📚 Study Agent — AI-Powered Study Assistant

An AI study assistant that lets you upload PDF or text documents, ask questions about them, and auto-generate quizzes to test your understanding — all powered by RAG (Retrieval-Augmented Generation).

Built with FastAPI · Vanilla JavaScript · Groq (LLaMA 3.3 70B) · Sentence Transformers

Live Demo: thejesh06.github.io/study-agent · API: study-agent-backend-5o1i.onrender.com

---

Features

• Document Q&A — Upload any PDF or .txt file and ask questions in natural language. Answers are grounded in your document and supplemented by the LLM's own knowledge.
• Conversation Memory — Follow-up questions work naturally; the assistant remembers the last 3 exchanges.
• Multi-document Support — Upload multiple documents and switch between them using the document selector. Queries stay scoped to the selected document.
• Quiz Generator — Auto-generate 3–10 multiple choice questions from any uploaded document. Instant scoring with correct/wrong highlighting and explanations.
• Persistent Storage — Uploaded documents are indexed and saved to disk — they survive server restarts.

---

How It Works

PDF / TXT
   │
   ▼
[Chunker] — splits text into 1000-char overlapping chunks
   │
   ▼
[Embedder] — converts each chunk to a vector (all-MiniLM-L6-v2, runs locally)
   │
   ▼
[Vector Store] — saves chunks + embeddings to disk (JSON)
   │
   ▼  (on query)
[Retriever] — embeds the question, finds top-5 most similar chunks via cosine similarity
   │
   ▼
[LLM] — Groq LLaMA 3.3 70B generates the answer using retrieved chunks as context

---

Tech Stack

Layer	Technology
Backend	FastAPI + Uvicorn
Frontend	Vanilla JavaScript + CSS (GitHub Pages)
LLM	Groq API (LLaMA 3.3 70B Versatile)
Embeddings	sentence-transformers/all-MiniLM-L6-v2 (local, no API needed)
PDF Parsing	pypdf
Vector Storage	JSON file (persistent, no external DB)

---

Project Structure

study-agent/
├── backend/
│   ├── app.py          # FastAPI app, CORS, router
│   ├── llm.py          # Groq client, multi-turn chat
│   ├── embeddings.py   # Local sentence-transformer model
│   ├── vector_db.py    # Persistent vector store (JSON)
│   ├── chunker.py      # Text splitting with overlap
│   └── routes/
│       └── api.py      # All API endpoints
├── docs/
│   └── index.html      # Frontend — HTML/CSS/JS (served via GitHub Pages)
├── requirements.txt
└── .env                # GROQ_API_KEY goes here

---

Setup

1. Clone the repo

git clone https://github.com/Thejesh06/study-agent.git
cd study-agent

2. Create a virtual environment

python -m venv venv

# Windows
venv\Scripts\activate

# Mac/Linux
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Add your Groq API key

Create a .env file in the study-agent/ directory:

GROQ_API_KEY=your_groq_api_key_here

Get a free API key at console.groq.com.

---

Running Locally

Terminal — Backend:

venv\Scripts\activate
uvicorn backend.app:app --reload

Then open docs/index.html directly in your browser — it talks to localhost:8000 by default.

---

Deployment

Service	Purpose
Render	Hosts the FastAPI backend (free tier)
GitHub Pages	Hosts the HTML/JS frontend (free, from /docs folder)

---

API Endpoints

Method	Endpoint	Description
GET	/api/ping	Health check
GET	/api/documents	List all indexed documents
POST	/api/upload	Upload and index a PDF or TXT file
POST	/api/query	Ask a question (RAG pipeline)
POST	/api/quiz	Generate MCQ quiz from a document

Example — Query

curl -X POST https://study-agent-backend-5o1i.onrender.com/api/query \
  -H "Content-Type: application/json" \
  -d '{"question": "What is photosynthesis?", "doc_filter": "biology_notes.pdf"}'

Example — Quiz

curl -X POST https://study-agent-backend-5o1i.onrender.com/api/quiz \
  -H "Content-Type: application/json" \
  -d '{"doc_filter": "biology_notes.pdf", "num_questions": 5}'

---

Usage

1. Upload a document — Use the sidebar to upload a PDF or TXT file and click "Upload & Start Chatting"
2. Select the document — Pick it from the "Active Document" dropdown to restrict search to that file
3. Chat — Type questions in the Chat tab; the assistant answers from your document
4. Quiz — Switch to the Quiz tab, set the number of questions, and click "Generate Quiz"
5. Score — Submit your answers to see correct/wrong answers with explanations

---

Environment Variables

Variable	Description
GROQ_API_KEY	Your Groq API key (required)