AI Agents
Document Processing

Building a Document Processing Agent

This example shows how to create a custom CrewAI tool that processes documents and extracts structured data:

from typing import Type
from crewai.tools import BaseTool
from pydantic import BaseModel, Field
from crewai_tools import FileWriterTool

class DocumentProcessorInput(BaseModel):
    """Input schema for Document Processor Tool."""
    url: str = Field(..., description="URL of the document to process")
    output_file: str = Field(..., description="Output JSON file path")

class DocumentProcessorTool(BaseTool):
    name: str = "Document Processor"
    description: str = "Processes documents and saves structured data to JSON"
    args_schema: Type[BaseModel] = DocumentProcessorInput

    def _run(self, url: str, output_file: str) -> str:
        # Process document using our API
        api_response = requests.post(
            "https://api.example.com/process",
            json={"url": url}
        )

        # Save results using FileWriterTool
        writer = FileWriterTool()
        writer._run(
            output_file,
            api_response.json(),
            "processed_documents"
        )

        return f"Document processed and saved to {output_file}"

# Usage in CrewAI
from crewai import Agent, Task, Crew

# Create agent with our tool
agent = Agent(
    role="Document Processor",
    goal="Process documents and extract structured data",
    tools=[DocumentProcessorTool()]
)

# Create task
task = Task(
    description="Process document and save results",
    agent=agent
)