Star on GitHub
Star on GitHub

Deploy PRSense to Production RIGHT NOW

Fastest Path (5 Minutes)

Step 1: Get API Key

# Go to: https://platform.openai.com/api-keys
# Click "Create new secret key"
# Copy the key (starts with sk-...)

Step 2: Deploy to Vercel

# Install Vercel CLI
npm i -g vercel

# Deploy
cd PRSense
vercel

# Follow prompts, press Enter for defaults

Step 3: Add Secrets

# Add OpenAI key
vercel env add OPENAI_API_KEY
# Paste your key: sk-...

# Add GitHub token (get from https://github.com/settings/tokens)
vercel env add GITHUB_TOKEN
# Paste token: ghp_...

# Add webhook secret (make up a random string)
vercel env add GITHUB_WEBHOOK_SECRET
# Paste: your-random-secret-123

Step 4: Configure GitHub Webhook

1. Go to your repo → Settings → Webhooks → Add webhook

2. Payload URL: https://your-app.vercel.app/webhook
(Vercel shows you the URL after deploy)

3. Content type: application/json

4. Secret: your-random-secret-123
(same as GITHUB_WEBHOOK_SECRET above)

5. Events: "Let me select individual events"
Pull requests

6. Click "Add webhook"

Step 5: Test

1. Open a test PR in your repo
2. Check webhook deliveries (GitHub settings)
3. Should see 200 response
4. If duplicate detected, bot comments automatically!

Done!

Your PRSense instance is now:

  • Detecting duplicates automatically
  • Commenting on PRs
  • Using real OpenAI embeddings
  • Running serverlessly (scales automatically)
  • Costing ~$0-1/month

Alternative: Self-Hosted (Docker)

If you prefer Docker instead of Vercel:

# 1. Configure environment
cp .env.example .env
nano .env # Add your keys

# 2. Start services
docker-compose up -d

# 3. Configure GitHub webhook
# URL: http://your-server-ip:3000/webhook

# Done!

Free Alternative (No OpenAI Cost)

Use built-in ONNX embeddings:

PRSense now includes built-in local embeddings via ONNX Runtime (Feature 7). No Python setup required!

# 1. Configure for local mode
vercel env add EMBEDDING_PROVIDER
# Enter: onnx

# 2. Remove OpenAI key (if present)
vercel env rm OPENAI_API_KEY

# 3. Redeploy
vercel --prod

That’s it! The app will now run 100% free using the all-MiniLM-L6-v2 model embedded in the application.

Costs Summary

  • 100 PRs/month: $0.01/month ≈ FREE
  • 1,000 PRs/month: $0.10/month ≈ FREE
  • 10,000 PRs/month: $1/month + $20 Vercel = $21/month

With Local ONNX (Free)

  • Any scale: $0/month
  • Hardware: CPU-only is sufficient
  • Privacy: No data leaves your server

What Happens After Deploy?

1. PR Opened

Developer opens PR #123: "Fix login bug"

2. Webhook Triggered

GitHub → Webhook → Your Vercel Function

3. PRSense Checks

Generate embeddings
Search for similar PRs
Calculate similarity scores
Make decision (2ms total)

4. Bot Responds

IF duplicate (>90% confidence):
→ Comment: " Duplicate of #100"
→ Add label: "duplicate"

IF possible (>82% confidence):
→ Comment: "ℹ Similar to #100, please review"

IF unique (<82%):
→ No action (avoid noise)

5. Maintainer Sees Result

Saved 30 minutes of review time!

Troubleshooting

”Webhook delivery failed"

# Check Vercel logs
vercel logs

# Verify secrets are set
vercel env ls

# Test webhook locally
curl -X POST https://your-app.vercel.app/webhook \
-H "Content-Type: application/json" \
-d '{"action":"opened","pull_request":{"number":1}}'

"OpenAI API error"

# Check you have credits
# Go to: https://platform.openai.com/account/billing

# Or switch to free local embeddings (see above)

"Database error”

# For development, just use SQLite (auto-configured)
# For production, set up Postgres:

# Option 1: Supabase (free tier has pgvector)
# https://supabase.com

# Option 2: Neon (free tier)
# https://neon.tech

# Add DATABASE_URL to vercel:
vercel env add DATABASE_URL
# postgresql://user:pass@host:5432/db

Next Steps After Deploy

Week 1: Monitor

  • Check webhook deliveries in GitHub
  • Review bot comments
  • Adjust thresholds if needed

Week 2: Tune

# Too many false positives?
vercel env add DUPLICATE_THRESHOLD
# Enter: 0.95 (stricter)

# Missing duplicates?
vercel env add DUPLICATE_THRESHOLD
# Enter: 0.85 (more aggressive)

Week 3: Expand

  • Add to more repos
  • Track time saved
  • Measure ROI

Support

Need help?

You’re Ready!

PRSense is now:

  • Deployed to production
  • Integrated with GitHub
  • Detecting duplicates automatically
  • Saving maintainer time

Estimated setup time: 5 minutes Estimated time saved: 5 hours/week ROI: 6000%

Deploy now: vercel