Deployment - PDF AI

Overview

PDF AI is designed to be deployed on Vercel with Edge Functions for optimal performance. The application leverages Vercel’s serverless infrastructure to provide fast, scalable PDF processing and AI-powered chat capabilities.

There is currently a known issue with Vercel Edge Functions deployment. The application works perfectly in local development but may experience issues in the deployed version. See the Known Issues section below.

Deploying to Vercel

Connect your repository

Push your code to GitHub, GitLab, or Bitbucket
Log in to Vercel
Click “New Project”
Import your repository

Vercel will automatically detect that this is a Next.js project and configure the build settings.

Configure environment variables

In the Vercel project settings, add all required environment variables:Database:

DATABASE_URL

OpenAI:

OPEN_AI_KEY

Pinecone:

PINECONE_API_KEY
PINECONE_ENVIRONMENT

AWS S3:

NEXT_PUBLIC_S3_ACCESS_KEY_ID
NEXT_PUBLIC_S3_SECRET_ACCESS_KEY
NEXT_PUBLIC_S3_BUCKET_NAME

Clerk:

NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY
CLERK_SECRET_KEY

Stripe:

STRIPE_API_KEY (use live key sk_live_...)
STRIPE_WEBHOOK_SECRET

Application:

NEXT_BASE_URL (e.g., https://your-domain.vercel.app)

Make sure to use production API keys and credentials, not test/development keys.

Configure build settings

Vercel should automatically detect these settings, but verify:

Framework Preset: Next.js
Build Command: next build
Output Directory: .next
Install Command: npm install
Development Command: next dev

Deploy

Click “Deploy” to start the build process. Vercel will:

Install dependencies
Run the Next.js build
Deploy to the edge network
Provide a production URL

The deployment typically takes 2-5 minutes.

Configure custom domain (optional)

To use a custom domain:

Go to your project settings
Navigate to “Domains”
Add your custom domain
Configure DNS records as instructed

Edge Functions

PDF AI uses Vercel Edge Functions for API routes to provide:

Low latency: Edge Functions run closer to your users
Global distribution: Automatically deployed to Vercel’s global edge network
Scalability: Automatically scales based on demand

Edge Runtime Compatibility

The following API routes are designed to work with Edge Runtime:

/api/chat - AI chat endpoint (src/app/api/chat/route.ts:13)
/api/stripe - Stripe checkout (src/app/api/stripe/route.ts:9)
/api/webhook - Stripe webhooks (src/app/api/webhook/route.ts:20)

Edge Functions have some limitations compared to Node.js serverless functions. Ensure all dependencies are compatible with the Edge Runtime.

Production Considerations

Memory Configuration

In production, you may need to adjust memory limits for processing large PDFs:

package.json

{
  "scripts": {
    "build": "cross-env NODE_OPTIONS=--max-old-space-size=8192 next build"
  }
}

Database Connection Pooling

PDF AI uses Neon’s serverless driver which automatically handles connection pooling. No additional configuration is needed, but ensure your Neon plan supports the expected number of concurrent connections.

S3 CORS Configuration

Ensure your S3 bucket has proper CORS configuration for your production domain:

[
  {
    "AllowedHeaders": ["*"],
    "AllowedMethods": ["GET", "PUT", "POST", "DELETE"],
    "AllowedOrigins": [
      "https://your-domain.vercel.app",
      "https://your-custom-domain.com"
    ],
    "ExposeHeaders": ["ETag"]
  }
]

Stripe Webhooks

Configure your Stripe webhook endpoint:

Go to Stripe Dashboard → Developers → Webhooks
Add endpoint: https://your-domain.vercel.app/api/webhook
Select events to listen for (e.g., checkout.session.completed)
Copy the webhook signing secret to STRIPE_WEBHOOK_SECRET

Clerk Domain Configuration

Update your Clerk application with the production domain:

Go to Clerk Dashboard → Your Application
Navigate to “Paths”
Update the Home URL and callback URLs to your production domain

Monitoring and Debugging

Vercel Logs

Access real-time logs from your Vercel dashboard:

Go to your project
Click on “Deployments”
Select a deployment
View logs in the “Functions” tab

Error Tracking

Consider integrating error tracking services:

Sentry: Comprehensive error tracking
LogRocket: Session replay and error tracking
Vercel Analytics: Built-in analytics and Web Vitals

Performance Monitoring

Monitor your application performance:

Vercel Analytics: Enable in project settings
Web Vitals: Track Core Web Vitals metrics
Custom Logging: Add logging to critical paths

Deployment Checklist

Before deploying to production:

All environment variables configured in Vercel
Production API keys (not test keys) are being used
Database migrations have been run
S3 bucket CORS is configured for production domain
Stripe webhook endpoint is configured
Clerk domain settings are updated
Custom domain DNS is configured (if applicable)
Error tracking is set up
Analytics are enabled

Known Issues

Edge Function Deployment Issue: There is currently a known issue where the application works perfectly in local development but encounters problems when deployed to Vercel Edge Functions.

Current Status

Local development: Fully functional
Vercel deployment: Experiencing issues with Edge Functions

Potential Causes

Edge Runtime Compatibility: Some dependencies may not be compatible with the Edge Runtime
Environment Variables: Edge Functions may have different variable scoping
Memory Limits: Edge Functions have stricter memory constraints
Cold Starts: First request after deployment may timeout

Workarounds

While the team works on resolving the Edge Function issue, consider:

Use Node.js Runtime: Modify API routes to use Node.js runtime instead of Edge:
```
export const runtime = 'nodejs'; // instead of 'edge'
```

Increase Function Duration: In vercel.json:

{
  "functions": {
    "api/**/*.ts": {
      "maxDuration": 60
    }
  }
}

Test Thoroughly: Deploy to a preview environment first:
```
vercel --prod=false
```

Contributing to the Fix

If you’d like to help resolve this issue:

Clone the repository
Test locally to confirm it works
Deploy to Vercel and document the error
Submit a pull request with potential fixes
Share your findings in the project issues

Continuous Deployment

Vercel automatically deploys:

Production: When you push to the main/master branch
Preview: When you create a pull request
Development: When you push to development branches

Each deployment gets a unique URL for testing before promoting to production.

Rollback

If a deployment causes issues:

Go to your project in Vercel
Navigate to “Deployments”
Find the last working deployment
Click the three dots menu
Select “Promote to Production”

This instantly rolls back to the previous version.

Getting Help

Vercel Documentation: vercel.com/docs
Next.js Deployment: nextjs.org/docs/deployment
Project Repository: github.com/Srijan-D/pdf.ai
Vercel Support: [email protected]

Help

Development

​Overview

​Deploying to Vercel

​Edge Functions

​Edge Runtime Compatibility

​Production Considerations

​Memory Configuration

​Database Connection Pooling

​S3 CORS Configuration

​Stripe Webhooks

​Clerk Domain Configuration

​Monitoring and Debugging

​Vercel Logs

​Error Tracking

​Performance Monitoring

​Deployment Checklist

​Known Issues

​Current Status

​Potential Causes

​Workarounds

​Contributing to the Fix

​Continuous Deployment

​Rollback

​Getting Help

Build docs developers (and LLMs) love

Overview

Deploying to Vercel

Edge Functions

Edge Runtime Compatibility

Production Considerations

Memory Configuration

Database Connection Pooling

S3 CORS Configuration

Stripe Webhooks

Clerk Domain Configuration

Monitoring and Debugging

Vercel Logs

Error Tracking

Performance Monitoring

Deployment Checklist

Known Issues

Current Status

Potential Causes

Workarounds

Contributing to the Fix

Continuous Deployment

Rollback

Getting Help