Tzurot v3 Observability & Operations

Use this skill when: Adding logging, debugging production issues, checking health, performing database/Redis operations, or adding new personalities.

Quick Reference

# Check service health
curl https://api-gateway-development-83e8.up.railway.app/health

# View logs
railway logs --service api-gateway
railway logs --service ai-worker | grep "ERROR"

# Find request across services
railway logs | grep "requestId\":\"abc-123"

// Structured logging pattern
logger.info({ requestId, userId, personalityId }, 'Processing request');
logger.error({ err: error, context: 'additional data' }, 'Operation failed');

Structured Logging

The Golden Pattern

logger.info(
  { contextObject }, // First param: structured data
  'Human readable message' // Second param: message string
);

// ✅ GOOD
logger.info({ personalityId, model }, 'Loaded personality');
logger.error({ err: error, requestId }, 'Failed to process');

// ❌ BAD - String interpolation loses structure
logger.info(`Loaded personality ${personalityId}`);

Log Levels

Error Logging

// Use 'err' key for Pino's special error serialization
logger.error({ err: error, requestId, userId }, 'Request failed');

Privacy Rules

NEVER log: API keys, tokens, passwords, email, IP, message content, DM content

SAFE to log: User IDs, channel IDs, counts, durations, error types

Correlation IDs

Track requests across services with requestId:

// bot-client: Generate and include in request
const requestId = randomUUID();
await fetch(url, { headers: { 'X-Request-ID': requestId } });

// All services: Include in logs
logger.info({ requestId, jobId }, 'Processing');

// Search across services
railway logs | grep "requestId\":\"abc-123"

Health & Metrics

Health endpoint: GET /health

curl https://api-gateway-development-83e8.up.railway.app/health

Check all services:

railway status
railway logs --service api-gateway --tail 50

Common Operations

Adding a Personality

1. Create personalities/name.json:

{
  "name": "PersonalityName",
  "systemPrompt": "Your description...",
  "model": "anthropic/claude-sonnet-4.5",
  "temperature": 0.8
}

2. Commit and push (Railway auto-deploys)

Database Quick Commands

# Check migrations
railway run npx prisma migrate status

# Apply migrations
railway run npx prisma migrate deploy

# Direct database access
railway run psql

# Open Prisma Studio
npx prisma studio

Redis Queue Status

railway run redis-cli
LLEN bull:ai-generation:wait
LLEN bull:ai-generation:active
LLEN bull:ai-generation:failed

Debugging Checklist

1. Check logs: railway logs --service <name>
2. Check health: curl .../health
3. Check env vars: railway variables --service <name>
4. Find errors: railway logs | grep '"level":"error"'
5. Trace request: railway logs | grep "requestId\":\"..."

Common Issues

Anti-Patterns

// ❌ BAD - console.log
console.log('Processing');

// ✅ GOOD - structured logger
logger.info({ messageId }, 'Processing');

// ❌ BAD - logging in loops
for (const item of items) {
  logger.info({ item }, 'Processing'); // Spams logs!
}

// ✅ GOOD - log summary
logger.info({ count: items.length }, 'Processing items');

// ❌ BAD - swallowing errors
try {
  doSomething();
} catch (e) {
  /* nothing */
}

// ✅ GOOD - log and handle
try {
  doSomething();
} catch (e) {
  logger.error({ err: e }, 'Failed');
  throw e;
}

Railway Log Commands

# Recent logs
railway logs --service api-gateway

# Follow in real-time
railway logs --service ai-worker --tail

# Since time
railway logs --since 1h

# Find errors
railway logs | grep '"level":"error"'

# Count error types
railway logs --since 24h | grep error | jq '.msg' | sort | uniq -c

Related Skills

• tzurot-deployment - Railway service management, rollbacks
• tzurot-db-vector - Database migrations, pgvector
• tzurot-security - Privacy in logging
• tzurot-async-flow - Job correlation, BullMQ

References

• Pino docs: https://getpino.io/
• Railway CLI: docs/reference/RAILWAY_CLI_REFERENCE.md
• Logger utility: packages/common-types/src/utils/logger.ts