diff --git a/README.md b/README.md index ca9785e..ea8a641 100644 --- a/README.md +++ b/README.md @@ -71,13 +71,14 @@ Our vision is to revolutionize how AI agents collaborate to solve real-world tas - [**Install Dependencies**](#install-dependencies) - [**Setup Environment Variables**](#setup-environment-variables) - [**Running with Docker**](#running-with-docker) - - [🚀 Quick Start](#-quick-start) +- [🧰 Toolkits and Capabilities](#-toolkits-and-capabilities) - [🌐 Web Interface](#-web-interface) - [🧪 Experiments](#-experiments) - [⏱️ Future Plans](#️-future-plans) - [📄 License](#-license) - [🖊️ Cite](#️-cite) +- [🤝 Contributing](#-contributing) - [🔥 Community](#-community) - [❓ FAQ](#-faq) - [⭐ Star History](#-star-history) @@ -87,7 +88,8 @@ Our vision is to revolutionize how AI agents collaborate to solve real-world tas - **[2025.03.11]**: We added MCPToolkit, FileWriteToolkit, and TerminalToolkit to enhance OWL agents with MCP tool calling, file writing capabilities, and terminal command execution. - **[2025.03.09]**: We added a web-based user interface that makes it easier to interact with the system. -- **[2025.03.07]**: We open-source the codebase of 🦉 OWL project. +- **[2025.03.07]**: We open-sourced the codebase of the 🦉 OWL project. +- **[2025.03.03]**: OWL achieved the #1 position among open-source frameworks on the GAIA benchmark with a score of 58.18. # 🎬 Demo Video @@ -106,6 +108,8 @@ https://private-user-images.githubusercontent.com/55657767/420212194-e813fc05-13 # 🛠️ Installation +OWL supports multiple installation methods to fit your workflow preferences. Choose the option that works best for you. + ## Option 1: Using uv (Recommended) ```bash @@ -253,9 +257,7 @@ For more detailed Docker usage instructions, including cross-platform support, o # 🚀 Quick Start - - -Run the following demo case: +After installation and setting up your environment variables, you can start using OWL right away: ```bash python owl/run.py @@ -323,13 +325,17 @@ print(f"\033[94mAnswer: {answer}\033[0m") OWL will then automatically invoke document-related tools to process the file and extract the answer. -Example tasks you can try: +### Example Tasks + +Here are some tasks you can try with OWL: + - "Find the latest stock price for Apple Inc." - "Analyze the sentiment of recent tweets about climate change" - "Help me debug this Python code: [your code here]" - "Summarize the main points from this research paper: [paper URL]" +- "Create a data visualization for this dataset: [dataset path]" -# 🧰 Configuring Toolkits +# 🧰 Toolkits and Capabilities > **Important**: Effective use of toolkits requires models with strong tool calling capabilities. For multimodal toolkits (Web, Image, Video), models must also have multimodal understanding abilities. @@ -392,16 +398,19 @@ Selecting only necessary toolkits optimizes performance and reduces resource usa # 🌐 Web Interface -OWL now includes a web-based user interface that makes it easier to interact with the system. To start the web interface, run: +OWL includes an intuitive web-based user interface that makes it easier to interact with the system. + +## Starting the Web UI ```bash +# Start the Chinese version python run_app.py -# For English Version +# Start the English version python run_app_en.py ``` -The web interface provides the following features: +## Features - **Easy Model Selection**: Choose between different models (OpenAI, Qwen, DeepSeek, etc.) - **Environment Variable Management**: Configure your API keys and other settings directly from the UI @@ -415,21 +424,25 @@ The web interface is built using Gradio and runs locally on your machine. No dat To reproduce OWL's GAIA benchmark score of 58.18: 1. Switch to the `gaia58.18` branch: -```bash -git checkout gaia58.18 -``` + ```bash + git checkout gaia58.18 + ``` -1. Run the evaluation script: -```bash -python run_gaia_roleplaying.py -``` +2. Run the evaluation script: + ```bash + python run_gaia_roleplaying.py + ``` + +This will execute the same configuration that achieved our top-ranking performance on the GAIA benchmark. # ⏱️ Future Plans -- [ ] Write a technical blog post detailing our exploration and insights in multi-agent collaboration in real-world tasks. -- [ ] Enhance the toolkit ecosystem with more specialized tools for domain-specific tasks. -- [ ] Develop more sophisticated agent interaction patterns and communication protocols +We're continuously working to improve OWL. Here's what's on our roadmap: +- [ ] Write a technical blog post detailing our exploration and insights in multi-agent collaboration in real-world tasks +- [ ] Enhance the toolkit ecosystem with more specialized tools for domain-specific tasks +- [ ] Develop more sophisticated agent interaction patterns and communication protocols +- [ ] Improve performance on complex multi-step reasoning tasks # 📄 License @@ -479,6 +492,14 @@ Join us for further discussions! A: If OWL determines that a task can be completed using non-browser tools (such as search or code execution), the browser will not be launched. The browser window will only appear when OWL determines that browser-based interaction is necessary. +**Q: Which Python version should I use?** + +A: OWL supports Python 3.10, 3.11, and 3.12. + +**Q: How can I contribute to the project?** + +A: See our [Contributing](#-contributing) section for details on how to get involved. We welcome contributions of all kinds, from code improvements to documentation updates. + # ⭐ Star History [![Star History Chart](https://api.star-history.com/svg?repos=camel-ai/owl&type=Date)](https://star-history.com/#camel-ai/owl&Date) diff --git a/README_zh.md b/README_zh.md index 9ed9625..20d3a92 100644 --- a/README_zh.md +++ b/README_zh.md @@ -73,13 +73,16 @@ - [**设置环境变量**](#设置环境变量) - [**使用Docker运行**](#使用docker运行) - [🚀 快速开始](#-快速开始) +- [🧰 工具包与功能](#-工具包与功能) - [🌐 网页界面](#-网页界面) - [🧪 实验](#-实验) - [⏱️ 未来计划](#️-未来计划) - [📄 许可证](#-许可证) - [🖊️ 引用](#️-引用) +- [🤝 贡献](#-贡献) - [🔥 社区](#-社区) - [❓ 常见问题](#-常见问题) +- [⭐ Star History](#-star-history) # 🔥 新闻 @@ -87,6 +90,7 @@ - **[2025.03.11]**: 我们添加了 MCPToolkit、FileWriteToolkit 和 TerminalToolkit,增强 OWL Agent的工具调用、文件写入能力和终端命令执行功能。 - **[2025.03.09]**: 我们添加了基于网页的用户界面,使系统交互变得更加简便。 - **[2025.03.07]**: 我们开源了 🦉 OWL 项目的代码库。 +- **[2025.03.03]**: OWL 在 GAIA 基准测试中取得 58.18 平均分,在开源框架中排名第一! # 🎬 演示视频 @@ -325,7 +329,7 @@ OWL 将自动调用与文档相关的工具来处理文件并提取答案。 - "帮我调试这段 Python 代码:[在此粘贴你的代码]" - "总结这篇研究论文的主要观点:[论文URL]" -# 🧰 配置工具包 +# 🧰 工具包与功能 > **重要提示**:有效使用工具包需要具备强大工具调用能力的模型。对于多模态工具包(Web、图像、视频),模型还必须具备多模态理解能力。 @@ -423,10 +427,12 @@ python run_gaia_roleplaying.py # ⏱️ 未来计划 -- [ ] 撰写一篇技术博客,详细介绍我们在现实任务中多智能体协作方面的探索与见解。 -- [ ] 通过引入更多针对特定领域任务的专业工具,进一步完善工具生态系统。 -- [ ] 开发更复杂的智能体交互模式和通信协议 +我们正在不断努力改进 OWL。以下是我们的路线图: +- [ ] 撰写技术博客,详细介绍我们在现实任务中多智能体协作方面的探索与见解 +- [ ] 通过引入更多针对特定领域任务的专业工具,进一步完善工具生态系统 +- [ ] 开发更复杂的智能体交互模式和通信协议 +- [ ] 提高复杂多步推理任务的性能 # 📄 许可证 @@ -476,6 +482,14 @@ python run_gaia_roleplaying.py A: 当OWL判断某个任务可以使用非浏览器工具(如搜索、代码分析等)完成时,浏览器就不会启动。只有在判断需要使用浏览器工具的时候,本地才会弹出浏览器窗口,并进行浏览器模拟交互。 +**Q: 我应该使用哪个Python版本?** + +A: OWL支持Python 3.10、3.11和3.12。为了与所有依赖项获得最佳兼容性,我们推荐使用Python 3.10。 + +**Q: 我如何为项目做贡献?** + +A: 请参阅我们的[贡献](#-贡献)部分,了解如何参与的详细信息。我们欢迎各种形式的贡献,从代码改进到文档更新。 + # ⭐ Star History [![Star History Chart](https://api.star-history.com/svg?repos=camel-ai/owl&type=Date)](https://star-history.com/#camel-ai/owl&Date)