legal_deed_review_system / HF_README.md
Pial2233's picture
Commit with fix
e089c13 verified
|
raw
history blame
8.26 kB
metadata
title: Legal Deed Reviewer
emoji: ⚖️
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.0.0
app_file: app.py
pinned: false
license: apache-2.0
tags:
  - mcp
  - mcp-in-action-productivity
  - legal
  - document-analysis
  - gradio
  - ai
  - legal-tech
  - property
  - deed-analysis

⚖️ Legal Deed Reviewer

AI-powered legal deed analysis using MCP (Model Context Protocol) servers

Upload property deed documents to receive comprehensive risk assessments, clause-by-clause breakdowns, and plain-language explanations of legal issues.

🎯 Overview

Legal Deed Reviewer is an intelligent system that helps property buyers, landlords, lawyers, and mortgage teams understand the risks and implications in legal deed documents. Built for the MCP-1st-Birthday Hackathon, this project demonstrates advanced MCP integration with multi-tool orchestration for legal document analysis.

What This Tool Does:

Deed Classification - Automatically identifies deed type (sale, mortgage, lease, gift, warranty, quitclaim)
Metadata Extraction - Extracts parties, property details, consideration, and jurisdiction
Clause Breakdown - Splits deeds into logical sections and clauses
Risk Analysis - Identifies legal risks with severity levels (LOW/MEDIUM/HIGH)
Plain-Language Explanations - Translates legal jargon into understandable language
Actionable Recommendations - Suggests next steps and areas requiring legal consultation

🚀 Quick Start

  1. Upload a PDF - Click "Upload Deed (PDF)" and select your property deed document
  2. Click "Analyze Deed" - The system will process your document (takes 10-30 seconds)
  3. Review Results - Navigate through tabs:
    • Overview: Deed metadata and quick stats
    • Clause Breakdown: All identified clauses with categorization
    • Risk Analysis: Clause-by-clause risk assessment with explanations
    • Extracted Text: Full text from the deed
  4. Download Report - Get a Markdown report for your records

Sample Deed Included

Try the system with our sample: usa_general_warranty_deed_sample.pdf (included in the Space)

🏗️ Architecture

MCP Integration

This project uses 5 MCP tools that work together to provide comprehensive deed analysis:

1. PDF Text Extraction

extract_text_from_deed_pdf(pdf_path: str) -> JSON
  • Direct text extraction from PDFs using PyMuPDF
  • OCR fallback for scanned documents
  • Returns full text + page-by-page breakdown

2. Clause Splitting

split_deed_into_clauses(text: str) -> JSON
  • Pattern-based clause detection
  • Identifies common deed sections (WITNESSETH, WHEREAS, NOW THEREFORE)
  • Categorizes clause types

3. Deed Classification

classify_deed_type(deed_text: str) -> JSON
  • LLM-powered deed type identification
  • Extracts jurisdiction, parties, property details
  • Structured JSON output

4. Risk Analysis

analyze_deed_risks(clauses: str, classification: str) -> JSON
  • Clause-by-clause risk assessment
  • Categories: TITLE, WARRANTY, ENCUMBRANCE, EASEMENT, RESTRICTION
  • Risk levels with explanations and recommendations

5. Comprehensive Report Generation

generate_comprehensive_deed_report(pdf_path: str) -> JSON
  • Orchestrates all tools in a pipeline
  • Returns complete analysis report
  • Single-command full analysis

Tech Stack

  • MCP Framework: Model Context Protocol for tool orchestration
  • Gradio 4: Web interface
  • FastAPI: REST API backend
  • Nebius Qwen2.5-VL-72B: Vision model for OCR
  • Meta Llama-3.3-70B: LLM for legal analysis
  • PyMuPDF: PDF processing

📊 Sample Output

Deed Classification

{
  "deed_type": "warranty",
  "jurisdiction": {
    "country": "United States",
    "state_province": "Illinois"
  },
  "key_parties": {
    "grantor": "Michael Austin Carter and wife Laura Jean Carter",
    "grantee": "Husband and wife Address: 7421 Meadowbrook Drive"
  },
  "consideration_amount": "$250,000.00"
}

Risk Analysis Example

RISK LEVEL: MEDIUM
CATEGORY: ENCUMBRANCE

EXPLANATION: The deed includes several exceptions and reservations that could 
affect the property's value and usability, including unpaid real estate taxes 
and existing easements.

RECOMMENDATION: Conduct a thorough title search to understand the full extent 
of encumbrances and consult with a real estate attorney to assess their impact.

🎓 Use Cases

For Property Buyers

  • Understand risks before closing
  • Identify unusual clauses
  • Know what questions to ask your lawyer

For Real Estate Lawyers

  • Quick first-pass review
  • Standardized risk assessment
  • Time-saving for routine deeds

For Mortgage Teams

  • Automated security deed screening
  • Risk flagging for approval workflow
  • Compliance checking

For Landlords

  • Lease deed analysis
  • Easement and restriction identification
  • Future resale impact assessment

⚠️ Legal Disclaimer

IMPORTANT: This tool provides analysis for informational purposes only and does not constitute legal advice.

  • Always consult with a qualified attorney licensed in your jurisdiction
  • Legal requirements vary by location
  • This tool cannot replace professional legal counsel
  • Use this as a starting point for discussion with your lawyer

🔧 How It Works

Multi-Step Reasoning Pipeline

The system uses intelligent multi-step reasoning:

  1. 📄 Text Extraction - Extracts text from PDF (direct or OCR)
  2. 🔍 Classification - Identifies deed type and jurisdiction
  3. ✂️ Clause Segmentation - Breaks document into logical sections
  4. ⚖️ Risk Scoring - Analyzes each clause for legal issues
  5. 📝 Report Generation - Compiles comprehensive analysis

MCP Tool Orchestration

All tools are MCP-compliant and can be called individually or chained:

# Example: Full analysis pipeline
report = generate_comprehensive_deed_report(pdf_path)

# Or: Individual tool calls
text = extract_text_from_deed_pdf(pdf_path)
classification = classify_deed_type(text)
clauses = split_deed_into_clauses(text)
risks = analyze_deed_risks(clauses, classification)

🏆 MCP-1st-Birthday Hackathon

This project was built for the MCP-1st-Birthday Hackathon in the Productivity Track.

Why This Project Uses MCP

  1. Modularity - Each legal analysis function is a separate MCP tool
  2. Composability - Tools can be chained for complex workflows
  3. Reusability - MCP tools work standalone or in pipelines
  4. Extensibility - Easy to add new analysis tools (RAG, jurisdiction-specific logic)
  5. Interoperability - Standard MCP interface for all tools

Future Enhancements

  • RAG System: Vector database with model clauses and legal precedents
  • Multi-Jurisdiction Support: Country-specific risk assessments
  • Clause Comparison: Visual diff against standard templates
  • Advanced Risk Scoring: ML-based risk prediction
  • Multi-MCP Architecture: Separate servers for PDF, RAG, and LLM

📚 Documentation

  • Main README: Project documentation
  • CLAUDE.md: AI assistant guidance for codebase
  • readme_main.md: Detailed project guidelines
  • main_project.md: Original requirements and roadmap

👥 Team

Built by the Legal-AI Team for MCP-1st-Birthday Hackathon:

  • Sojib: Frontend (Gradio UI, report export)
  • Pial & Sojib: MCP servers (PDF + RAG tools)
  • Takib: LLM orchestration and legal prompts

🔗 Links

📄 License

Apache-2.0 License - See LICENSE file for details


Made with ⚖️ for the MCP-1st-Birthday Hackathon

Empowering users to understand legal documents through AI