How ossamalouati/legal-ai_project works
Project type and purpose
• Full-stack web application for automated legal document analysis.
• Frontend: Next.js/TypeScript web app (UI, authentication, API proxy).
• Backend: Flask server (machine-learning inference, NLP pipelines, data files).
Main components
a. User & Browser
b. Frontend Service (“web_app” container)
– Next.js server (Node.js + React + TypeScript)
– NextAuth authentication module
– Pages & API routes (pages/api/*)
– Static assets (public/assets)
c. Backend Service (“flask_server” container)
– Flask web server (app.py)
– ML modules:
• predict.py (reading comprehension, SQuAD-based QA)
• paraphrase.py (T5-based paraphrasing)
• sentiment analysis (TextBlob)
– Data store (flat files: contracts, questions, responses.json, nbest.json)
d. Container orchestration
– Docker Compose (docker-compose.yml)
e. External resources (optionally highlight datasets/ML models on Hugging Face, SQuAD, CUAD)
Relationships & data flows
• Users ↔ Browser → HTTP/HTTPS → Next.js Frontend
• Frontend (API routes) → HTTP → Flask server endpoints (e.g., /predict, /paraphrase)
• Flask server → local data files & ML modules → compute inferences → returns JSON → Frontend
• Frontend → renders results + charts/summary → Browser
• Docker Compose → orchestrates two containers + shared network
Architectural patterns & design principles
• Client-Server separation (UI vs. ML backend)
• Microservice-style split (two independently containerized services)
• Layered architecture: Presentation (Next.js), Application (Flask APIs), Data/ML layer
• Use of containerization for environment parity