LangManus community-driven AI automation framework

By /

LangManus is a community-driven AI automation framework that aims to combine language models with professional tools such as web search, crawlers, and Python code execution to automate complex tasks. (langmanus/README_zh.md at main · Darwin-lfl … – GitHub)

Project introduction

LangManus is built on the excellent work of the open source community, with the goal of combining language models with professional tools to handle tasks such as web search, crawling, and Python code execution. (langmanus/README_zh.md at main · Darwin-lfl … – GitHub)

The project adopts a multi-agent system architecture and mainly includes the following roles:

  1. Coordinator (Coordinator): Handle initial interactions and distribute tasks.
  2. Planner: Analyze tasks and develop execution strategies.
  3. Supervisor (Supervisor): Supervise and manage the execution of other agents.
  4. Researcher: Collect and analyze information.
  5. Coder: Responsible for code generation and modification.
  6. Browser: Perform web browsing and information retrieval.
  7. Reporter (Reporter): Generate reports and summaries of workflow results. (Darwin-lfl/langmanus – GitHub, pre-commit – Darwin-lfl/langmanus – GitHub)

🚀Main functions

  • Language model integration: Supports open source models (such as Qwen) and OpenAI-compatible API interfaces, suitable for tasks of different complexity.
  • Search and information retrieval: Web search is carried out through Tavily API, neural search is implemented with Jina, and advanced content extraction functions.
  • Python integrated: Built-in Python REPL, supports code execution environment, and performs package management through uv. (Darwin-lfl/langmanus – GitHub)

Example demonstration

One example task is to calculate the influence index of the DeepSeek R1 model on HuggingFace. LangManus’s automated processes include: (langmanus/README.md at main – GitHub)

  1. Search online for the latest information on “DeepSeek R1” and “HuggingFace”.
  2. Use the Chromium instance to visit the HuggingFace official website and retrieve the latest data of “DeepSeek R1”, such as followers, likes and downloads.
  3. Find formulas to calculate the influence of your model through search engines and web scraping.
  4. Use Python to calculate the influence index based on the data collected.
  5. Present comprehensive reports to users. (langmanus/README.md at main – GitHub)

ˇStart quickly

To start the project quickly, follow these steps: (langmanus/README.md at main – GitHub)

#Clone warehouse
git clone https://github.com/langmanus/langmanus.git
cd langmanus

#Install dependencies, uv will handle the creation of Python interpreters and virtual environments
uv sync

#Install Playwright to use Chromium by default for browser operations
uv run playwright install

#Configuration environment
cp .env.example .env
#Edit the.env file and add your API key

#Run the project
uv run main.py

Web interface

LangManus provides a Web UI that allows users to interact with the system through a graphical interface, submit tasks and view results.

Open source protocol

The project uses an MIT license and welcomes community members to contribute code, improve documentation, or add new features.

For more information or to contribute, please visit the project homepage: (Darwin-lfl/langmanus – GitHub)。

Github：https://github.com/Darwin-lfl/langmanus

Oil tubing:

Scroll to Top