This project provides a multi-agent A-share intelligent investment research system, simulating institutional collaboration and debate to generate structured trading advice. It targets investors and researchers seeking AI-driven insights, offering a visualized, interactive platform that leverages advanced LLMs for deep market analysis.

How It Works

The system employs a multi-agent architecture inspired by TradingAgents, decomposing complex investment research into specialized roles. Fourteen AI agents, including analysts (fundamentals, sentiment, news, tech, macro, funds), researchers (bull/bear debaters), and decision-makers (traders, risk control, portfolio managers), collaborate in a simulated institutional workflow. This approach facilitates claim-driven structured debate and risk-controlled decision-making, aiming for more robust and validated investment plans.

Quick Start & Requirements

• Docker: Recommended for one-click deployment. Pull the image (ghcr.io/kylinmountain/tradingagents-ashare:latest) and run the container, mapping a local data directory and setting TA_APP_SECRET_KEY. Access via http://localhost:8000.
• Source Install: Requires Python 3.10+ for the backend and Node.js 18+ for the frontend. Clone the repository, sync backend dependencies (uv sync), build the frontend (npm install, npm run build), configure .env, and run the backend (uvicorn api.main:app --port 8000).
• Prerequisites: LLM API keys (OpenAI, Anthropic, Google Gemini, etc.) are configured via the frontend settings. A secret key (TA_APP_SECRET_KEY) is recommended for production.
• Links: Online Experience, Releases, OpenClaw Skill.

Highlighted Details

• Interactive Visualization: Real-time, token-level streaming of multi-agent debates, including dialogue, timelines, and decision summaries.
• Natural Language Interface: Supports intent-driven queries (e.g., "调研茅台短线") for automatic analysis initiation.
• Multi-Model Support: Integrates with a wide range of LLM providers including OpenAI, Anthropic, Google Gemini, DeepSeek, Moonshot, Zhipu, and SiliconFlow.
• API & Integration: Offers a REST API for programmatic access and integrates with OpenClaw via the tradingagents-analysis skill.
• Persistence: Features database persistence for watchlists, portfolio tracking, and structured research reports, enabling historical retrieval and analysis.

Maintenance & Community

The project is a derivative work, with core architecture inspired by TauricResearch/TradingAgents. Specific community channels, active contributors, or sponsorship details are not detailed in the provided README.

Licensing & Compatibility

The project is licensed under a dual-license: Apache 2.0 for the original TradingAgents components and PolyForm Noncommercial 1.0.0 for new modules and modifications. The PolyForm Noncommercial license strictly prohibits commercial use.

Limitations & Caveats

This system is intended strictly for academic research, technical demonstration, and learning purposes; it does not constitute investment advice. Users assume all risks associated with actual trading, as the system's outputs are algorithmic results and not liable for investment performance. Data sources may exhibit latency or biases, requiring verification against official exchange information. The non-commercial license restricts deployment in any business context.