Turn your notes into interactive quizzes and track your learning progress.
-
User Authentication
Secure authentication with JWT tokens and email login, ensuring safe access to your account. -
Upload Notes & Auto-Generate Questions
Upload your study notes in PDF Format and automatically generate questions in multiple formats:- MCQ (Multiple Choice Questions)
- MRQ (Multiple Response Questions)
- True/False
-
Create Chapters and Questions
Organize your content efficiently by creating chapters and questions for quizzes and study material. -
Take Quizzes & Track Attempts
Test your knowledge with interactive quizzes. Review answers, monitor progress, and analyze performance over multiple attempts. -
Dashboard with Analytics and Charts
Gain insights into your learning progress through detailed analytics, including charts and performance metrics.
- Support Category creation
- An Analytics page for detailed insights and reporting.
- Asynchronous chapter and question generation powered by Celery for improved performance.
- Notes summary generator
Login
HomePage / Dashboard
Upload PDFs
Chapters List
Chapter Attempts
In Attempt (Question View)
In Quiz (Answering)
- Backend: Create .env under backend/ and set
GEMINI_API_KEYto your Gemini API Key - Frontend: Create .env under frontend/ and set
NEXT_PUBLIC_API_URLto your API URL (backend)
All common setup and run commands are available via the Makefile at the project root. Run these from the project root directory:
make helpmake backend-install
make frontend-installmake backend-migratemake backend-run
make frontend-run-devmake frontend-build
make frontend-run-prod- Architecture: Single Page Application
- Backend: Django REST Framework, JWT Auth, OpenAI integration for question generation.
- Frontend: Next.js (App Router), React, Tailwind CSS, shadcn/ui, Recharts for charts.
- Database: SQLite (default, can be swapped for Postgres/MySQL)
- Database Modelling:
The application uses a Gemini fallback chain for AI generation:
- Try gemini-2.5-pro first for best reasoning and summary quality.
- If it returns transient overload / 503 after limited retries, fall back to gemini-2.5-flash.
- If Flash also overloads, fall back to gemini-2.5-flash-lite (fastest / cheapest) to still return something.
- If all three fail, a clear error is returned.
Only text input/output is currently used (no image/video/audio responses stored), but full multimodal support is available in all three Gemini 2.5 models.
(Shows standard vs batch where relevant; Flash Lite & Flash have same rate ≤ / > 200K for text.)
| Model | Input ≤200K | Input >200K | Output Text (≤200K→>200K) | Batch Input ≤200K | Batch Input >200K | Batch Output (≤200K→>200K) | Max Input Tokens | Max Output Tokens | Notes |
|---|---|---|---|---|---|---|---|---|---|
| gemini-2.5-pro | $1.25 | $2.50 | $10 → $15 | $0.625 | $1.25 | $5.00 → $7.50 | 1,048,576 | 65,535 | Highest quality / deep reasoning |
| gemini-2.5-flash | $0.30 | $0.30 | $2.50 (flat) | $0.15 | $0.15 | $1.25 (flat) | 1,048,576 | 65,535 | Balanced cost/speed |
| gemini-2.5-flash-lite | $0.10 | $0.10 | $0.40 (flat) | $0.05 | $0.05 | $0.20 (flat) | 1,048,576 | 65,536 | Lowest cost / fallback & bulk |
| gpt-4o-mini (ref) | $0.50 | $0.50 | $1.50 | – | – | – | 128,000 | ~16,000+ | Alternative; polish tradeoff |
- Pro delivers best structured summaries & nuanced distractors for MCQ/MRQ.
- Flash handles the majority of volume cheaply when Pro is temporarily busy.
- Flash Lite ensures resilience under sustained platform load (keeps user flow unblocked).
- Gemini chosen for this personal project because of a generous free usage tier → near-zero iteration cost while prototyping; GPT kept as an easy swap path.
- The service adds an internal
_model_usedfield for downstream logging/analytics. - Fallback & retry logic lives in
backend/api/utils/ai.py(call_gemini_model). Adjust order viaMODEL_CHAIN.
MIT