ds-trading-agents: Autonomous Trading Agents for Paradex

ds-trading-agents is a multi-agent system designed for automated trading on the Paradex platform. It allows you to run up to four autonomous agents, each powered by a different large language model (LLM), to execute trading strategies, manage goals, and interact with the Paradex API.

Features

• Multi-Agent System: Run up to four agents concurrently, each in its own process.
• Diverse AI Models: Each agent is powered by a different LLM for a variety of responses and strategies:
  • Agent 1: Google Gemini 2.0 Flash
  • Agent 2: xAI Grok Beta (via OpenRouter)
  • Agent 3: OpenAI GPT-4.1 Mini (via OpenRouter)
  • Agent 4: Anthropic Claude 3.5 Haiku (via OpenRouter)
• Paradex Integration: A comprehensive set of tools for interacting with the Paradex API, including:
  • Account information retrieval
  • Order management (open, cancel, list open orders)
  • Market data (list available markets)
  • Position tracking
• Goal Management: A flexible goal management system that allows agents to create, update, and delete long-term, medium-term, and short-term goals.
• Extensible Architecture: Built on @daydreamsai/core, the system is highly extensible. It includes a command-line interface (CLI) for manual control and an autonomous mode for hands-off operation.
• Real-time Data: Includes a "Model Context Protocol" (MCP) extension for connecting to real-time data feeds like LunarCrush for asset context, price history, and market sentiment analysis.
• Configurable: Uses environment variables for easy configuration of API keys, Starknet wallets, and other settings.

Available Tools

The agents have access to a variety of tools for trading and goal management:

Trading Tools

• getAccountInfo: Retrieves comprehensive account information, including balance, free collateral, and trading details.
• openOrder: Creates a new trading order on Paradex. Supports both MARKET and LIMIT orders.
• cancelOrder: Cancels an existing order using its order ID.
• listOpenOrders: Retrieves all currently open (unfilled) orders for the account.
• listAvailableMarkets: Fetches all available trading markets and their symbols on Paradex.
• getPositions: Retrieves all current open positions for the trading account.

Goal Management Tools

• addTask: Creates and adds a new task to the agent's goal list with a specified priority and timeframe (long-term, medium-term, or short-term).
• setGoalPlan: Sets the complete goal planning structure for the agent.
• updateGoal: Updates the properties of a specific goal by its ID.
• deleteGoal: Removes a goal from the agent's goal list.

Market Intelligence Tools (LunarCrush Integration)

The agents leverage LunarCrush API through Model Context Protocol (MCP) to gain comprehensive market intelligence:

• Asset Context: Detailed information about cryptocurrencies and tokens, including fundamental data and market metrics.
• Price History: Historical price data and trends to inform trading decisions and technical analysis.
• Market Sentiment: Real-time social sentiment analysis from various platforms to gauge market mood and potential price movements.
• Token Analytics: Overall market feeling and community sentiment about specific tokens to enhance trading strategies.
• Social Data: Aggregated social media mentions, engagement metrics, and influence scores for better market understanding.

Getting Started

Prerequisites

• Bun installed on your system.
• A Paradex account and a Starknet wallet.
• API keys for Google, OpenRouter, and LunarCrush.

Installation

1. Clone the repository:

   git clone https://github.com/your-username/ds-agents.git
   cd ds-agents

2. Install dependencies:

   bun install

Configuration

1. Create a .env file: Copy the .env.example file to a new file named .env:

   cp .env.example .env

2. Edit the .env file: Fill in the required environment variables in the .env file. This includes your API keys, Starknet wallet details, and other configuration options.

    **Required:**
   

   • LUNARCRUSH_API_KEY - For market sentiment, price history, and asset context data
   • FIREBASE_PROJECT_ID - Firebase project ID for persistent memory storage
   • FIREBASE_CLIENT_EMAIL - Firebase service account email
   • FIREBASE_PRIVATE_KEY - Firebase private key (base64 encoded)
   • AGENT1_STARKNET_ADDRESS - Starknet wallet address for Agent 1
   • AGENT1_STARKNET_PRIVATE_KEY - Starknet private key for Agent 1
   • AGENT1_GOOGLE_API_KEY - Google AI Studio API key for Gemini 2.0 Flash

   Optional (depending on which agents you run):

   • OPENROUTER_API_KEY (for agents 2, 3, and 4)
   • AGENT2_STARKNET_ADDRESS, AGENT2_STARKNET_PRIVATE_KEY
   • AGENT3_STARKNET_ADDRESS, AGENT3_STARKNET_PRIVATE_KEY
   • AGENT4_STARKNET_ADDRESS, AGENT4_STARKNET_PRIVATE_KEY

Usage

You can run one or more agents using the following bun scripts:

• Run all four agents:

  bun run start-all

• Run a specific number of agents (1-4):

  bun run src/run-agents.ts <number_of_agents>

  For example, to run two agents:

  bun run src/run-agents.ts 2

• Run a single agent by its number (1-4):

  bun run start-1
  bun run start-2
  bun run start-3
  bun run start-4

Each agent will start in its own process and log its output to the console with a unique color for easy identification.

Project Structure

/
├───.env.example        # Example environment variables
├───package.json        # Project dependencies and scripts
├───README.md           # This file
├───src/
│   ├───run-agents.ts   # Main script for running the agents
│   ├───actions/        # Agent actions (goals, tools)
│   ├───agents/         # Agent creation and configuration
│   ├───extensions/     # CLI and autonomous mode extensions
│   ├───prompts/        # System prompts for the agents
│   ├───schema/         # Zod schemas for data validation
│   └───utils/          # Utility functions (Paradex, Starknet)
└───vendored/           # Local, vendored dependencies

Contributing

Contributions are welcome! Please feel free to open an issue or submit a pull request.

1. Fork the repository.
2. Create a new branch (git checkout -b feature/your-feature).
3. Make your changes.
4. Commit your changes (git commit -m 'Add some feature').
5. Push to the branch (git push origin feature/your-feature).
6. Open a pull request.

---

This project is for educational and research purposes only. Trading cryptocurrencies involves significant risk. Use this software at your own risk.