Project: AI SaaS (Full Stack)

Level: Advanced Time: 1-2 Weeks Stack: Next.js, Postgres, Stripe, Clerk, Vercel AI SDK

Overview

Build a production-ready "Chat with PDF" SaaS platform where users pay a monthly subscription.

Architecture

Step 1: Authentication (Clerk)

Protect your routes.

npm install @clerk/nextjs

// middleware.ts
import { authMiddleware } from "@clerk/nextjs";
export default authMiddleware({
  publicRoutes: ["/", "/api/webhook/stripe"],
});

Step 2: Database Schema

You need to track:

1. Users (Subscription status)
2. Files (Uploaded PDFs)
3. Chats (History)
4. Messages

model User {
  id String @id
  stripeCustomerId String?
  isPro Boolean @default(false)
}

Step 3: Payments (Stripe)

Create a Checkout Session.

// app/api/stripe/route.ts
const session = await stripe.checkout.sessions.create({
  line_items: [{ price: 'price_123', quantity: 1 }],
  mode: 'subscription',
  success_url: `${origin}/dashboard`,
  metadata: { userId: user.id },
});

Step 4: The RAG Pipeline

When a user uploads a file:

1. Index: Read PDF -> Chunk -> Embed -> Store in Pinecone/Supabase.
2. Namespace: IMPORTANT! Tag vectors with userId or fileId so users can't search each other's files.

// Metadata filter is CRITICAL for security
const results = await vectorStore.similaritySearch(query, 5, {
  userId: currentUser.id, // Only search my files
});

Step 5: Rate Limiting

Prevent abuse.

• Free Users: 5 messages/day.
• Pro Users: 100 messages/day. Use Upstash Redis for this.

Deployment Checklist

1. Environment Variables set in Vercel.
2. Stripe Webhooks configured.
3. Database migrations applied.
4. Domain DNS configured.