◈ Local Contract Analyzer

A privacy-first contract analysis tool that runs 100% offline using Microsoft Foundry Local. Upload a PDF contract and get a plain-English breakdown — risk level, red flags, key terms, important dates, and questions to ask before signing. Nothing ever leaves your device.

🎯 Overview

This project demonstrates how to build a private AI-powered contract analyzer using Foundry Local. All processing happens on your local machine — no internet connection required after setup, no data sent to external servers, and no information used for model training.

✨ Features

• 100% Private: All data stays on your device
• Offline Operation: No internet required after initial model download
• PDF Support: Upload contracts directly as PDF files
• Structured Analysis: Risk banner, red flags, key terms, dates, numbers, recommendation, and questions to ask
• No Data Training: Your contracts are never used to train AI models
• Clean UI: Minimal, demo-ready interface

🛠️ Tech Stack

• Foundry Local: Microsoft's local AI runtime (OpenAI-compatible API)
• phi-4-mini: Fast, lightweight model optimised for document analysis
• Streamlit: Web application frontend
• pymupdf (fitz): PDF parsing
• Python: Application runtime

💻 Requirements

• OS: macOS (Apple Silicon recommended), Windows 10/11, or Linux
• RAM: 8GB minimum, 16GB recommended
• Disk: At least 5GB free space
• Python: 3.8 or higher
• Internet: Required once to download the model — fully offline after that

🚀 Getting Started

1. Install Foundry Local

macOS:

brew tap microsoft/foundrylocal
brew install foundrylocal

Windows:

winget install Microsoft.FoundryLocal

After installing, verify it works:

foundry --version

---

2. Download the model (one-time, requires internet)

foundry model run phi-4-mini

Wait until you see the service is running ✅

This downloads phi-4-mini (~3.7GB) and caches it locally. You only need to do this once — after that it runs completely offline.

---

3. Clone this repository

git clone https://github.com/ifiecas/contract-analyzer.git
cd contract-analyzer

---

4. Set up a virtual environment (recommended)

python3 -m venv venv
source venv/bin/activate      # macOS/Linux
venv\Scripts\activate         # Windows

---

5. Install dependencies

pip install -r requirements.txt

---

▶️ Running the App

Every time you want to use the app, you need two terminal tabs:

Terminal 1 — Start the AI model:

foundry model run phi-4-mini

Wait until you see the service is running ✅

Terminal 2 — Launch the app:

cd contract-analyzer
bash start.sh

Then open your browser to the local URL shown in the terminal (usually http://localhost:8501).

Two commands. That's the entire setup.

---

🏗️ How It Works

┌─────────────────────────────┐
│        Your Machine         │
│                             │
│  ┌─────────────────────┐   │
│  │   Foundry Local      │   │  ← Serves phi-4-mini locally
│  │  (localhost API)     │   │  ← OpenAI-compatible endpoint
│  └─────────────────────┘   │
│            ↕                │
│  ┌─────────────────────┐   │
│  │  Contract Analyzer   │   │  ← Streamlit frontend
│  │     (app.py)         │   │
│  └─────────────────────┘   │
└─────────────────────────────┘
         ↕
   Your PDF Contract
   (Never leaves your device)

The app connects to Foundry Local via a local OpenAI-compatible API running on localhost. Your contract is sent to the model running on your own machine — no external servers involved.

---

📋 What You Get

Upload a PDF contract and the app returns:

• Risk Level → High / Medium / Low with a plain-English summary
• What is this contract → 2-3 sentence overview
• Who is involved → Parties and their obligations
• Red Flags → Issues ranked by severity with suggestions
• Terms Explained → Legal jargon in plain English
• Key Dates & Numbers → Important figures at a glance
• Recommendation → Sign, negotiate, or avoid
• Questions to Ask → Before you sign

📄 Supported Contract Types

• Employment agreements
• Service contracts
• NDAs
• Lease agreements
• Vendor agreements
• Partnership agreements
• License agreements

🔧 Troubleshooting

"Could not connect to Foundry Local" → Make sure foundry model run phi-4-mini is running in a separate terminal tab first.

"streamlit: command not found" → Use bash start.sh instead of calling streamlit directly. This activates the correct environment.

App is slow on first run → The model takes a moment to load into memory. Subsequent analyses in the same session will be faster.

Model not found → Run foundry model list to confirm phi-4-mini is downloaded. If not, run foundry model run phi-4-mini again with an internet connection.

---

🔐 Privacy & Security

• No Cloud Processing: Everything runs locally on your machine
• No Telemetry: No usage data collected
• No Model Training: Your data never trains the AI
• Offline After Setup: Internet only needed for first-time model download

---

📧 Contact

Portfolio: ifiecas.com

Blog post: A Private Contract Analyzer That Never Shares Your Data

---

⚠️ Disclaimer: This tool is intended for educational and experimental purposes only and is a demonstration of Foundry Local's capabilities for running AI models privately on your local machine.

This tool does not provide legal advice. The analysis generated by this application should not be relied upon for legal decisions. Always consult a qualified attorney or legal professional for advice on contracts or any legal matters. The author assumes no liability for any decisions made based on the output of this tool.