Document ORM

Extract Thinker

An ORM (Object-Relational Mapping) style library for extracting structured data from documents using Large Language Models. Interact with documents using Pydantic models, just like you interact with databases.

🧠 What is Extract Thinker?

Extract Thinker abstracts away the complexity of prompt engineering and document loading. It allows you to define Contracts (your data schema) and lets the library handle the extraction, classification, and validation against multiple document sources (PDFs, Images, etc.).

⚡

ORM Style

Pydantic -> Docs

🔄

Multi-Loader

OCR, Azure, AWS

🏷️

Classification

Auto-route files

🧱

LLM Agnostic

Any Model

✨ Key Capabilities

Pydantic Contracts

Define extraction schema using standard Pydantic models. The library guarantees output matches this structure.

Flexible Loaders

Swap between OCR engines (Tesseract, Azure, AWS Textract, Google DocAI) without changing extraction logic.

Document Classification

Automatically classify documents (e.g., "Invoice" vs "Driver License") and route to the correct extractor.

LLM Agnostic

Works with OpenAI, Anthropic, Cohere, and local models via LiteLLM. Optimize for cost or accuracy.

💻 Code Example

from extract_thinker import DocumentLoaderPyPdf, Extractor, Contract
from pydantic import Field

# 1. Define your Data Model (The Contract)
class InvoiceContract(Contract):
    invoice_number: str = Field(description="The unique number of the invoice")
    date: str = Field(description="The date of the invoice")
    total_amount: float = Field(description="The total amount to be paid")
    vendor_name: str = Field(description="Name of the vendor/company")

# 2. Setup Loader and Extractor
loader = DocumentLoaderPyPdf()
extractor = Extractor()
extractor.load_document(loader, "invoice.pdf")

# 3. Extract!
result = extractor.extract(InvoiceContract)

print(f"Invoice #: {result.invoice_number}")
print(f"Total: ${result.total_amount}")

🎯 Ideal Use Cases

🏦

Financial Parsing

Extracting data from complex Invoices, Receipts, and Financial Statements.

🆔

KYC Automation

Extracting fields from ID cards, Passports, and Driver Licenses.

⚖️

Legal Contracts

Extracting key clauses, dates, and parties from legal agreements.

📚 Resources

GitHub Repository

Star the project

PyPI Package

pip install extract-thinker

Test Your Knowledge

Score 8/10 or higher to pass