Skip to main content
This tutorial guides you through building a trading agent that fetches trending coins, gets market data, performs analysis, and executes token swaps using the Galadriel framework.

​
Prerequisites

  • Python >=3.10
  • Git
  • Familiarity with blockchain concepts and Solana development (optional)

​
Setup

  1. Clone the Galadriel repository: 
    git clone <repository_url>
cd galadriel
    Replace <repository_url> with the actual URL of the Galadriel repository.
  2. Create a virtual environment: 
    python3 -m venv venv
source venv/bin/activate
  3. Install the Galadriel framework: 
    pip install galadriel
  4. Configure environment variables:
    • Rename template.env to .env and add your OpenAI API and Coingecko API keys: 
      OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
COINGECKO_API_KEY=<YOUR_COINGECKO_API_KEY>
SOLANA_NETWORK="devnet" # devnet, mainnet
    • Rename template.agents.env to agents.env and add your Solana keypair path: 
      AGENT_NAME="trading_agent"
SOLANA_KEY_PATH=<YOUR_SOLANA_KEYPAIR_PATH>
      If you don’t have a Solana keypair, you can create one and get funded using the Galadriel CLI (see Wallet Setup): 
      galadriel wallet create --path <YOUR_SOLANA_KEYPAIR_PATH>
      And then: 
      galadriel wallet airdrop

​
Agent Implementation

  1. Create the agent.py file: Create a new file named agent.py in the examples/trading directory and add the following code: 
    import asyncio
import os
from pathlib import Path

from dotenv import load_dotenv

from galadriel import AgentRuntime
from galadriel.agent import CodeAgent
from galadriel.core_agent import LiteLLMModel
from galadriel.clients import Cron
from galadriel.tools.web3.market_data import market_data_devnet
from galadriel.tools.web3.onchain.solana import raydium_cpmm
from galadriel.tools.web3.onchain.solana import common as solana_common

TRADING_INTERVAL_SECONDS = 300

# Set up a complex trading prompt which explains the trading strategy
TRADING_PROMPT = """
        You are an expert crypto trading advisor. Based on the user's portfolio, current market data, and trading patterns, your task is to suggest one of three actions for each token: Buy, Sell, or Hold, and execute your decision using the available tools.
        """

load_dotenv(dotenv_path=Path(".") / ".env", override=True)
load_dotenv(dotenv_path=Path(".") / ".agents.env", override=True)

model = LiteLLMModel(
    model_id="openai/gpt-4o",
    api_key=os.getenv("OPENAI_API_KEY"),
)

# Prepare a Web3 specific toolkit, relevant for the trading agent
tools = [
    market_data_devnet.fetch_mock_market_data,
    raydium_cpmm.BuyTokenWithSolTool(),
    solana_common.GetAdminWalletAddressTool(),
    solana_common.GetUserBalanceTool(),
]

# Create a trading agent
trading_agent = CodeAgent(
    prompt_template=TRADING_PROMPT,
    model=model,
    tools=tools,
    add_base_tools=True,
    additional_authorized_imports=["json"],
    max_steps=8,  # Make the trading agent more reliable by increasing the number of steps he can take to complete the task
)

# Set up the runtime
runtime = AgentRuntime(
    inputs=[Cron(TRADING_INTERVAL_SECONDS)],
    outputs=[],
    agent=trading_agent,
)

# Run the agent
asyncio.run(runtime.run())

​
Code Explanation

​
Imports

The code imports necessary modules from the Galadriel framework and standard Python libraries.

​
Configuration

The code loads environment variables from .env and agents.env files, initializes the LiteLLM model with your OpenAI API key, and defines the trading interval.

​
Tools

The code defines the tools that the agent will use:
  • market_data_devnet.fetch_mock_market_data: Fetches mock market data for development purposes.
  • raydium_cpmm.BuyTokenWithSolTool(): Tool for buying tokens with SOL using Raydium CPMM.
  • solana_common.GetAdminWalletAddressTool(): Tool to get the admin wallet address.
  • solana_common.GetUserBalanceTool(): Tool to get the user’s balance for a specific token.

​
Agent Initialization

The code initializes the CodeAgent with the trading prompt, LiteLLM model, and the defined tools. It also enables additional authorized imports such as json that might be required by the tools. Setting max_steps makes the trading agent more reliable by increasing the number of steps it can take to complete the task

​
Runtime Configuration

The code configures the AgentRuntime with a Cron input to schedule agent execution, no outputs (as the agent executes trades directly), and the initialized trading agent.

​
Start the Agent

The code runs the agent runtime, starting the autonomous trading loop.

​
Running the Agent

  1. Make sure the solana keypair defined in .agents.env has some funds.
  2. Run the agent: 
    python agent.py
The agent will then start fetching trending coins, getting market data, performing analysis, and executing token swaps every 5 minutes.