratiCode | TypeScript | AGPL-3.0 | 3.4k | Active |
Considerations
- The README describes lilbee as active beta software and warns that interfaces, command names, and on-disk formats may change. Check current release and installation instructions before adopting it.
- The project is active, but contributor activity is highly concentrated in one person. That creates a maintenance and continuity risk despite recent releases and quick issue handling.
- Local inference does not require a GPU, but performance and the size of models you can run depend on available hardware and memory. GPU-specific builds and requirements vary by platform.
- A Python installation needs Python 3.11 or newer. The standalone builds offer a separate route that bundles the runtime and model engine.
- Cloud models are optional, but using them sends questions and relevant retrieved snippets to the selected provider. The README says lilbee otherwise keeps the index and files local.
Found this useful?
Share it with someone who would like lilbee.