PDFs are a common format for documents that contain rich information in a structured format. The Opper SDK provides powerful capabilities to extract and process content from PDFs, making it accessible for analysis, transformation, and integration into your applications.

With the Opper API, you can define tasks in a declarative manner, where you specify input data (PDF documents), output structures and optional configuration on how to complete the task. The Opper API returns completions as specified and builds optimal prompts for completing the task.

​
Extracting structured information

You can use LLMs to extract structured data from PDF documents. This can be useful to make the content of PDFs programmable and searchable.

Here is a simple example of extracting structured information from a PDF document:

import base64
import os
import sys
from opperai import Opper
from pydantic import BaseModel, Field

opper = Opper(os.getenv("OPPER_API_KEY"))

class InvoiceData(BaseModel):
    invoice_number: str = Field(description="The invoice number")
    total_amount: float = Field(description="The total amount in dollars")
    vendor_name: str = Field(description="The name of the vendor or company")
    date: str = Field(description="The invoice date")
    items: list = Field(description="List of items and their prices")

def extract_invoice_data(pdf_path: str) -> dict:
    media_input = f"data:application/pdf;base64,{base64.b64encode(open(pdf_path, 'rb').read()).decode('utf-8')}"
    response = opper.call(
        name="extract_invoice",
        instructions="Extract structured invoice data from the PDF document",
        input={
            "_opper_media_input": media_input,
        },
        output_schema=InvoiceData.model_json_schema()
    )
    return response.json_payload

def main():
    if len(sys.argv) < 2:
        print("Usage: python extract_invoice.py <path_to_pdf>")
        return

    pdf_path = sys.argv[1]
    result = extract_invoice_data(pdf_path)
    print(result)

if __name__ == "__main__":
    main()

# Output:
# {
#     'invoice_number': 'INV-2024-001',
#     'total_amount': 1250.00,
#     'vendor_name': 'Tech Solutions Inc.',
#     'date': '2024-01-15',
#     'items': [
#         {'item': 'Software License', 'price': 1000.00},
#         {'item': 'Support Services', 'price': 250.00}
#     ]
# }