doc-intelligence 0.0.7

Document AI - Intelligent document processing and extraction

Document AI

Documentation: https://zeel-04.github.io/document-ai/

A library for parsing, formatting, and processing documents that can be used to build AI-powered document processing pipelines with structured data extraction and citation support.

Features

• Extract structured data from PDF documents using LLMs
• Automatic citation tracking with page numbers, line numbers, and bounding boxes
• Support for digital PDFs
• Type-safe data models using Pydantic
• OpenAI integration with support for reasoning models

Installation

Requirements

• Python >= 3.10
• OpenAI API key

Install uv

First, install uv if you haven't already:

curl -LsSf https://astral.sh/uv/install.sh | sh

Install from Source

Clone the repository and install the package:

git clone https://github.com/zeel-04/document-ai.git
cd document-ai
uv sync

Install from Git (Alternative)

You can also install directly from the git repository:

uv pip install git+https://github.com/zeel-04/document-ai.git

Quick Start

Set up your OpenAI API key:

echo "OPENAI_API_KEY=your-api-key-here" > .env

Here's a simple example to extract structured data from a PDF:

from dotenv import load_dotenv
from document_ai.processer import DocumentProcessor
from document_ai.llm import OpenAILLM
from pydantic import BaseModel

# Load environment variables
load_dotenv()

# Initialize the LLM
llm = OpenAILLM()

# Create a processor from a PDF file
processor = DocumentProcessor.from_digital_pdf(
    uri="path/to/your/document.pdf",
    llm=llm,
)

# Define your data model
class Balance(BaseModel):
    ending_balance: float

# Configure extraction with citations
config = {
    "response_format": Balance,
    "llm_config": {
        "model": "gpt-5",
        "reasoning": {"effort": "minimal"},
    },
    "extraction_config": {
        "include_citations": True,
        "extraction_mode": "single_pass",
        "page_numbers": [0, 1],  # Optional: specify which pages to process
    }
}

# Extract structured data
response = processor.extract(config)

# Get the extracted data and citations
data, citations = response
print(f"Extracted data: {data}")
print(f"Citations: {citations}")

Sample Output

The extract method returns a tuple containing the extracted data and citation information:

(Balance(ending_balance=111.61),
 {'ending_balance': {'value': 111.61,
   'citations': [{'page': 0,
     'bboxes': [{'x0': 0.058823529411764705,
       'top': 0.6095707475757575,
       'x1': 0.5635455037254902,
       'bottom': 0.6221969596969696}]}]}})

Documentation

For more detailed documentation, see the docs directory or visit the documentation site.

Development Setup

Prerequisites:

• Python 3.10+
• uv

git clone https://github.com/zeel-04/document-ai.git
cd document-ai
uv venv
uv sync