Skip to main content
This tutorial demonstrates how to create a research agent that receives SOL payments from users before executing tasks. The agent analyzes investment-related queries and retrieves data from Web3 tools.

​
Overview

The agent leverages:
  • gpt-4o model from OpenAI for language processing.
  • CodeAgent to orchestrate a series of steps toward a final result.
  • AgentRuntime to connect clients to the agent and manage execution.
  • Clients:
    • SimpleMessageClient for local testing with predefined messages.
    • (Optional) TwitterMentionClient for processing mentions on Twitter.
  • Web3 Tools:
    • get_coin_price (CoinGecko) to fetch real-time cryptocurrency prices.
    • get_token_profile (DexScreener) to retrieve token profiles.
  • A Pricing Mechanism using Solana (SOL) payments. The agent will validate payment before executing the task.

​
Payment & Transaction Requirements

To use the agent, a client (user) must provide a task with either:
  • A link to a Solana transaction (e.g., from Solscan)
  • A transaction signature on the Solana blockchain

​
Setup

  1. Local Environment & Galadriel Installation: 
    python3 -m venv venv
source venv/bin/activate
pip install galadriel
  2. Agent Wallet Creation: The agent needs a Solana wallet to receive payments. If you don’t have one, create a new one. The tutorial is described in Wallet Tutorial.
  3. Environment Variables Configuration: Rename template.env to .env and populate it with your OpenAI API key, agent wallet address, and optional Twitter API credentials (if using TwitterMentionClient): 
    OPENAI_API_KEY=<YOUR_OPENAI_API_KEY>
TWITTER_CONSUMER_API_KEY=<YOUR_TWITTER_CONSUMER_API_KEY> #Optional
TWITTER_CONSUMER_API_SECRET=<YOUR_TWITTER_CONSUMER_API_SECRET> #Optional
TWITTER_ACCESS_TOKEN=<YOUR_TWITTER_ACCESS_TOKEN> #Optional
TWITTER_ACCESS_TOKEN_SECRET=<YOUR_TWITTER_ACCESS_TOKEN_SECRET> #Optional
TWITTER_USER_ID=<YOUR_TWITTER_USER_ID> #Optional
    • Make sure you have configured SOLANA_KEY_PATH in .agents.env file to point the path to your Solana wallet
  4. Agent Code (agent.py): Create a file named agent.py with the following Python code. Remember to replace placeholders with your actual values. 
    import asyncio
import os
from pathlib import Path

from dotenv import load_dotenv

from galadriel import AgentRuntime, CodeAgent
from galadriel.clients import SimpleMessageClient
# from galadriel.clients import TwitterMentionClient #Uncomment to enable Twitter client
from galadriel.core_agent import LiteLLMModel
from galadriel.entities import Pricing
from galadriel.tools.web3 import dexscreener, coingecko

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

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

# Set up a researcher who will perform Web3 related tasks
researcher = CodeAgent(
    tools=[
        coingecko.get_coin_price,
        dexscreener.get_token_profile,
    ],
    model=model,
    add_base_tools=True,
    additional_authorized_imports=["json"],
)

# Configure agent's pricing information
agent_pricing = Pricing(
    wallet_address="<YOUR_AGENT_WALLET_ADDRESS>",  # Replace with the agent's wallet address
    cost=0.001,  # Price per task in SOL
)

# The client will pass a research task to the agent with either
# - a link to the transaction on Solscan
# - a signature of transaction of Solana
# Example transaction passed to SimpleMessageClient below:
# https://explorer.solana.com/tx/5aqB4BGzQyFybjvKBjdcP8KAstZo81ooUZnf64vSbLLWbUqNSGgXWaGHNteiK2EJrjTmDKdLYHamJpdQBFevWuvy
simple_client = SimpleMessageClient(
    "Is Bitcoin good investment now with high prices? https://solscan.io/tx/5aqB4BGzQyFybjvKBjdcP8KAstZo81ooUZnf64vSbLLWbUqNSGgXWaGHNteiK2EJrjTmDKdLYHamJpdQBFevWuvy",
)

# Alternatively, Twitter (or Discord, Telegram) client which can be passed to inputs and outputs in order to get tasks from social channels
# twitter_client = TwitterMentionClient(
#     TwitterCredentials(
#         consumer_api_key=os.getenv("TWITTER_CONSUMER_API_KEY"),
#         consumer_api_secret=os.getenv("TWITTER_CONSUMER_API_SECRET"),
#         access_token=os.getenv("TWITTER_ACCESS_TOKEN"),
#         access_token_secret=os.getenv("TWITTER_ACCESS_TOKEN_SECRET"),
#     ),
#     user_id=os.getenv("TWITTER_USER_ID"),
# )

# Combine all elements into runtime
runtime = AgentRuntime(
    inputs=[simple_client],  # Runtime inputs, pass twitter_client if you use it
    outputs=[simple_client],  # Runtime outputs, pass twitter_client if you use it
    agent=researcher,
    pricing=agent_pricing,
)

# Run the runtime with agent
asyncio.run(runtime.run())
    Code Explanation:
    • Imports: Imports necessary libraries and modules.
    • Environment Variables: Loads environment variables from a .env file.
    • Model Setup: Initializes the LiteLLMModel with the specified model ID and API key.
    • Tool Setup:
      • Initializes coingecko.get_coin_price and dexscreener.get_token_profile
    • Agent Setup:
      • Initializes the CodeAgent with the specified tools
    • Client Setup:
      • Initializes the SimpleMessageClient with a sample question and Solana transaction link.
    • Runtime Setup:
      • Initializes the AgentRuntime with the agent, pricing, input, and output.
  5. Run the agent: 
    python agent.py

​
Optional: Running the agent with Twitter Client

To use the Twitter client, you have to set TWITTER_CONSUMER_API_KEY, TWITTER_CONSUMER_API_SECRET, TWITTER_ACCESS_TOKEN and TWITTER_ACCESS_TOKEN_SECRET environment variables defined inside .env file.
  1. Uncomment Twitter related code snippets:
    • TwitterMentionClient import statement
    • twitter_client initialization with the proper credentials
    • Pass twitter_client client to the AgentRuntime arguments for inputs and outputs
  2. Run the agent: 
    python agent.py
The agent will now monitor Twitter mentions, check the transaction signature for payment, and provide its analysis if payment validation is successful.
⌘I