Youtu-agent is a flexible and high-performance framework for building, running, and evaluating autonomous agents, specifically designed for users who want to leverage open-source models for complex tasks. It offers a cost-effective and accessible solution for applications like data analysis, file organization, and research, while achieving state-of-the-art performance on challenging benchmarks.

How It Works

Youtu-agent is built upon the openai-agents SDK, providing a robust foundation for agent development. Its core design emphasizes modularity and configurability, allowing for easy integration of diverse model APIs (from DeepSeek to other open-source options), custom toolkits, and various framework implementations. A key innovation is its automatic agent generation capability, which uses simple YAML-based configurations and an interactive "meta-agent" to capture user requirements and streamline setup, reducing manual overhead. The framework also features a fully asynchronous architecture for high performance and an in-depth tracing and analysis system for monitoring agent trajectories.

Quick Start & Requirements

• Install: Clone the repository, navigate to the directory, and run uv sync (or make sync), followed by source ./.venv/bin/activate.
• Prerequisites: Python 3.12+ is required. uv is recommended for dependency management.
• Setup: The project provides a quickstart example that can be launched with python scripts/cli_chat.py --stream --config default.
• Documentation: Quickstart Documentation, Examples Documentation, Evaluation Documentation.

Highlighted Details

• Achieved 71.47% on WebWalkerQA (pass@1) and 72.8% on GAIA (text-only subset, pass@1) using DeepSeek-V3 models.
• Offers out-of-the-box support for CSV analysis, literature review, file organization, and upcoming podcast/video generation.
• Features YAML-based configurations and automatic agent generation via an interactive meta-agent.
• Built on openai-agents, supporting streaming, tracing, and compatibility with various model APIs.

Maintenance & Community

The project is developed by Tencent Youtu Lab. Further community engagement details (e.g., Discord/Slack, roadmap) are not explicitly detailed in the README.

Licensing & Compatibility

The repository's license is not specified in the provided README. Compatibility for commercial use or closed-source linking would require clarification on the licensing terms.

Limitations & Caveats

The README mentions that multimodal tools and trajectories for the full GAIA benchmark will be released in the future, indicating these features are not yet available. Tool generation and automated optimization are also listed as future supported features.