RAG Tool Agent

1
import os
2
import textwrap
3
4
from composio_crewai import Action, App, ComposioToolSet
5
from crewai import Agent, Crew, Process, Task
6
from dotenv import load_dotenv
7
from langchain_openai import ChatOpenAI
8
9
# Load environment variables
10
load_dotenv()
11
12
# Initialize the language model
13
llm = ChatOpenAI(model="gpt-4o")
14
15
# Set up Composio tools
16
composio_toolset = ComposioToolSet()
17
18
# Get tools for RAG operations
19
rag_tools = composio_toolset.get_tools(apps=[App.RAG])
20
21
# Define the RAG Agent
22
rag_agent = Agent(
23
    role="RAG Knowledge Base Manager",
24
    goal="Manage a knowledge base using RAG tools",
25
    backstory="""You are an expert in managing knowledge bases and retrieving information.
26
    Your job is to add content to the knowledge base and retrieve answers to user queries.
27
    You use RAG (Retrieval-Augmented Generation) tools to efficiently store and retrieve information.""",
28
    verbose=True,
29
    allow_delegation=False,
30
    tools=rag_tools,
31
    llm=llm
32
)
33
34
# Define task for adding content
35
add_content_task = Task(
36
    description="""Add the following information to the knowledge base:
37
    
38
    1. Paris is the capital city of France.
39
    2. London is the capital city of the United Kingdom.
40
    3. Washington D.C. is the capital city of the United States.
41
    4. Tokyo is the capital city of Japan.
42
    5. Berlin is the capital city of Germany.
43
    
44
    Make sure to confirm that each piece of information is successfully added.""",
45
    expected_output="Confirmation that all information was added to the knowledge base",
46
    agent=rag_agent
47
)
48
49
# Define task for querying
50
query_task = Task(
51
    description="""Query the knowledge base to answer the following question:
52
    "What is the capital of France?"
53
    
54
    Return the answer with any supporting information from the knowledge base.""",
55
    expected_output="The answer to the question based on the knowledge base",
56
    agent=rag_agent,
57
    context=[add_content_task]  # This task depends on content being added first
58
)
59
60
# Create a crew with the agent and tasks
61
crew = Crew(
62
    agents=[rag_agent],
63
    tasks=[add_content_task, query_task],
64
    verbose=2,
65
    process=Process.sequential  # Tasks must run in order
66
)
67
68
# Execute the workflow
69
result = crew.kickoff()
70
print(textwrap.fill(f"Final Result: {result}", width=80))