[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-73634":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":23,"hasPages":25,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":29,"readmeContent":30,"aiSummary":31,"trendingCount":16,"starSnapshotCount":16,"syncStatus":15,"lastSyncTime":32,"discoverSource":33},73634,"QuantAgent","Y-Research-SBU\u002FQuantAgent","Y-Research-SBU","Official Repository for QuantAgent","https:\u002F\u002Fy-research-sbu.github.io\u002FQuantAgent\u002F",null,"HTML",2723,591,25,2,0,7,26,89,21,96.22,"MIT License",false,"main",true,[27,28],"agentic-ai","large-language-models","2026-06-12 04:01:10","\u003Cdiv align=\"center\">\n\n![QuantAgent Banner](assets\u002Fbanner.png)\n\u003Ch2>QuantAgent: Price-Driven Multi-Agent LLMs for High-Frequency Trading\u003C\u002Fh2>\n\n\u003C\u002Fdiv>\n\n\n\n\u003Cdiv align=\"center\">\n\n\u003Cdiv style=\"position: relative; text-align: center; margin: 20px 0;\">\n  \u003Cdiv style=\"position: absolute; top: -10px; right: 20%; font-size: 1.2em;\">\u003C\u002Fdiv>\n  \u003Cp>\n    \u003Ca href=\"https:\u002F\u002Fmachineily.github.io\u002F\">Fei Xiong\u003C\u002Fa>\u003Csup>1,2 ★\u003C\u002Fsup> \n    \u003Ca href=\"https:\u002F\u002Fwyattz23.github.io\">Xiang Zhang\u003C\u002Fa>\u003Csup>3 ★\u003C\u002Fsup> \n    \u003Ca href=\"https:\u002F\u002Fscholar.google.com\u002Fcitations?user=hFhhrmgAAAAJ&hl=en\">Aosong Feng\u003C\u002Fa>\u003Csup>4\u003C\u002Fsup> \n    \u003Ca href=\"https:\u002F\u002Fintersun.github.io\u002F\">Siqi Sun\u003C\u002Fa>\u003Csup>5\u003C\u002Fsup> \n    \u003Ca href=\"https:\u002F\u002Fchenyuyou.me\u002F\">Chenyu You\u003C\u002Fa>\u003Csup>1\u003C\u002Fsup>\n  \u003C\u002Fp>\n  \n  \u003Cp>\n    \u003Csup>1\u003C\u002Fsup> Stony Brook University    \n    \u003Csup>2\u003C\u002Fsup> Carnegie Mellon University   \n    \u003Csup>3\u003C\u002Fsup> University of British Columbia    \u003Cbr>\n    \u003Csup>4\u003C\u002Fsup> Yale University    \n    \u003Csup>5\u003C\u002Fsup> Fudan University    \n    ★ Equal Contribution \u003Cbr>\n  \u003C\u002Fp>\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\" style=\"margin: 20px 0;\">\n  \u003Ca href=\"README.md\">English\u003C\u002Fa> | \u003Ca href=\"README_CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Cbr>\n\u003Cp align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2509.09995\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F💡%20ArXiv-2509.09995-B31B1B?style=flat-square\" alt=\"Paper\">\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002FY-Research-SBU.github.io\u002FQuantAgent\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject-Website-blue?style=flat-square&logo=googlechrome\" alt=\"Project Website\">\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FY-Research-SBU\u002FQuantAgent\u002Fblob\u002Fmain\u002Fassets\u002Fwechat_0203.jpg\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWeChat-Group-green?style=flat-square&logo=wechat\" alt=\"WeChat Group\">\n  \u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002Ft9nQ6VXQ\">\n    \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-Community-5865F2?style=flat-square&logo=discord\" alt=\"Discord Community\">\n  \u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n\nA sophisticated multi-agent trading analysis system that combines technical indicators, pattern recognition, and trend analysis using LangChain and LangGraph. The system provides both a web interface and programmatic access for comprehensive market analysis.\n\n\n\u003Cdiv align=\"center\">\n\n🚀 [Features](#-features) | ⚡ [Installation](#-installation) | 🎬 [Usage](#-usage) | 🔧 [Implementation Details](#-implementation-details) | 🤝 [Contributing](#-contributing) | 📄 [License](#-license)\n\n\u003C\u002Fdiv>\n\n## 🚀 Features\n\n\u003C!-- - **Multi-Agent Analysis**: Four specialized agents working together: -->\n  \n  ### Indicator Agent\n  \n  • Computes five technical indicators—including RSI to assess momentum extremes, MACD to quantify convergence–divergence dynamics, and the Stochastic Oscillator to measure closing prices against recent trading ranges—on each incoming K‑line, converting raw OHLC data into precise, signal-ready metrics.\n\n  ![indicator agent](assets\u002Findicator.png)\n  \n ### Pattern Agent\n  \n  • Upon a pattern query, the Pattern Agent first uses the agent draws the recent price chart, spots its main highs, lows, and general up‑or‑down moves, compares that shape to a set of familiar patterns, and returns a short, plain‑language description of the best match.\n  \n  ![indicator agent](assets\u002Fpattern.png)\n  \n  ### Trend Agent\n  \n  • Leverages tool-generated annotated K‑line charts overlaid with fitted trend channels—upper and lower boundary lines tracing recent highs and lows—to quantify market direction, channel slope, and consolidation zones, then delivers a concise, professional summary of the prevailing trend.\n  \n  ![trend agent](assets\u002Ftrend.png)\n\n  ### Decision Agent\n  \n  • Synthesizes outputs from the Indicator, Pattern, Trend, and Risk agents—including momentum metrics, detected chart formations, channel analysis, and risk–reward assessments—to formulate actionable trade directives, clearly specifying LONG or SHORT positions, recommended entry and exit points, stop‑loss thresholds, and concise rationale grounded in each agent’s findings.\n  \n  ![alt text](assets\u002Fdecision.png)\n\n### Web Interface\nModern Flask-based web application with:\n  - Real-time market data from Yahoo Finance\n  - Interactive asset selection (stocks, crypto, commodities, indices)\n  - Multiple timeframe analysis (1m to 1d)\n  - Dynamic chart generation\n  - API key management\n\n## 📦 Installation\n\n### 1. Create and Activate Conda Environment\n\n```bash\nconda create -n quantagents python=3.11\nconda activate quantagents\n```\n\n### 2. Install Dependencies\n\n```bash\npip install -r requirements.txt\n```\n\nIf you encounter issues with TA-lib-python, \ntry\n\n```bash\nconda install -c conda-forge ta-lib\n```\n\nOr visit the [TA-Lib Python repository](https:\u002F\u002Fgithub.com\u002Fta-lib\u002Fta-lib-python) for detailed installation instructions.\n\n### 3. Set Up LLM API Key\nYou can set it in our Web InterFace Later,\n\n![alt text](assets\u002Fapibox.png)\n\nOr set it as an environment variable:\n```bash\n# For OpenAI\nexport OPENAI_API_KEY=\"your_openai_api_key_here\"\n\n# For Anthropic (Claude)\nexport ANTHROPIC_API_KEY=\"your_anthropic_api_key_here\"\n\n# For Qwen (DashScope, based in Singapore — delays may occur)\nexport DASHSCOPE_API_KEY=\"your_dashscope_api_key_here\"\n\n# For MiniMax (204K context, OpenAI-compatible API)\nexport MINIMAX_API_KEY=\"your_minimax_api_key_here\"\n\n```\n\n\n\n\n\n## 🚀 Usage\n\n### Start the Web Interface\n\n```bash\npython web_interface.py\n```\n\nThe web application will be available at `http:\u002F\u002F127.0.0.1:5000`\n\n### Web Interface Features\n\n1. **Asset Selection**: Choose from available stocks, crypto, commodities, and indices\n2. **Timeframe Selection**: Analyze data from 1-minute to daily intervals\n3. **Date Range**: Select custom date ranges for analysis\n4. **Real-time Analysis**: Get comprehensive technical analysis with visualizations\n5. **API Key Management**: Update your OpenAI API key through the interface\n\n## 📺 Demo\n\n![Quick preview](assets\u002Fdemo.gif)\n\n\n## 🔧 Implementation Details\n\n\n**Important Note**: Our model requires an LLM that can take images as input, as our agents generate and analyze visual charts for pattern recognition and trend analysis.\n\n### Python Usage\n\nTo use QuantAgents inside your code, you can import the trading_graph module and initialize a TradingGraph() object. The .invoke() function will return a comprehensive analysis. You can run web_interface.py, here's also a quick example:\n\n```python\nfrom trading_graph import TradingGraph\n\n# Initialize the trading graph\ntrading_graph = TradingGraph()\n\n# Create initial state with your data\ninitial_state = {\n    \"kline_data\": your_dataframe_dict,\n    \"analysis_results\": None,\n    \"messages\": [],\n    \"time_frame\": \"4hour\",\n    \"stock_name\": \"BTC\"\n}\n\n# Run the analysis\nfinal_state = trading_graph.graph.invoke(initial_state)\n\n# Access results\nprint(final_state.get(\"final_trade_decision\"))\nprint(final_state.get(\"indicator_report\"))\nprint(final_state.get(\"pattern_report\"))\nprint(final_state.get(\"trend_report\"))\n```\n\nYou can also adjust the default configuration to set your own choice of LLMs or analysis parameters in web_interface.py.\n\n```python\nif provider == \"anthropic\":\n    # Set default Claude models if not already set to Anthropic models\n    if not analyzer.config[\"agent_llm_model\"].startswith(\"claude\"):\n        analyzer.config[\"agent_llm_model\"] = \"claude-haiku-4-5-20251001\"\n    if not analyzer.config[\"graph_llm_model\"].startswith(\"claude\"):\n        analyzer.config[\"graph_llm_model\"] = \"claude-haiku-4-5-20251001\"\n\nelif provider == \"qwen\":\n    # Set default Qwen models if not already set to Qwen models\n    if not analyzer.config[\"agent_llm_model\"].startswith(\"qwen\"):\n        analyzer.config[\"agent_llm_model\"] = \"qwen3-max\"\n    if not analyzer.config[\"graph_llm_model\"].startswith(\"qwen\"):\n        analyzer.config[\"graph_llm_model\"] = \"qwen3-vl-plus\"\n\nelif provider == \"minimax\":\n    # Set default MiniMax models (204K context window)\n    if not analyzer.config[\"agent_llm_model\"].startswith(\"MiniMax\"):\n        analyzer.config[\"agent_llm_model\"] = \"MiniMax-M2.7\"\n    if not analyzer.config[\"graph_llm_model\"].startswith(\"MiniMax\"):\n        analyzer.config[\"graph_llm_model\"] = \"MiniMax-M2.7\"\n\nelse:\n    # Set default OpenAI models if not already set to OpenAI models\n    if analyzer.config[\"agent_llm_model\"].startswith((\"claude\", \"qwen\", \"MiniMax\")):\n        analyzer.config[\"agent_llm_model\"] = \"gpt-4o-mini\"\n    if analyzer.config[\"graph_llm_model\"].startswith((\"claude\", \"qwen\", \"MiniMax\")):\n        analyzer.config[\"graph_llm_model\"] = \"gpt-4o\"\n\n```\n\nFor live data, we recommend using the web interface as it provides access to real-time market data through yfinance. The system automatically fetches the most recent 30 candlesticks for optimal LLM analysis accuracy.\n\n### Configuration Options\n\nThe system supports the following configuration parameters:\n\n- `agent_llm_model`: Model for individual agents (default: \"gpt-4o-mini\")\n- `graph_llm_model`: Model for graph logic and decision making (default: \"gpt-4o\")\n- `agent_llm_temperature`: Temperature for agent responses (default: 0.1)\n- `graph_llm_temperature`: Temperature for graph logic (default: 0.1)\n\n**Note**: The system uses default token limits for comprehensive analysis. No artificial token restrictions are applied.\n\nYou can view the full list of configurations in `default_config.py`.\n\n## 🤝 Contributing\n\n1. Fork the repository\n2. Create a feature branch\n3. Make your changes\n4. Add tests if applicable\n5. Submit a pull request\n\n## 📄 License\n\nThis project is licensed under the MIT License - see the LICENSE file for details.\n\n## 🔖 Citation\n```\n@article{xiong2025quantagent,\n  title={QuantAgent: Price-Driven Multi-Agent LLMs for High-Frequency Trading},\n  author={Fei Xiong and Xiang Zhang and Aosong Feng and Siqi Sun and Chenyu You},\n  journal={arXiv preprint arXiv:2509.09995},\n  year={2025}\n}\n```\n\n\n## 🙏 Acknowledgements\n\nThis repository was built with the help of the following libraries and frameworks:\n\n- [**LangGraph**](https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flanggraph)\n- [**OpenAI**](https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-python)\n- [**Anthropic (Claude)**](https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fanthropic-sdk-python)\n- [**Qwen**](https:\u002F\u002Fgithub.com\u002FQwenLM\u002FQwen)\n- [**MiniMax**](https:\u002F\u002Fplatform.minimaxi.com\u002F) — 204K context, OpenAI-compatible API\n- [**yfinance**](https:\u002F\u002Fgithub.com\u002Franaroussi\u002Fyfinance)\n- [**Flask**](https:\u002F\u002Fgithub.com\u002Fpallets\u002Fflask)\n- [**TechnicalAnalysisAutomation**](https:\u002F\u002Fgithub.com\u002Fneurotrader888\u002FTechnicalAnalysisAutomation\u002Ftree\u002Fmain)\n- [**tvdatafeed**](https:\u002F\u002Fgithub.com\u002FrongardF\u002Ftvdatafeed)\n## ⚠️ Disclaimer\n\nThis software is for educational and research purposes only. It is not intended to provide financial advice. Always do your own research and consider consulting with a financial advisor before making investment decisions.\n\n## 🐛 Troubleshooting\n\n### Common Issues\n\n1. **TA-Lib Installation**: If you encounter TA-Lib installation issues, refer to the [official repository](https:\u002F\u002Fgithub.com\u002Fta-lib\u002Fta-lib-python) for platform-specific instructions.\n\n2. **LLM API Key**: Ensure your API key is properly set in the environment or through the web interface.\n\n3. **Data Fetching**: The system uses Yahoo Finance for data. Some symbols might not be available or have limited historical data.\n\n4. **Memory Issues**: For large datasets, consider reducing the analysis window or using a smaller timeframe.\n\n### Support\n\nIf you encounter any issues, please:\n\n0. Try refresh and re-enter LLM API key\n1. Check the troubleshooting section above\n2. Review the error messages in the console\n3. Ensure all dependencies are properly installed\n4. Verify your API key is valid and has sufficient credits\n\n## 📧 Contact\n\nFor questions, feedback, or collaboration opportunities, please contact:\n\n**Email**: [chenyu.you@stonybrook.edu](mailto:chenyu.you@stonybrook.edu), [siqisun@fudan.edu.cn](mailto:siqisun@fudan.edu.cn)\n\n\n## Star History\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=Y-Research-SBU\u002FQuantAgent&type=Date)](https:\u002F\u002Fwww.star-history.com\u002F#Y-Research-SBU\u002FQuantAgent&Date)\n","QuantAgent 是一个专为高频交易设计的价格驱动多代理大型语言模型系统。它利用LangChain和LangGraph技术，集成了技术指标计算、模式识别及趋势分析等功能，通过多个专门化代理协同工作，如指示器代理负责计算包括RSI、MACD在内的五种关键技术指标，将原始OHLC数据转换为精确的信号指标；而模式代理则能根据用户查询识别特定市场模式。该项目提供网页界面与编程接口两种访问方式，适用于需要进行深入市场分析并基于实时价格变动做出快速决策的金融场景。","2026-06-11 03:46:29","high_star"]