Contract Data Extraction Tool

A Python command-line tool that uses LLMs to extract structured information from contract documents and outputs the data as JSON.

Overview

This tool processes contract text files and extracts key information including:

• Contract summary and type
• Parties involved (with addresses and roles)
• Effective dates and duration
• Contract scope
• Important clauses (categorized by type)
• Governing law
• Total contract value

The extraction uses Azure AI Foundry's GPT-oss-120b model via LangChain with structured output to ensure consistent, schema-validated results.

Features

• Structured Extraction: Uses Pydantic models to ensure consistent output format
• Smart Date Parsing: Handles various date formats and converts to ISO 8601
• Clause Categorization: Automatically categorizes clauses into predefined types
• ISO Standards: Uses ISO 3166-1 alpha-2 country codes and ISO 8601 duration format
• Token Counting: Reports token usage for each contract processed
• Automatic Directory Creation: Creates output directory if it doesn't exist

Installation

1. Clone this repository or download the files

2. Install dependencies:

pip install -r requirements.txt

3. Set up your environment variables:
   • Copy .env.example to .env
   • Fill in your Azure AI Foundry credentials

Configuration

Create a .env file with the following variables:

FOUNDRY_MODEL_ENDPOINT=https://your-foundry-endpoint.azure.com/openai/v1
OPENAI_API_KEY=your-api-key-here
DEPLOYMENT_NAME=gpt-oss-120b

See .env.example for a template.

Usage

Basic Usage

python import_contract.py --input path/to/contract.txt

Example

python import_contract.py --input "contracts/my_contract.txt"

Help

python import_contract.py --help

Output

The tool generates JSON files in the output/ directory with the naming pattern:

{filename}_extracted.json

Output Schema

{
  "summary": "Contract summary...",
  "contract_type": "Service",
  "parties": [
    {
      "name": "Company Name",
      "location": {
        "address": "123 Street",
        "city": "City",
        "state": "State",
        "country": "US"
      },
      "role": "Provider"
    }
  ],
  "effective_date": "2023-01-01",
  "contract_scope": "Description of what the contract covers...",
  "duration": "P1Y",
  "end_date": "2024-01-01",
  "total_amount": 100000.00,
  "governing_law": {
    "country": "US"
  },
  "clauses": [
    {
      "summary": "Clause summary...",
      "clause_type": "Renewal & Termination"
    }
  ],
  "file_id": "contract_filename"
}

Clause Types

The tool categorizes clauses into the following types:

• Renewal & Termination: Contract duration, renewal terms, termination conditions
• Confidentiality & Non-Disclosure: Protection of confidential information, NDAs
• Non-Compete & Exclusivity: Non-compete agreements, exclusivity provisions
• Liability & Indemnification: Liability limits, indemnification, warranties, penalties
• Service-Level Agreements: Performance metrics, service standards, quality requirements

Contract Types

Supported contract types include:

• Affiliate Agreement
• Development
• Distributor
• Endorsement
• Franchise
• Hosting
• IP
• Joint Venture
• License Agreement
• Maintenance
• Manufacturing
• Marketing
• Non Compete/Solicit
• Outsourcing
• Promotion
• Reseller
• Service
• Sponsorship
• Strategic Alliance
• Supply
• Transportation

Requirements

• Python 3.8+
• Azure AI Foundry access with GPT-oss-120b deployment
• See requirements.txt for Python package dependencies

Troubleshooting

Authentication Errors

If you see authentication errors, verify:

1. Your .env file exists and contains valid credentials
2. The FOUNDRY_MODEL_ENDPOINT URL is correct
3. The OPENAI_API_KEY is valid and not expired

File Not Found

Ensure the input file path is correct and the file exists. Use quotes around paths with spaces:

python import_contract.py --input "contracts/file with spaces.txt"

Poor Extraction Quality

For better results:

• Ensure the input file is clean text (not scanned images)
• Remove excessive formatting or special characters
• Verify the contract is in English (or adjust prompts for other languages)

License

This project is provided as-is for contract analysis purposes.

Contributing

To improve extraction quality, you can:

1. Adjust field descriptions in the Pydantic models
2. Refine the extraction prompt in the process_contract() function
3. Add additional clause types or contract types to the enums

Credit

The code in this example repo is based on an approach to processing CUAD_v1 posted by Tomaz Bratanic Blog Post on his blog.