ContractEx: Modern Contract Intelligence for Python

LLM-powered contract analysis | CUAD taxonomy | Risk detection | Privacy-first

ContractEx is a production-ready Python library for intelligent contract analysis using large language models. Extract clauses, identify parties, analyze risks, and extract financial terms from legal documents with a clean, intuitive API.

---

Features

• ** Simple API**: Extract contracts with a single line of code
• ** Multi-LLM Support**: OpenAI (GPT-4o), Anthropic (Claude), Google (Gemini), local models (Llama via Ollama)
• ** CUAD Taxonomy**: 41 standard clause types from the Contract Understanding Atticus Dataset
• ** Risk Analysis**: Automatic detection of unfavorable terms and potential risks
• ** Financial Extraction**: Extract payment terms, amounts, and conditions
• ** Privacy-First**: Local LLM support for sensitive documents
• ** Named Entity Recognition**: Extract parties, dates, and legal entities using spaCy/Blackstone
• ** Dataset Loaders**: Built-in access to ACORD, CUAD, and LePaRD benchmarks
• ** Extensible**: LangChain and spaCy compatibility
• ** Export**: JSON, Excel, CSV output formats
• ** Fast**: Batch processing with parallel execution
• ** Type-Safe**: Full type hints and Pydantic models

---

Installation

Quick Install

# Clone repository
git clone https://github.com/aahepburn/Contract-Clause-Extractor.git
cd Contract-Clause-Extractor

# Install all dependencies (single requirements file)
pip install -r requirements.txt

# Or install as editable package
pip install -e .

Using pyproject.toml (Optional Feature Groups)

# Install specific feature groups
pip install -e ".[ocr]"        # OCR support for scanned PDFs
pip install -e ".[spacy]"      # Named Entity Recognition
pip install -e ".[langchain]"  # LangChain integration
pip install -e ".[local]"      # Local LLM support (Ollama)
pip install -e ".[storage]"    # PostgreSQL storage
pip install -e ".[datasets]"   # Dataset loaders (ACORD, CUAD, LePaRD)
pip install -e ".[all]"        # All features

Configuration

# Create .env file with your API keys
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=your-google-api-key

---

Quick Start

Basic Usage (< 10 lines)

from contractex import extract_contract

# Extract contract with one line
contract = extract_contract("contract.pdf")

# Access results
print(f"Parties: {', '.join([p.name for p in contract.parties])}")
print(f"Clauses: {len(contract.clauses)}")
print(f"Risks: {len(contract.risks)} ({len(contract.critical_risks)} critical)")

# Export
contract.to_json("output.json")
contract.to_excel("output.xlsx")

Advanced Usage

from contractex import ContractExtractor
from contractex.llm import OpenAIProvider
from contractex.loaders import PDFLoader
from contractex.chunking import ClauseAwareChunker

# Configure custom components
llm = OpenAIProvider(model="gpt-4o", temperature=0.0)
loader = PDFLoader(ocr_enabled=True, preserve_layout=True)
chunker = ClauseAwareChunker(max_chunk_size=4000, overlap=200)

# Create extractor
extractor = ContractExtractor(
    llm_provider=llm,
    document_loader=loader,
    chunking_strategy=chunker,
    confidence_threshold=0.8
)

# Extract with options
contract = extractor.extract(
    "complex_contract.pdf",
    analyze_risks=True,
    extract_financial=True
)

---

Dataset Loading

Load popular legal contract datasets for training and evaluation:

from contractex.data import load_cuad, load_acord, load_lepard

# Load CUAD (Contract Understanding Atticus Dataset)
cuad_df = load_cuad(split='train')
print(f"Loaded {len(cuad_df)} contracts with 41 clause types")

# Load ACORD (clause retrieval benchmark)
acord_df = load_acord(split='train')

# Load LePaRD (legal passage retrieval)
lepard_df = load_lepard()

See contractex/data/README.md for full documentation.

---

Use Cases

Legal Teams

• Contract Review & Due Diligence
• Risk Assessment & Compliance
• M&A Document Analysis

Procurement Teams

• Vendor Agreement Review
• Payment Terms Verification
• SLA Analysis

Sales & Business Development

• Deal Analysis & Redlining Support
• Contract Comparison
• Archive Search

---

LLM Providers

• OpenAI (GPT-4o): Best accuracy (~$0.025/contract)
• Anthropic (Claude): Large documents (~$0.030/contract)
• Google (Gemini): Fast and cost-effective (~$0.002/contract)
• Local (Llama): Privacy-first, zero cost

---

Documentation & Examples

• CHANGELOG.md - Version history and release notes
• Examples Directory - Ready-to-run examples:
  • basic_extraction.py - Simple usage
  • advanced_extraction.py - Custom configuration
  • batch_processing.py - Multiple contracts
  • langchain_integration.py - LangChain usage
  • local_llm_example.py - Privacy-first local
  • fastapi_service.py - REST API
  • dataset_loading.py - Working with legal datasets
  • ner_example.py - Named entity recognition
  • storage_example.py - PostgreSQL persistence

Run examples: python examples/basic_extraction.py

---

Testing & Development

# Run all tests
pytest

# With coverage
pytest --cov=contractex --cov-report=html

# Code quality
black contractex/           # Format code
ruff check contractex/ --fix  # Lint
mypy contractex/             # Type check

---

Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

---

License

Apache 2.0 License - see LICENSE for details.

---