AI Study Assistant

A multi-user AI-powered study assistant built with Flask. Upload PDFs, paste YouTube links, or type questions — and get context-aware responses powered by Groq. Each user gets isolated sessions with persistent chat history backed by PostgreSQL and a two-layer Redis caching architecture.

---

Features

• Multi-user authentication — JWT-based auth with secure cookie handling and CSRF protection
• Session management — create, switch between, and persist multiple named chat sessions per user
• Context-aware responses — last 10 messages of each session are held in Redis for fast, stateful conversation
• PDF support — upload PDFs and ask questions about their content
• YouTube support — paste a YouTube URL and get a transcript-based AI response
• Two-layer Redis caching
  • Stateless cache: SHA-256 keyed prompt → response cache for repeated factual questions
  • Stateful detection: follow-up prompts ("explain that", "give me an example") are never cached
  • Session cache: per-session message history stored as Redis lists, rebuilt from PostgreSQL on cache miss
• Token revocation — logout invalidates the JWT via a Redis blocklist (TTL-matched to token expiry)
• Blueprint architecture — modular route structure with an app factory pattern

---

Tech Stack

Layer	Technology
Backend	Python, Flask
Database	PostgreSQL + Flask-SQLAlchemy + Flask-Migrate
Cache	Redis
Auth	Flask-JWT-Extended
LLM	Groq API
PDF Parsing	PyMuPDF (fitz)
YouTube	youtube-transcript-api
Server	Gunicorn
Containerization	Docker + Docker Compose

---

Project Structure

app/
├── __init__.py              # App factory
├── extensions.py            # db, redis, jwt init
├── models/
│   ├── user.py
│   ├── sessions.py
│   ├── messages.py
│   └── token_blocklist.py
├── routes/
│   ├── auth.py              # Register, login, logout
│   ├── prompt.py            # Core LLM interaction
│   ├── session_extractor.py # Session listing/switching
│   └── uploads.py           # PDF upload handling
└── services/
    ├── llm_services/        # Groq provider
    ├── llm_cache_services/  # Stateless text + PDF cache
    ├── session_cache_services/ # Per-session Redis history
    ├── session_services/    # Session creation
    ├── message_services/    # DB read/write for messages
    ├── pdf_services/        # PDF text extraction
    └── yt_services/         # YouTube transcript fetching

---

Getting Started

Prerequisites

• Docker and Docker Compose
• A Groq API key

Setup

1. Clone the repo

git clone https://github.com/jayssSmm/StudyMindAI.git
cd StudyMindAI

2. Create a .env file

SECRET_KEY=your-secret-key
JWT_SECRET_KEY=your-jwt-secret
GROQ_API_KEY=your-groq-api-key

DATABASE_URL=postgresql://postgres:password@db:5432/studydb
REDIS_URL=redis://redis:6379/0

3. Start the containers

docker compose up --build

4. Run database migrations

docker compose exec web flask db upgrade

The app will be available at http://localhost:5000.

---

How It Works

Request Flow

User sends prompt
       │
       ▼
Is it stateful? (e.g. "explain that") ──► Yes ──► Skip cache, go to LLM
       │ No
       ▼
Cache hit in Redis? ──► Yes ──► Return cached response
       │ No
       ▼
Is it a YouTube URL? ──► Yes ──► Fetch transcript ──► LLM
       │ No
       ▼
Fetch last 10 messages from Redis (or rebuild from PostgreSQL)
       │
       ▼
Send prompt + history to Groq
       │
       ▼
Store response in Redis cache + session history + PostgreSQL

Stateful Detection

Prompts like "give me an example", "explain that", "continue", or "as you mentioned above" are detected as context-dependent via regex pattern matching and are never cached — they always go through the full LLM pipeline with session history attached.

---

API Endpoints

Method	Endpoint	Description
POST	/auth/register	Create a new user account
POST	/auth/login	Login, returns JWT cookie
POST	/auth/logout	Invalidates token via Redis blocklist
POST	/prompt	Send a prompt, get an AI response
GET	/sessions	List all sessions for current user
POST	/uploads/pdf	Upload a PDF for context

All protected routes require a valid JWT cookie.

---

Environment Variables

Variable	Description
SECRET_KEY	Flask secret key
JWT_SECRET_KEY	JWT signing key
GROQ_API_KEY	Groq API key
DATABASE_URL	PostgreSQL connection string
REDIS_URL	Redis connection string

---