trading-agents

trading-agents is a multi-agent financial trading framework re-implemented in Go, based on the original TradingAgents.

Data Flow

The system operates through a sequential flow of specialized agents, each contributing to the final trading decision.

graph LR
    classDef analyst fill:#e0f2f7,stroke:#333,stroke-width:2px,color:black;
    classDef block fill:#f0f0f0,stroke:#333,stroke-width:2px,color:black; 
    classDef researcher fill:#e6ffe6,stroke:#333,stroke-width:2px,color:black;
    classDef risk fill:#fff2e0,stroke:#333,stroke-width:2px,color:black;
    classDef trader fill:#f0e0ff,stroke:#333,stroke-width:2px,color:black;
    classDef ioNode fill:#f5f5f5,stroke:#999,stroke-width:2px,font-weight:bold,color:black;
    

    Input[Position]:::ioNode

    subgraph Analysts
        direction LR
        FA[Fundamentals Analyst]:::analyst
        MA[Market Analyst]:::analyst
        NA[News Analyst]:::analyst
        SMA[Social Media Analyst]:::analyst
    end

    subgraph Researchers
        direction LR
        BR[Bull Researcher]:::researcher
        BeR[Bear Researcher]:::researcher
    end

    RSM[Research Manager]:::researcher

    Trader[Trader]:::trader

    subgraph RM_Team[Risk Management]
        direction LR
        AD[Aggressive Debator]:::risk
        CD[Conservative Debator]:::risk
        ND[Neutral Debator]:::risk
    end
    
    RKM[Risk Manager]:::risk

    Output[Action]:::ioNode

    Input --> Analysts
    
    Analysts --> Researchers
    Researchers --> RSM
    
    RSM --> Trader
    Analysts --> Trader

    Trader --> RM_Team
    Analysts --> RM_Team

    RM_Team --> RKM

    RKM --> Output

Loading

Getting Started

Prerequisites

• Go: Ensure you have Go (version 1.23.2 or later) installed.
• Environment Variables: Set up the necessary API keys and configuration in your environment.

Configuration

The application is configured using environment variables, typically managed via an .envrc file (which is ignored by Git for security). An example .envrc is provided:

export PORT=3001
export LOG_FORMAT="text"

export LLM_PROVIDER="gemini" # or "ollama"
export LLM_MODEL="gemini-2.5-flash" # or "gemma3:4b" etc...

# Ollama specific
export LLM_API_URL="http://localhost:11434"

# Gemini specific (using Vertex AI backend)
export GOOGLE_CLOUD_PROJECT="your-gcp-project-id"
export GOOGLE_CLOUD_LOCATION="your-gcp-region"

export FINNHUB_API_KEY="YOUR_FINNHUB_API_KEY"
export SIMFIN_API_KEY="YOUR_SIMFIN_API_KEY"

export API_TIMEOUT="10s"
export API_CACHE_TTL="8760h"

Contributing

Contributions are welcome! Please open an issue or submit a pull request with your changes. Ensure you follow the project's coding standards.