Implementing Multiple AI Agents with Google Gini and Crew AI

Introduction

In this article, we will explore how to implement multiple AI agents for real-world use cases using Crew AI and Google's Gini model. This step-by-step guide demonstrates setting up a coding environment, creating agents, defining tasks, and executing the process to produce a comprehensive output. We chose Google Gini for its free usage benefits compared to the paid OpenAI API.

Setting Up the Environment

To begin, we need to set up our environment. We will use Python 3.10 and create a virtual environment. Here are the steps:

• Create a virtual environment using Python 3.10:

python3.10 -m venv env        

• Activate the virtual environment:

# On Windows
.\env\Scripts\activate

# On macOS/Linux
source env/bin/activate        

• Create a requirements.txt file listing the necessary libraries, including Crew AI and Langchain's Google Gini library:

crewai
langchain-google-gini        

• Install the dependencies:

pip install -r requirements.txt        

Coding Environment Configuration

Our coding environment consists of several key files and folders:

• task.py: Defines the tasks for the AI agents.
• agents.py: Contains the definitions and configurations of the AI agents.
• tools.py: Includes the tools that the agents will use, such as the Google search tool.
• crew.py: The main file to execute the process and coordinate the agents.

Creating AI Agents

agents.py

from langchain_google_gini import GiniAgent

class NewsResearcher(GiniAgent):
    def __init__(self):
        super().__init__(
            name="News Researcher",
            description="Uncovers groundbreaking technologies",
            tools=["google_search"]
        )

class NewsWriter(GiniAgent):
    def __init__(self):
        super().__init__(
            name="News Writer",
            description="Crafts engaging narratives based on research findings",
            tools=["google_search"]
        )        

Implementing Tools

tools.py

from crewai.tools import Tool
from serpdev import GoogleSearch

class GoogleSearchTool(Tool):
    def __init__(self, api_key):
        self.api_key = api_key
        self.search_engine = GoogleSearch(api_key=api_key)

    def search(self, query):
        return self.search_engine.search(query)        

Defining Tasks

task.py

class ResearchTask:
    def __init__(self):
        self.objective = "Identify the next big trend in AI"
        self.output = None

    def execute(self, researcher):
        query = "latest trends in AI 2024"
        self.output = researcher.tools["google_search"].search(query)
        return self.output

class WritingTask:
    def __init__(self, research_output):
        self.objective = "Compose a detailed article based on the research findings"
        self.research_output = research_output
        self.output = None

    def execute(self, writer):
        self.output = writer.write_article(self.research_output)
        return self.output        

Execution Process

crew.py

from agents import NewsResearcher, NewsWriter
from tools import GoogleSearchTool
from task import ResearchTask, WritingTask

def main():
    # Initialize tools
    google_search_tool = GoogleSearchTool(api_key="YOUR_SERPDEV_API_KEY")

    # Initialize agents
    researcher = NewsResearcher()
    writer = NewsWriter()

    # Assign tools to agents
    researcher.add_tool("google_search", google_search_tool)
    writer.add_tool("google_search", google_search_tool)

    # Define tasks
    research_task = ResearchTask()
    research_output = research_task.execute(researcher)

    writing_task = WritingTask(research_output)
    article = writing_task.execute(writer)

    # Print final output
    print("Final Article:", article)

if __name__ == "__main__":
    main()        

Final Output

Upon executing the process, the AI agents collaborate to produce a comprehensive article and blog post. This demonstrates the power of multiple AI agents working together to perform complex tasks and generate useful outputs for real-world applications.

This tutorial provides a practical example of using Crew AI and Google Gini to implement AI agents, highlighting their potential to automate research and content creation effectively.

Conclusion

Implementing multiple AI agents using Crew AI and Google Gini is a powerful approach to automate and streamline complex tasks. By following this guide, you can set up a similar environment and explore the capabilities of AI agents in your projects. The collaboration between agents to perform research and content creation showcases the potential of AI in various applications, from technology to media.