⚖️ LexGuard — Enterprise AI Contract Intelligence

Demystifying legal agreements through multi-pass AI analysis, vector-grounded RAG, and executive-grade precision UX.

---

📖 Overview

People and businesses sign contracts every day without fully understanding the underlying liabilities. Legal language is deliberately nuanced, lengthy, and opaque. LexGuard transforms dense contracts (PDFs & DOCX) into transparent, risk-assessed executive summaries in seconds.

Built with a high-performance FastAPI backend, a state-of-the-art Google Gemini GenAI SDK pipeline, local ChromaDB vector embedding for real-time RAG chat, and a premium Next.js frontend inspired by institutional financial terminals, LexGuard empowers signees to negotiate with absolute confidence.

---

✨ Core Features

🔍 1. Multi-Pass Contract Deconstruction

• Structural Ingestion: Directly parses complex PDF and DOCX files using robust text extraction algorithms (PyMuPDF / python-docx).
• Pass 1 — Structural Extraction: Deconstructs contracts into distinct legal clauses, identifying document types, governing laws, and calculating initial suspicion scores (0.0 to 10.0) using structured Pydantic schema outputs.

🛡️ 2. Deep Risk Taxonomy & Plain English Translation

• Pass 2 — Granular Risk Analysis: Concurrently evaluates flagged clauses across standard legal categories (IP transfer, non-compete, indemnification, termination, arbitration).
• Plain English Breakdown: Translates legalese into 2-3 sentence non-lawyer summaries.
• Direct Consequences: Generates hard-hitting "If you sign this..." consequence statements.
• Verbatim Red Flags: Extracts precise 3-5 word high-risk phrases directly from the document text.
• Actionable Negotiation Tips: Recommends concrete carve-outs, time limits, and scope restrictions for counter-proposals.

🤖 3. Grounded RAG Chat Assistant

• Local Embedding: Automatically chunks contract text and generates dense semantic vector embeddings via sentence-transformers/all-MiniLM-L6-v2.
• ChromaDB Vector Store: Houses document embeddings locally for sub-millisecond retrieval.
• SSE Real-Time Streaming: Streams answers in real-time through Server-Sent Events (SSE), strictly grounded in document context with verbatim clause citations (e.g., [Clause: Governing Law]).

🏛️ 4. Enterprise Persistence & Cloud Infrastructure

• Google Cloud Firestore: Seamlessly persists session states, parsed contract structures, risk report snapshots, and chat history.
• Firebase Cloud Storage: Safely archives uploaded contract artifacts for secure retrieval and auditability.

🎨 5. Precision Terminal UX

• Bloomberg / Surgeon Table Aesthetic: Information-dense, high-contrast typography and layout with zero fluff.
• Curated Severity Dashboard: Instantly highlights Critical, High, Medium, and Low risks with custom color-coded badges and filtering.
• Cinematic Micro-Animations: Smooth layout transitions powered by GSAP and subtle interactive Three.js WebGL background canvases.

---

🏗️ System Architecture

flowchart TB
    subgraph Client [Browser / Client]
        UI[Next.js Premium Terminal UI]
        ChatUI[Real-time SSE Chat Panel]
    end

    subgraph Backend [FastAPI Backend]
        API_Upload[POST /upload]
        API_Analyze[POST /analyze]
        API_Chat[POST /chat SSE Stream]
        
        Parser[Document Extraction\nPyMuPDF / python-docx]
        Chunker[Clause Splitter & Regex Chunker]
        Embedder[MiniLM Embedder & ChromaDB]
        
        Pass1[Pass 1 AI: Structural Ingestion\nGoogle Gemini GenAI SDK]
        Pass2[Pass 2 AI: Concurrent Risk Analysis\nGoogle Gemini GenAI SDK]
        Agg[Pass 3 AI: Executive Aggregation]
    end

    subgraph Cloud [Google Cloud Platform]
        Firestore[(Firestore Session & Risk Store)]
        Storage[(Firebase Contract Bucket)]
    end

    UI -- "1. Upload Contract" --> API_Upload
    API_Upload --> Parser
    Parser --> Chunker
    Chunker --> Embedder
    Parser --> Pass1
    API_Upload --> Storage
    Pass1 --> Firestore
    
    UI -- "2. Run Risk Analysis" --> API_Analyze
    API_Analyze --> Pass2
    Pass2 --> Agg
    Agg --> Firestore
    
    ChatUI -- "3. Query Specific Clause" --> API_Chat
    API_Chat <--> Embedder
    API_Chat -- "Stream SSE" --> ChatUI

Loading

---

🚀 Quickstart & Installation

📋 Prerequisites

• Python 3.11+
• Node.js 20+ & npm
• Google Gemini API Key (GEMINI_API_KEY)
• Google Cloud / Firebase Project (Firestore & Firebase Storage enabled)

---

⚙️ 1. Backend Setup

# Navigate to the backend directory
cd backend

# Create and activate a Python virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Environment Variables (backend/.env)

Create a .env file in the backend/ directory:

# Mode configuration
# DEMO_MODE=true bypasses live Gemini/Firebase calls for instant offline testing
DEMO_MODE=false

# Gemini AI Credentials
GEMINI_API_KEY=your_actual_gemini_api_key_here

# Firebase / GCP Project Configuration
GOOGLE_CLOUD_PROJECT=your-gcp-project-id
FIREBASE_STORAGE_BUCKET=your-project.appspot.com

# Optional: Path to Firebase service account JSON key (for local development)
GOOGLE_APPLICATION_CREDENTIALS=/path/to/serviceAccountKey.json

Run the Backend Server

uvicorn main:app --reload --port 8000

The FastAPI documentation will be instantly accessible at http://localhost:8000/docs.

---

💻 2. Frontend Setup

# Navigate to the frontend directory
cd frontend

# Install Node dependencies
npm install

Environment Variables (frontend/.env.local)

Create a .env.local file in the frontend/ directory:

NEXT_PUBLIC_API_URL=http://localhost:8000

Run the Frontend Development Server

npm run dev

Open http://localhost:3000 in your browser to experience LexGuard.

---

📡 Key API Endpoints

Method	Endpoint	Description
GET	/health	API health check and version status.
POST	/upload	Ingests PDF/DOCX contract, runs Pass 1 structural extraction, indexes into ChromaDB, and returns clauses.
POST	/analyze	Executes Pass 2 concurrent risk analysis and Pass 3 overall contract executive aggregation.
POST	/chat	SSE streaming endpoint for real-time document-grounded RAG query and response.
GET	/report/{session_id}	Retrieves complete persisted contract analysis session and risk report from Firestore.
GET	/chat/{session_id}/history	Retrieves historical chat messages for a specific contract session.

---

🧪 Demo / Sandbox Mode

For testing or development without consuming Gemini API tokens or requiring Firebase setup, LexGuard includes a robust Demo Mode.

In backend/.env:

DEMO_MODE=true

When enabled, LexGuard bypasses external network requests and instantly serves highly realistic, pre-computed contract intelligence data models for seamless UI and workflow evaluation.

---

🛡️ Security & Privacy Notice

LexGuard is designed with enterprise security in mind:

• Ephemeral Processing: In-memory vector embedding ensures that chunked data resides locally inside ChromaDB.
• Strict Pydantic Enforcement: All LLM outputs are rigorously validated against strict Pydantic schemas, eliminating prompt injection anomalies and malformed responses.
• No Model Training: Contracts processed through the Gemini API under standard enterprise terms are not used to train underlying foundation models.

---

⚖️ LexGuard — Empowering clarity in every agreement.