Guru-Setu-AI

Smart assistant suite for teachers — question paper generator, class summarizer, scheduler & voice assistant.

---

Table of contents ✅

• Guru-Setu-AI
  • Table of contents ✅
  • Project overview
  • Architecture & Tech stack 🔧
  • Features ✨
  • Quickstart — Development 🚀
    • Prerequisites
    • Backend setup
    • Frontend setup
  • Configuration & Environment Variables ⚙️
  • APIs / Endpoints 📡
    • Users
    • Question Generator
    • Class Notes & Summary
    • Scheduler
    • Voice Assistant
  • Data models (brief)
  • Google Classroom & OAuth notes 📚

---

Project overview

Guru-Setu-AI is a full-stack project aimed at helping teachers build/organize teaching material quickly. It combines:

• PDF-based question paper generation (RAG + GenAI)
• Video/YouTube class summarization + optional Google Classroom upload
• Scheduler with email reminders
• Voice assistant (Whisper STT → GenAI → TTS)

The backend is a Flask app and the frontend is a Next.js (App Router) SPA.

---

Architecture & Tech stack 🔧

• Backend: Python, Flask, Flask-SQLAlchemy, Flask-Mail, APScheduler
• Frontend: Next.js 16 (React 19), Tailwind CSS, TypeScript
• LLMs & AI: Google GenAI (Gemini models), sentence-transformers embeddings, Chroma DB (local), Whisper for STT
• Video & audio: yt_dlp, ffmpeg, gTTS
• PDF generation: reportlab
• Storage: local uploads/ and summary/ folders (dev). SQLite used for small persistent storage.

---

Features ✨

• Upload PDFs and generate custom question papers using RAG + LLM.
• Extract audio from video or download YouTube, transcribe, summarize, and export PDF summary.
• Upload generated summary directly to Google Classroom (optional, requires credentials).
• Scheduler UI to create reminders; backend sends email reminders via SMTP.
• Voice assistant endpoint: upload speech, backend transcribes, runs LLM, returns TTS audio.

---

Quickstart — Development 🚀

Prerequisites

• Python 3.10+ (recommended)
• Node 18+ (for Next.js)
• ffmpeg binary available in PATH
• (Optional) Google Cloud credentials for Classroom API

Backend setup

1. Create a virtual environment and install dependencies (create requirements.txt if you prefer):

python -m venv .venv
.venv\Scripts\activate   # Windows
pip install -r backend/requirements.txt  # if you add one
# or install individually
pip install flask flask_sqlalchemy flask_mail flask_cors python-dotenv sentence-transformers pypdf chromadb google-generativeai yt_dlp ffmpeg-python openai-whisper reportlab google-auth google-auth-oauthlib google-api-python-client apscheduler gTTS

2. Add environment variables in backend/.env (see config section below).
3. Ensure backend/credentials.json is present if you want Google Classroom integration.
4. Start the backend:

python backend/app.py

The app will create users.db and schedule.db on first run.

Frontend setup

1. From frontend/ folder:

cd frontend
npm install   # or pnpm install
npm run dev

2. Open http://localhost:3000 (default Next dev port). The frontend is currently configured to call backend at http://localhost:5000.

---

Configuration & Environment Variables ⚙️

Place secrets in backend/.env (do NOT commit to git).

• GEMINI_API_KEY — Google Generative AI API key
• MAIL_USERNAME — SMTP username (Gmail used in dev config)
• MAIL_PASSWORD — SMTP password (or app password)

Files (not committed):

• backend/credentials.json — Google OAuth client secrets
• backend/token.json — OAuth token saved after consent (auto-generated)

Note: backend/.gitignore already excludes these files.

---

APIs / Endpoints 📡

Base: http://localhost:5000

Users

• GET /users — list users
• GET /users/<id> — get user
• POST /user — create user (JSON: { name, password, email })
• PUT /users/<id> — update name
• DELETE /users/<id> — delete

Question Generator

• POST /submit_question (multipart/form-data)
  • Files: pdfs (multiple)
  • Fields: numQuestions, difficulty_1, ..., question_1, etc.
  • Returns: { generated_questions: [{ input: {...}, generated_question }] }

Example cURL (simplified):

curl -X POST "http://localhost:5000/submit_question" \
  -F "pdfs=@mynotes.pdf" \
  -F "numQuestions=1" \
  -F "difficulty_1=medium" \
  -F "marks_1=2" \
  -F "type_1=short" \
  -F "question_1=Explain concept X"

Class Notes & Summary

• POST /classnotes/sendfile — form: either youtube_url or video_file (multipart). Optional upload_to_classroom=true
• GET /classnotes/download/<job_id> — returns summary/class_summary.pdf

Scheduler

• POST /scheduler/add — JSON: { title, time (ISO), email, remind_before }
• GET /scheduler/<email> — list scheduled tasks for email

Voice Assistant

• POST /voice_upload — form: audio file -> returns { teacher_text, assistant_text, tts_audio }
• GET /voice/<file> — serves TTS audio

---

Data models (brief)

• User (bind users_db) — id, name, password, email
• Schedule (bind schedule_db) — id, title, time (DateTime UTC), remind_before (minutes), notified (bool), email

Google Classroom & OAuth notes 📚

• To use upload to Classroom: place credentials.json (OAuth client secret) in backend/ and run the flow. The server writes token.json after user consent. The OAuth scope is limited to classroom announcements & drive file upload.
• course_id.txt stores chosen course id after listing courses.

---