VedaAI Assessment Creator

VedaAI is a full-stack AI assessment creator for teachers. It follows the provided Figma/UI references in ui/ and implements the flow from assignment creation to background generation to a structured printable question paper.

Figma reference:

• https://www.figma.com/design/nB2HMm1BhTpmHcHrmEslGB/VedaAI---Hiring-Assignment?node-id=0-1&t=UjYQLgEek4u99AA4-1

Architecture

• apps/web: Next.js + TypeScript frontend with Zustand state management and Socket.IO client updates.
• apps/api: Express + TypeScript backend with REST APIs, Socket.IO, MongoDB persistence, Redis/BullMQ jobs, OpenAI structured generation, and local fallbacks.
• packages/shared: Shared TypeScript contracts for assignments, generation progress, and generated papers.

The backend is designed around clean boundaries:

• Routes validate input and call application services.
• Repository stores assignments and generated papers in MongoDB when MONGODB_URI is present, otherwise memory for quick demos.
• Queue uses BullMQ when REDIS_URL is present, otherwise an in-process async worker.
• Generator uses OpenAI structured JSON when OPENAI_API_KEY is present, validates the response with Zod, and maps it into app-owned paper objects. It never renders raw model text.

Features

• Assignment list and empty state based on the supplied desktop/mobile references.
• Create assignment form with file upload selection UI, due date, question type rows, count/marks steppers, source material, and validation.
• Real-time generation progress over WebSocket.
• Structured question paper output with student info lines, sections, instructions, difficulty tags, and marks.
• Backend PDF download endpoint powered by a BullMQ PDF job.
• Responsive desktop and mobile layouts.

Setup

Install dependencies:

npm install

Start MongoDB and Redis for the complete backend flow:

docker compose up -d

Create apps/api/.env:

PORT=4000
CLIENT_URL=http://localhost:3000
MONGODB_URI=mongodb://localhost:27017/veda-ai
REDIS_URL=redis://localhost:6379
OPENAI_API_KEY=your_key_here
OPENAI_MODEL=gpt-4o-mini

Create apps/web/.env.local:

NEXT_PUBLIC_API_URL=http://localhost:4000

Run the API:

npm run dev:api

Run the web app:

npm run dev:web

Open:

http://localhost:3000

Runtime Modes

Full assignment mode:

• MongoDB stores assignments and generated papers.
• Redis powers BullMQ queue state and short-lived assignment/paper response caching.
• BullMQ processes generation and PDF jobs in the background.
• Socket.IO sends real-time progress to the frontend.
• OpenAI returns structured JSON, which is validated and rendered as sections/questions.

Demo mode:

• If MONGODB_URI, REDIS_URL, or OPENAI_API_KEY are missing, the app still runs with memory storage, an in-process worker, and deterministic structured generation.

Useful Commands

npm run typecheck
npm run build
npm run dev:api
npm run dev:web