Medo-Veda is an AI-powered consumer health analysis platform focused on packaged food intelligence. It reads product labels, analyzes ingredient risk, verifies marketing claims, personalizes advice by user health profile, and generates a clinical-style report.

This README is updated to match the current codebase architecture and behavior.

Medo-Veda supports:

• Product scan and analysis flow (image-first)
• Ingredient-level risk classification
• Claim vs reality verification
• Profile-aware advice (age, goals, conditions)
• Nutrition snapshot normalization and fallback
• Report history and dashboard summary
• Downloadable report PDF from report view

System overview:

1. Frontend (React + Vite) captures user input and polls status.
2. Backend (Node + Express) orchestrates multi-agent analysis.
3. Vision + AI agents generate structured report data.
4. PostgreSQL persists users, profiles, scans, and reports.
5. Cloudinary stores uploaded images.
6. Redis caches computed responses where applicable.

Flow summary:

1. User uploads image or submits scan request.
2. Backend starts pipeline and tracks progress by scanId.
3. Frontend polls scan status endpoint.
4. Once complete, frontend fetches final report payload.
5. Report renders in Analysis Report UI and can be exported to PDF.

• React 19
• Vite 8
• Tailwind CSS 4
• React Router
• Axios
• Framer Motion
• React Hot Toast
• html-to-image + jsPDF + html2canvas (report export)
• MUI dependencies (available in project)

• Node.js + Express 5
• PostgreSQL (pg)
• Cloudinary
• Upstash Redis
• Multer
• Sharp
• JWT + bcryptjs
• Axios
• jsonrepair
• OpenAI SDK and Google/Vertex packages present in dependencies

• Vision-first orchestration
• Multi-agent pipeline with safe fallbacks and timeout wrappers
• Report assembly and output validation
• Guideline enrichment via Wikipedia-backed lookup service

Core backend entry:

• Express app bootstraps middleware, routes, health checks, and timeout controls.

Important route groups:

1. Auth
   • Register and login
2. Profile
   • Setup and get profile
3. Scan lifecycle
   • Create scan
   • Poll status
   • Fetch result
   • History/list and single record
4. Dashboard summary
5. Vision/image extraction endpoints

Pipeline behavior (orchestrator):

1. Wave 1: Vision or product discovery + persona context
2. Wave 2: Ingredient + claims + web research fallback
3. Wave 3: Personalization + evidence + alternatives + verdict
4. Wave 4: Assembly of final report schema

Stability design:

• Each critical external/agent call is wrapped by a safe timeout helper.
• Fallback paths are used when an agent fails or returns sparse output.
• Guideline text is sanitized so UI receives renderable values.

Core frontend app shell:

• Route-based pages with fixed navbar and mobile bottom navigation patterns.
• Protected routes for authenticated pages.
• Mobile-safe spacing and viewport handling are implemented globally.

Key pages:

1. Landing page
2. Auth page (separate mobile/desktop layouts)
3. Dashboard
4. Scan page
5. Report loading page (status polling)
6. Analysis report page
7. History page
8. Profile setup page

Report rendering notes:

• Ingredient and nutrition values are normalized defensively.
• Personal advice is layout-optimized for mobile readability.
• PDF export path captures rendered DOM and writes a downloadable PDF.

1. Node.js 18+
2. npm
3. PostgreSQL database connection string
4. Cloudinary account keys
5. NVIDIA API keys for pipeline calls

git clone https://github.com/Dineshkumar2006471/Medo-veda.git
cd Medo-veda

cd backend
npm install

Create backend .env with (current expected keys):

1. DATABASE_URL
2. JWT_SECRET
3. NVIDIA_API_KEY
4. CLOUDINARY_CLOUD_NAME
5. CLOUDINARY_API_KEY
6. CLOUDINARY_API_SECRET
7. UPSTASH_REDIS_REST_URL
8. UPSTASH_REDIS_REST_TOKEN
9. PORT (optional, defaults from code)
10. WIKI_CONTACT_EMAIL (optional, fallback exists)

Start backend:

npm run dev

Health check:

curl http://localhost:3001/health

cd ../frontend
npm install

Create frontend .env:

1. VITE_API_URL=http://127.0.0.1:3001

Start frontend:

npm run dev

Open:

• http://localhost:5173

1. User logs in and profile is available.
2. User uploads/starts scan.
3. Backend creates scan and starts orchestrator.
4. Frontend polls status endpoint.
5. Backend assembles final report.
6. Frontend loads Analysis Report with normalized payload.
7. User downloads PDF copy from report page.

• Deploy backend folder as web service.
• Set all required env vars listed above.
• Ensure outbound access to AI and Cloudinary endpoints.

• Deploy frontend as Vite app.
• Set VITE_API_URL to deployed backend URL.

• Verify DATABASE_URL and SSL behavior of your host.
• Check backend logs for pool connection timeout.

• Check scan status endpoint response.
• Confirm AI keys are set and valid.
• Check image upload path and Cloudinary env values.

• Confirm frontend build is up to date.
• Validate report page renders fully before export.
• Check browser console for DOM capture or CORS errors from external images.

Recently stabilized areas:

1. Mobile auth and navigation behavior
2. Report rendering robustness
3. Ingredient guideline fallback/safety handling
4. Pipeline safety wrappers for external enrichment
5. PDF export flow hardening

This repository is currently used for hackathon/academic and product prototyping workflows.