ContractLens Documentation

Welcome to the documentation for ContractLens, a system for automated compliance analysis of PDF contracts.

This project parses PDF contracts, builds a vectorstore of their contents, and runs compliance analysis using a language model (Ollama / llama3) with pre-defined compliance rules.

Contents:

Getting Started

Requirements:

  • Python >= 3.11

  • Redis server running locally

  • Celery for background task processing

  • PDFplumber, langchain-ollama, chromadb, flask, pydantic

Installation:

# Create the environment from the YAML
conda env create -f environment.yml

# Activate it
conda activate <env_name_from_yaml>

Running the project:

# Start Redis
redis-server

# Start Celery worker
celery -A celery_app worker --loglevel=info

# Start Flask backend
flask run --port=5000

Then open your browser at http://localhost:5000

—

Project Structure

The repository has two main components: backend and frontend.

  • backend/

    • analyzer.py → Runs compliance analysis using LLM and vectorstore

    • app.py → Flask API for upload, status, and chat

    • celery_app.py → Celery configuration

    • parser.py → Parses PDF pages and tables

    • prompts.py → Defines compliance questions and analysis prompt template

    • redis_store.py → Simple Redis-based job store

    • schemas.py → Pydantic models for validation

    • tasks.py → Celery tasks for end-to-end pipeline

    • vectorstore.py → Builds Chroma vectorstore from PDF text and tables

  • frontend/

    • index.html → File upload page

    • chat.html → Chat interface to query contract

—

API Overview

The backend exposes the following endpoints:

  • GET / → Returns the upload page

  • POST /upload → Upload a PDF, returns a job_id

  • GET /status/<job_id> → Check status (queued, parsing, indexing, analyzing, done)

  • GET /vectorstore_preview/<job_id> → Preview first chunks of the vectorstore

  • GET/POST /chat/<job_id> → Ask questions about a processed contract

—

Modules