debug-prod-issues

What I Do

• Debug production issues on Vercel
• View and filter Vercel function logs
• Wait for deployments correctly
• Connect to production Neon database
• Inspect tables and data for debugging

Vercel Debugging (0 Finance)

IMPORTANT: Always use --scope prologe for the 0 Finance project:

# All Vercel commands need --scope prologe
vercel logs www.0.finance --scope prologe
vercel ls --scope prologe
vercel env ls --scope prologe

Viewing Logs

# Stream live logs (waits for new logs)
vercel logs www.0.finance --scope prologe

# Logs since a specific time
vercel logs www.0.finance --scope prologe --since 5m
vercel logs www.0.finance --scope prologe --since 1h

# Filter logs (pipe to grep)
vercel logs www.0.finance --scope prologe 2>&1 | grep -i "error"
vercel logs www.0.finance --scope prologe 2>&1 | grep "ai-email"

Note: vercel logs can be slow/hang. Use timeout or Ctrl+C if needed:

timeout 30 vercel logs www.0.finance --scope prologe --since 5m 2>&1 | head -100

Checking Deployments

# List recent deployments
vercel ls --scope prologe | head -10

# Check specific deployment status
vercel inspect <deployment-url> --scope prologe

# Get latest deployment URL
vercel ls --scope prologe 2>/dev/null | head -1

Waiting for Deployments

DO NOT just sleep and hope. Use vercel inspect --wait:

# Get the latest deployment URL
LATEST=$(vercel ls --scope prologe 2>/dev/null | head -1)

# Wait for it to be ready (up to 5 minutes)
vercel inspect "$LATEST" --scope prologe --wait --timeout 5m

# Check the status in the output:
#   status: ● Building  -> still building
#   status: ● Ready     -> deployed and live!
#   status: ● Error     -> build failed

Full workflow after pushing code:

# 1. Push your changes
git push origin main

# 2. Wait a few seconds for Vercel to pick it up
sleep 5

# 3. Get the new deployment URL
LATEST=$(vercel ls --scope prologe 2>/dev/null | head -1)
echo "Waiting for: $LATEST"

# 4. Wait for it to complete
vercel inspect "$LATEST" --scope prologe --wait --timeout 5m

# 5. Now your changes are live!

Check deployment details:

# See build info, aliases, and status
vercel inspect <deployment-url> --scope prologe

# See build logs if something failed
vercel inspect <deployment-url> --scope prologe --logs

Triggering a Redeploy

# Redeploy production from current state
vercel --prod --scope prologe

# Or push to git and wait
git push origin main
# Then use the waiting method above

Environment Variables

# List env vars
vercel env ls --scope prologe

# Add an env var (will prompt for value)
echo "value" | vercel env add VAR_NAME production --scope prologe

# Pull env vars to local .env
vercel env pull .env.local --scope prologe

Database Debugging

Environment Setup

CRITICAL: Always load .env.production.local for production database:

import * as dotenv from 'dotenv';
import path from 'path';

// Load production env
dotenv.config({
  path: path.resolve(__dirname, '../.env.production.local'),
});

Or in a script:

cd /path/to/zerofinance/packages/web
pnpm tsx -e "
import * as dotenv from 'dotenv';
dotenv.config({ path: '.env.production.local' });

// Now use db...
import { db } from './src/db';
"

Database Locations

Zero Finance uses Neon Postgres:

Environment	Host Pattern	Used By
Development	ep-aged-cherry-*	Local dev, some scripts
Production	ep-wispy-recipe-* or similar	Vercel deployment, prod data

Always verify which database you're connecting to:

[DB] Connecting to database host: ep-xxxxx-pooler.us-east-1.aws.neon.tech

Key Tables

Table	Purpose
user_safes	Safe addresses linked to users/workspaces
incoming_deposits	Incoming USDC transfers (synced from Safe API)
outgoing_transfers	Outgoing transactions from Safes
users	User records with primaryWorkspaceId
workspace_members	User-workspace membership
earn_deposits	Vault deposit records
ai_email_sessions	AI email agent conversation sessions

Script Template

import * as dotenv from 'dotenv';
dotenv.config({ path: '.env.production.local' });

import { db } from './src/db';
import { userSafes, incomingDeposits } from './src/db/schema';
import { eq } from 'drizzle-orm';

async function main() {
  // Your debugging code here
  const safes = await db.select().from(userSafes);
  console.log('Total safes:', safes.length);
}

main()
  .then(() => process.exit(0))
  .catch(console.error);

Common Issues

1. "Internal server error" from API endpoint

Debug steps:

1. Check Vercel logs for the specific endpoint
2. Look for stack traces or error messages
3. Test with curl to see response:

   curl -s -X POST "https://www.0.finance/api/endpoint" \
     -H "Content-Type: application/json" \
     -d '{"test": true}'
   

2. Code changes not taking effect

Causes:

• Deployment not complete yet
• Environment variables not updated (need redeploy)
• Caching issues

Fix:

1. Verify new deployment exists: vercel ls --scope prologe | head -3
2. Check deployment time matches your push
3. Force redeploy if needed: vercel --prod --scope prologe

3. Safe Not Found for Workspace

Debug:

const safe = await db.query.userSafes.findFirst({
  where: eq(userSafes.safeAddress, '0x...'),
});
const user = await db.query.users.findFirst({
  where: eq(users.privyDid, safe.userDid),
});
console.log('Safe workspace:', safe.workspaceId);
console.log('User primary workspace:', user.primaryWorkspaceId);

4. Environment variable not working

# Check it's set in Vercel
vercel env ls --scope prologe | grep VAR_NAME

# Check which environments it's set for (production, preview, development)
# May need to redeploy for changes to take effect

When to Use This Skill

• Investigating production API errors
• Checking if deployments completed
• Viewing function logs
• Debugging database/data issues
• Verifying environment variables
• Running one-off database scripts

---

Integration with Testing Strategy

This skill is part of the testing pyramid. Use it when:

1. Staging tests pass but production fails → Check prod logs
2. Need to verify fix in production → Inspect DB state
3. Feature works locally but not in prod → Compare environments

Related Skills

Scenario	Skill to Load
Testing on staging first	test-staging-branch
Making code testable	testability
After debugging, capture learnings	skill-reinforcement

Debugging Workflow (Fast → Slow)

1. Check Vercel logs first (fastest)
2. If unclear, inspect production DB
3. If still unclear, reproduce locally
4. After fix, test on staging before prod
5. Update this skill with new patterns

---

Learnings Log

Append new discoveries here

2026-01-13: prod DB script must load env before db import

Symptom: POSTGRES_URL missing or wrong host while using .env.production.local. Root Cause: packages/web/src/db/index.ts loads .env.local on import, which can run before dotenv config. Fix: In one-off scripts, call dotenv.config({ path: '.env.production.local' }) first and dynamically import ./src/db afterward. Prevention: Avoid static imports of ./src/db in CLI scripts; use dynamic import after dotenv setup.

Template for New Learnings

### YYYY-MM-DD: [Issue Description]

**Symptom**: [What you saw]
**Root Cause**: [Why it happened]
**Fix**: [How to resolve]
**Prevention**: [How to avoid in future]