Regtech Video Compliance Checker

An AI-powered web application for automatically checking video compliance against regulatory guidelines using 100% open source models.

Features

• Video Content Extraction

  • Frame extraction with scene change detection (OpenCV + FFmpeg)
  • Audio transcription (Whisper - local)
  • Visual analysis (LLaVA/BLIP-2 + YOLO v8)
  • OCR text extraction (EasyOCR/PaddleOCR)

• Semantic Vectorization

  • Embedding generation (sentence-transformers)
  • Vector search (Weaviate)
  • Semantic matching against guidelines

• Compliance Checking

  • Guideline parsing and structuring (LangChain)
  • Automated compliance analysis (Llama 3.1 via Ollama)
  • Evidence-based violation detection (LangGraph agent)
  • Interactive timeline reports

Technology Stack

Backend

• FastAPI - Modern async web framework
• LangChain + LangGraph - AI orchestration
• Celery + Redis - Task queue for video processing
• PostgreSQL - Relational database
• MinIO - S3-compatible object storage
• Weaviate - Vector database

AI/ML Models (All Open Source, Run Locally)

• Whisper - Audio transcription
• LLaVA 1.6/BLIP-2 - Vision-language understanding
• YOLO v8 - Object detection
• EasyOCR - Optical character recognition
• sentence-transformers - Text embeddings
• Llama 3.1 - Large language model (via Ollama)

Frontend

• React + TypeScript - UI framework
• Vite - Build tool
• TailwindCSS - Styling
• React Query - Data fetching

Prerequisites

Hardware Requirements

• Minimum: 16GB RAM, 8GB VRAM GPU (RTX 3070, GTX 1080 Ti or better)
• Recommended: 32GB RAM, 16GB+ VRAM GPU (RTX 4080/4090, A4000)
• CPU: Multi-core processor (8+ cores recommended)
• Storage: 100GB+ free space (for models and video storage)

Software Requirements

• Python 3.10+
• Node.js 18+
• Docker & Docker Compose
• NVIDIA CUDA 12.1+ (for GPU acceleration)
• FFmpeg
• Ollama (https://ollama.ai)

Quick Start

1. Clone and Setup

git clone <repository-url>
cd Regtech
cp .env.example .env

Edit .env with your configuration.

2. Install Ollama and Pull Models

# Install Ollama from https://ollama.ai
# Then pull the required model
ollama pull llama3.1:8b

3. Start Infrastructure with Docker

cd docker
docker-compose up -d

This starts:

• PostgreSQL (port 5433)
• Redis (port 6379)
• Weaviate (port 8080)
• MinIO (port 9000, console: 9001)

4. Setup Backend

cd backend

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run migrations
alembic upgrade head

# Start FastAPI server
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

5. Start Celery Worker (New Terminal)

cd backend
source venv/bin/activate  # On Windows: venv\Scripts\activate
celery -A app.tasks.celery_app worker --loglevel=info

6. Setup Frontend (New Terminal)

cd frontend
npm install
npm run dev

Access the app at http://localhost:5173

Usage

1. Upload Guideline Document

   • Go to Guidelines page
   • Upload a PDF regulatory document
   • System will parse and structure the requirements

2. Upload Video for Compliance Check

   • Go to Video Upload page
   • Drag and drop or select video file
   • Monitor processing progress in real-time

3. Review Compliance Report

   • View overall compliance status
   • Interactive timeline showing violations at specific timestamps
   • Click on violations to see evidence frames and guideline references
   • Export report as PDF

Development

Run Tests

# Backend tests
cd backend
pytest tests/ -v --cov=app

# Frontend tests
cd frontend
npm test

Project Structure

Regtech/
├── backend/          # FastAPI backend
│   ├── app/
│   │   ├── api/      # API routes
│   │   ├── models/   # Database models
│   │   ├── services/ # Business logic
│   │   ├── langchain_components/  # LangChain chains & agents
│   │   └── tasks/    # Celery tasks
│   └── tests/
├── frontend/         # React frontend
│   └── src/
│       ├── components/
│       ├── pages/
│       └── services/
└── docker/           # Docker configuration

Cost Comparison

Using Open Source Models (This Project)

• API Costs: $0 per video
• Operational Costs: Electricity only (~$0.10-0.50 per video hour)

Alternative: Using Commercial APIs

• API Costs: ~$10-23 per video hour
  • OpenAI Whisper API: $0.36/hour
  • GPT-4 Vision: $5-15/hour
  • AWS Rekognition: $0.60/hour
  • Google Cloud Vision: $1.50/hour
  • OpenAI Embeddings: $0.10/hour
  • Claude API: $2-5/hour

Savings: 100% of API costs eliminated!

Performance

• Video Processing: 5-10 minutes per hour of video
• Compliance Check: 2-5 minutes per guideline set
• Report Generation: < 30 seconds
• Concurrent Videos: 10+ (with 4 Celery workers)

License

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

Contributing

This is a personal/portfolio project. Issues and pull requests are welcome, but there's no formal contribution process in place yet. Feel free to open an issue to discuss a change before submitting a PR.

Support

For issues and questions, please open a GitHub issue.