This document describes TendSocial's CI/CD pipelines and how they're optimized for performance.

Overview

TendSocial uses GitHub Actions for continuous integration and deployment:

Workflow	Purpose	Trigger
backend-ci.yml	Test, build, deploy backend to Cloud Run	Push/PR to main
frontend-ci.yml	Lint, test, build frontend	Push/PR to main
marketing-ci.yml	Lint, test, build marketing site	Push/PR to main
documentation-ci.yml	Build documentation	Push/PR to main

Backend CI Pipeline

Job Structure

flowchart LR
    subgraph Parallel["Parallel Execution"]
        Test[Test Job]
        Build[Build Job]
    end
    Test & Build --> Deploy[Deploy Job]
    Deploy --> CloudRun[Cloud Run]

Jobs

1. test - Runs in parallel with build

   • Starts PostgreSQL service container
   • Runs pnpm install
   • Generates Prisma client
   • Runs database migrations
   • Lints TypeScript (tsc --noEmit)
   • Runs Vitest tests

2. build - Runs in parallel with test

   • Runs pnpm install
   • Generates Prisma client
   • Builds TypeScript (tsc)

3. deploy - Only on main branch, after test + build

   • Builds Docker image
   • Pushes to Google Artifact Registry
   • Deploys to Cloud Run with --cpu-boost

Performance Optimizations

Caching Strategy

Cache Type	Key Strategy	Benefit
pnpm store	Lockfile hash	Skip ~20s package download
Turbo cache	Lockfile hash	Skip unchanged package builds
Docker layers	GHA cache (type=gha)	Skip unchanged layers

Key Optimizations

1. Parallel Jobs

   • test and build run simultaneously
   • Saves ~50s vs sequential execution

2. Skip Puppeteer Chrome Download

   env:
     PUPPETEER_SKIP_DOWNLOAD: 'true'
     PUPPETEER_SKIP_CHROMIUM_DOWNLOAD: 'true'

   • Saves ~8s per workflow
   • Production uses Alpine's system Chromium instead

3. Cloud Run CPU Boost

   flags: '--allow-unauthenticated --cpu-boost'

   • Free temporary CPU boost during cold starts
   • Reduces cold start latency

4. Turbo Cache Key

   key: turbo-backend-${{ hashFiles('pnpm-lock.yaml') }}

   • Uses lockfile hash instead of git SHA
   • Enables cache reuse across commits

Dockerfile Optimization

The backend Dockerfile is optimized for:

1. Multi-stage builds - Smaller final image
2. Turbo prune - Only includes necessary monorepo packages
3. System Chromium - Uses Alpine's chromium package instead of Puppeteer's bundled version
4. Skip browser downloads - PUPPETEER_SKIP_DOWNLOAD=true in base stage

Troubleshooting

Cache Not Found

If you see "Cache not found" in logs:

• First run after lockfile changes will have cache miss
• Subsequent runs should hit the cache

Slow pnpm Install

Check for:

• "pnpm cache is not found" message → cache miss, expected first run
• ~7s install time = cache hit ✓
• ~25s install time = cache miss

Deploy Job Slow

The deploy job includes:

• Docker build (~2-3 min with cache hits)
• Docker push (~1 min)
• Cloud Run deploy (~1-2 min)
• GHA cache export (~2-3 min)

This is expected for production deployments.

---

Last Updated: December 2024
Maintainer: TendSocial Team