TestMeGemini

Turn your notes into interactive quizzes and track your learning progress.

Features

• 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.

Upcoming Features

• 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

Screenshots

The following captures show the current local demo flow, from sign-in through quiz completion and answer review.

Login

HomePage / Dashboard

Upload PDFs

Chapters List

Chapter Attempts

In Attempt (Question View)

In Quiz (Answering)

Installation & Local Development

Quick local start

After installing dependencies and creating backend/.env, start Django and Next.js together from the repository root:

make local-run

The default local URLs are:

• Frontend: http://127.0.0.1:3000
• Backend: http://127.0.0.1:8001

Port 8001 is used for Django so the app can coexist with tools that commonly occupy port 8000. Override either port when needed:

TESTME_BACKEND_PORT=8100 TESTME_FRONTEND_PORT=3100 make local-run

The dashboard, chapters, quizzes, and attempt history work without calling an AI provider. Uploading a PDF to generate new questions requires a valid GEMINI_API_KEY.

To verify that both servers start, proxy requests correctly, and shut down cleanly:

make local-check

Environment Variables

• Backend: copy backend/.env.example to backend/.env, then set a unique SECRET_KEY and your GEMINI_API_KEY. Production deployments must also set DEBUG=False, ALLOWED_HOSTS, CORS_ALLOWED_ORIGINS, and CSRF_TRUSTED_ORIGINS explicitly.
• Frontend: create frontend/.env and set NEXT_PUBLIC_API_URL to the backend origin.
• SQLite is intended for local development. Configure DATABASE_ENGINE and the related database variables in backend/.env.example for a production SQL database.

Using the Makefile (Recommended)

All common setup and run commands are available via the Makefile at the project root. Run these from the project root directory:

For full list of commands

make help

1. Install dependencies (backend & frontend)

make backend-install
make frontend-install

2. Run committed migrations

make backend-migrate

3. Start the development servers (in separate terminals)

make backend-run
make frontend-run-dev

4. Build for production

make frontend-build
make frontend-run-prod

Tests

make test-backend
make test-frontend
make test-e2e
make test-all

The Playwright command is added by the full QA setup and uses an isolated test database; it does not call Gemini.

For verified click-by-click behavior, every versioned endpoint, Django/DRF implementation details, and the latest test record, see docs/QA.md.

Architecture Overview

• Architecture: Single Page Application
• Backend: Django REST Framework, cookie-based JWT authentication, CSRF protection, and Google Gemini 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:

API

New clients should use the versioned /api/v1/ routes:

• /api/v1/auth/ for CSRF bootstrap, login, logout, and refresh
• /api/v1/users/ for registration and the authenticated user profile
• /api/v1/chapters/ for chapter CRUD and chapter questions
• /api/v1/chapter-generations/ for throttled PDF-to-question generation
• /api/v1/attempts/ for attempt submission and history

Browser clients must first request /api/v1/auth/csrf/ and send the resulting CSRF token on unsafe cookie-authenticated requests. Older /api/ routes remain temporarily available for compatibility.

The default SameSite=Lax cookie mode is intended for same-site deployments. If the frontend and backend are genuinely cross-site, configure SameSite=None and secure HTTPS cookies explicitly.

AI Model

The application currently uses the Google Gen AI SDK with this Gemini fallback chain:

1. gemini-3.8-flash — primary model, selected for its output quality and current stable free-tier availability.
2. gemini-3.6-flash — stable general-purpose fallback when the primary model is temporarily overloaded.
3. gemini-3.5-flash-lite — final, cost-efficient stable fallback for availability.

All three model IDs are stable Google Gemini API models, accept text/PDF input, support structured output, and are currently eligible for free input/output usage within a project's free-tier quota. Exact RPM, TPM, and daily limits vary by project and are shown in Google AI Studio. Free-tier prompts and uploaded content may be used by Google to improve its products, so do not upload private or sensitive documents without reviewing Google's current data-use terms.

The application extracts text from uploaded PDFs and asks Gemini to return the chapter summary and quiz questions as JSON matching the Pydantic response schema. The selected question count is enforced in the prompt, with JSON Schema minItems/maxItems, and through server-side validation before anything is saved.

Fallback occurs only for transient overload, service-unavailable, or deadline-timeout failures (HTTP 503/504). Each model is retried twice before the application advances to the next model. Other provider errors stop immediately and return a safe generation error.

Model configuration is defined by MODEL_CHAIN in backend/api/utils/ai.py. Authentication and timeout configuration use:

GEMINI_API_KEY=your-google-ai-api-key
GEMINI_TIMEOUT_SECONDS=30

Google model references: Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, pricing, and rate limits.

Observability

• The service adds an internal _model_used field for downstream logging/analytics.
• Fallback and retry logic lives in backend/api/utils/ai.py (call_gemini_model). Adjust the order with MODEL_CHAIN.

License

MIT