Tabular Review for Bulk Document Analysis

An AI-powered document review workspace that transforms unstructured legal contracts into structured, queryable datasets. Designed for legal professionals, auditors, and procurement teams to accelerate due diligence and contract analysis.

🚀 Features

• AI-Powered Extraction: Automatically extract key clauses, dates, amounts, and entities from PDFs using Google Gemini 2.5 Pro / 3.0.
• High-Fidelity Conversion: Uses Docling (running locally) to convert PDFs and DOCX files to clean Markdown text, preserving formatting and structure without hallucination.
• Dynamic Schema: Define columns with natural language prompts (e.g., "What is the governing law?").
• Verification & Citations: Click any extracted cell to view the exact source quote highlighted in the original document.
• Spreadsheet Interface: A high-density, Excel-like grid for managing bulk document reviews.
• Integrated Chat Analyst: Ask questions across your entire dataset (e.g., "Which contract has the most favorable MFN clause?").
• Real-Time Progress: Progress bar showing extraction status (X/Y cells completed).

🎬 Demo

TabularReview.Final.mp4

🏗 Architecture

┌─────────────────────────────────┐
│         React Frontend          │
│  (Vite + TypeScript + Tailwind) │
│                                 │
│  ┌─────────┐  ┌──────────────┐  │
│  │DataGrid │  │ Gemini SDK   │  │
│  │Sidebar  │  │ (extraction  │  │
│  │Chat     │  │  & chat)     │  │
│  └─────────┘  └──────┬───────┘  │
│                      │          │
│       Google Gemini API         │
└──────────┬──────────────────────┘
           │ /convert (file upload)
┌──────────▼──────────────────────┐
│      FastAPI Backend            │
│  (Python + Docling)             │
│                                 │
│  PDF/DOCX → Markdown conversion │
└─────────────────────────────────┘

Frontend (/): React 19 SPA. Handles the grid UI, Gemini API calls for extraction/chat (client-side via @google/genai SDK), and document viewing.

Backend (/server): Python FastAPI server running Docling for document conversion. Converts uploaded PDF/DOCX files to Markdown text. The frontend sends files here before storing them.

🛠 Tech Stack

• Frontend: React 19, TypeScript, Tailwind CSS, Vite
• AI Integration: Google GenAI SDK (Gemini 2.5 Flash, 2.5 Pro, 3.0 Pro)
• Backend: Python, FastAPI, Docling (document conversion)

📦 Getting Started

Prerequisites

• Node.js 18+ (with npm)
• Python 3.10+
• Google Gemini API Key — get one from Google AI Studio

1. Clone the repository

git clone https://github.com/astevens-lmds/Tabular_Review.git
cd Tabular_Review

2. Environment Variables

Copy the example file and add your API key:

cp .env.example .env

Edit .env and set:

VITE_GEMINI_API_KEY=your_google_api_key_here
VITE_API_URL=http://localhost:8000

Variable	Required	Description
VITE_GEMINI_API_KEY	Yes	Google Gemini API key for AI extraction and chat
VITE_API_URL	No	Backend URL (defaults to http://localhost:8000)

3. Setup Frontend

npm install

4. Setup Backend (Docling)

The backend is required for document conversion (PDF/DOCX → Markdown).

cd server
python3 -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install -r requirements.txt

5. Run

Start the backend (in one terminal):

cd server
source venv/bin/activate
python main.py
# Server runs at http://localhost:8000

Start the frontend (in another terminal):

npm run dev
# App runs at http://localhost:3000

🐳 Docker Deployment (Alternative)

cp .env.example .env
# Edit .env and add your Google Gemini API key

docker-compose up --build

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000
• API docs: http://localhost:8000/docs

🧪 Testing

Frontend (Vitest)

npx vitest run          # Run all tests once
npx vitest              # Watch mode

Tests live in tests/ and cover components (App, DataGrid, AddColumnMenu, KeyboardShortcutsHelp, ErrorBoundary, BatchUploadProgress), utilities (CSV/PDF export, column templates, theming), and type definitions.

Backend (pytest)

cd server
source venv/bin/activate
pytest test_main.py -v

Backend tests cover the /convert and /health endpoints, rate limiting, file size limits, CORS headers, and filename validation.

📁 Project Structure

├── App.tsx                    # Main application component
├── index.tsx                  # React entry point
├── types.ts                   # TypeScript type definitions
├── components/
│   ├── DataGrid.tsx           # Spreadsheet-like grid
│   ├── VerificationSidebar.tsx # Cell inspection & document viewer
│   ├── ChatInterface.tsx      # AI chat analyst
│   ├── AddColumnMenu.tsx      # Column creation/editing
│   ├── BatchUploadProgress.tsx # Batch upload progress overlay
│   ├── ColumnTemplateMenu.tsx # Pre-built column templates
│   ├── ErrorBoundary.tsx      # React error boundary
│   ├── KeyboardShortcutsHelp.tsx # Shortcuts modal
│   ├── ProjectManager.tsx     # Project save/load
│   └── Icons.tsx              # Icon re-exports from lucide-react
├── hooks/
│   ├── useTheme.ts            # Dark mode hook
│   └── useKeyboardShortcuts.ts # Keyboard shortcut handler
├── services/
│   ├── geminiService.ts       # Gemini API integration
│   ├── documentProcessor.ts   # Frontend → backend file conversion
│   ├── batchExport.ts         # Multi-format export
│   └── projectStore.ts        # LocalStorage project persistence
├── tests/                     # Vitest test suite
├── utils/
│   ├── sampleData.ts          # Built-in sample documents
│   └── columnTemplates.ts     # Column template definitions
├── server/
│   ├── main.py                # FastAPI backend
│   └── requirements.txt       # Python dependencies
├── .env.example               # Environment variable template
├── vite.config.ts             # Vite configuration
├── tsconfig.json              # TypeScript configuration
├── docker-compose.yml         # Docker setup
├── Dockerfile.frontend        # Frontend Docker image
└── Dockerfile.backend         # Backend Docker image

📖 API Documentation

The FastAPI backend includes auto-generated interactive API documentation:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc
• OpenAPI JSON: http://localhost:8000/openapi.json

Endpoints

Method	Path	Description
POST	/convert	Upload a document (PDF, DOCX, etc.) and receive Markdown text. Rate-limited to 30 req/min per IP.
GET	/health	Health check endpoint. Returns {"status": "ok"}.

🛡 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Disclaimer: This tool is an AI assistant and should not be used as a substitute for professional legal advice. Always verify AI-generated results against the original documents.