CrewAi

Integration of Agentipy with crewAI

This serves as a practical example of how to integrate Agentipy toolkit with the crewAI multi-agent framework. The integration showcases:

  1. Bringing Blockchain to AI Agents: How to equip AI agents with blockchain capabilities through Agentipy

  2. Tool Integration: How to wrap agentipy functionality as crewAI tools

  3. Configuration-Driven Design: Using YAML files to configure blockchain agents

Integration Pattern

The integration follows this pattern:

# 1. Import both libraries
from crewai import Agent, Crew, Task
from agentipy.agent import SolanaAgentKit

# 2. Create a Solana kit instance
solana_kit = SolanaAgentKit(
    private_key=os.getenv("SOLANA_PRIVATE_KEY"),
    rpc_url="https://api.devnet.solana.com"
)

# 3. Create tools from agentipy capabilities
@tool("get_sol_balance")
def get_sol_balance() -> str:
    # Use agentipy BalanceFetcher to get blockchain data
    balance = loop.run_until_complete(BalanceFetcher.get_balance(agent))
    return f"The current SOL balance is {balance:.4f} SOL"

# 4. Attach tools to crewAI agents
solana_agent = Agent(
    config=agents_config["solana_operator"],
    tools=[make_balance_tool(agent=solana_kit)],
    llm=ChatOpenAI(model=os.getenv("MODEL"), temperature=0)
)

# 5. Create tasks using these agents
solana_task = Task(
    description="Analyze instruction and take action: {user_input}",
    agent=solana_agent
)

# 6. Orchestrate with crewAI
crew = Crew(
    agents=[solana_agent],
    tasks=[solana_task],
    process=Process.sequential
)

This pattern demonstrates how to leverage both frameworks' strengths:

  • agentipy provides blockchain interaction capabilities

  • crewAI provides agent orchestration, task management, and goal-oriented behavior

Architecture

The repository is structured as follows:

src/solana_agent_agentipy/
├── config/
│   ├── agents.yaml         # Agent configuration definitions
│   └── tasks.yaml          # Task configuration definitions
├── tools/
│   ├── custom_tool.py      # Custom tools for the Solana agent
│   └── __init__.py
├── crew.py                 # Main crew definition and orchestration
├── main.py                 # Entry point for running the system
└── __init__.py

Components

1. SolanaWalletCrew

The core component of the system is the SolanaWalletCrew class defined in crew.py. This class:

  • Initializes a SolanaAgentKit with provided credentials

  • Defines the Solana agent's capabilities and tools

  • Sets up tasks for the agent to perform

  • Orchestrates the execution flow in a sequential process

2. Custom Tools

The custom_tool.py file defines tools that the agent can use to interact with the Solana blockchain, such as:

  • get_sol_balance: A tool to fetch and display the current SOL balance of the wallet

3. Configuration

The system uses YAML configuration files to define:

  • agents.yaml: Defines the Solana operator agent's role, goals, and backstory

  • tasks.yaml: Defines the operations to be performed based on user input

Setup and Installation

Prerequisites

  • Python >=3.10 <3.13

  • UV for dependency management

Installation Steps

  1. Install UV if not already installed:

pip install uv

  1. Clone the repository and navigate to the project directory

  2. Install dependencies (using the crewAI CLI):

crewai install

  1. Create a .env file with the following variables:

OPENAI_API_KEY=your_openai_api_key
SOLANA_PRIVATE_KEY=your_solana_private_key
MODEL=gpt-4-turbo

Usage

Run the project from the root folder with:

crewai run

Example input:

"Check my balance and send 1 SOL if I'm above $1000"

This will:

  1. Initialize the SolanaWalletCrew

  2. Process the natural language input

  3. Check the current SOL balance using the Solana blockchain

  4. Analyze if conditional requirements are met

  5. Execute the requested operations if conditions are satisfied

Customization

The system is designed to be easily customizable:

  1. Adding New Agents: Modify the agents.yaml file to define additional agents with different roles and capabilities

  2. Defining New Tasks: Update the tasks.yaml file to create new task definitions for your agents

  3. Creating Custom Tools: Extend the tools available to your agents by adding new tools in the tools/ directory

  4. Custom Logic: Modify the crew.py file to implement custom logic, additional tools, or specific arguments

Integration with agentipy

This project leverages the agentipy library to interact with the Solana blockchain, providing:

  • Wallet management

  • Balance fetching

  • Transaction capabilities

Dependencies

  • crewAI: Framework for creating and orchestrating AI agent systems

  • agentipy: Solana blockchain interaction toolkit

  • langchain-openai: LLM integration for natural language processing

  • python-dotenv: Environment variable management

Advanced Use Cases

The system can be extended to support:

  • Multi-token wallet management

  • Complex conditional operations

  • Integration with other blockchain operations

  • Custom token transfers and transactions

Troubleshooting

If you encounter issues:

  • Ensure your Solana private key is valid and has funds on the devnet

  • Verify your OpenAI API key is correctly set and has sufficient credits

  • Check that you're using the correct Solana RPC URL (default is devnet)

PreviousLangchainNextOpenAI

Last updated