Deployment Pipeline

How to go from code to running system — for cloud, on-premise, and air-gapped environments.

---

Overview

┌─────────────────────┐     ┌──────────────┐     ┌─────────────────────┐
│  Internet Machine    │     │   Transfer   │     │  Target Server      │
│  (your laptop or CI) │ ──► │   (USB/SCP)  │ ──► │  (VXRail/EC2/bare)  │
│                      │     │              │     │                     │
│  1. yarn install     │     │  .tar.gz     │     │  4. docker load     │
│  2. yarn build       │     │  (~3-6 GB)   │     │  5. docker compose  │
│  3. docker build     │     │              │     │  6. seed data       │
│  3b. docker save     │     │              │     │                     │
└─────────────────────┘     └──────────────┘     └─────────────────────┘

---

Phase 1: Build (Internet-Connected Machine)

1a. Install Dependencies

cd medOS-ultra

# Fix known dep issue
sed -i 's/"moleculer-db-adapter-mongoose": "\^0.9.12"/"moleculer-db-adapter-mongoose": "^0.10.1"/' services/interoperability/package.json

# Install
yarn install --network-timeout 600000

1b. Build Shared Packages + API Services

# Build shared platform packages first (required by services)
yarn build:platform

# Build all microservices
yarn build:api

# Verify — each service should have a dist/ folder
ls services/gateway/dist/
ls services/auth/dist/
ls services/clinical/dist/
# etc.

1c. Build Docker Images

# Build the single API image (used by all 13 NestJS services)
docker build -t medos-api:latest \
  -f docker/api/Dockerfile \
  --network=host \
  --build-arg SCOPE=ever-api-gateway .

# Build supporting images
docker build -t medos-eform:latest -f docker/eform/Dockerfile .   # if eform paths are set up
docker pull mongo:5
docker pull postgres:14.5
docker pull nats:2-alpine
docker pull nginx:1.22
docker pull ipfs/go-ipfs:master

1d. Save Images for Transfer

docker save \
  medos-api:latest \
  mongo:5 postgres:14.5 nats:2-alpine nginx:1.22 \
  "ipfs/go-ipfs:master" \
  | gzip > medos-images.tar.gz

ls -lh medos-images.tar.gz  # ~3-5 GB

1e. Or Use the Packaging Script

cd infrastructure
./scripts/package-airgap.sh thailand
# Output: medos-airgap-thailand-YYYYMMDD.tar.gz

---

Phase 2: Transfer

Method	Speed	Use When
USB 3.0 drive	~100 MB/s, ~60s	Air-gapped hospital
SCP/rsync over LAN	~100 MB/s	Same network
S3 upload → download	Varies	Cloud staging

# USB
cp medos-airgap-thailand-*.tar.gz /Volumes/USB_DRIVE/

# SCP
scp medos-airgap-thailand-*.tar.gz user@hospital-server:/opt/medos/

# rsync (resumable)
rsync -avzP medos-airgap-thailand-*.tar.gz user@server:/opt/medos/

---

Phase 3: Deploy on Target Server

3a. Prerequisites on Server

# Ubuntu 22.04 with Docker installed
docker --version   # 24.x+
docker compose version  # v2.x

3b. Extract and Load

cd /opt/medos
tar xzf medos-airgap-thailand-*.tar.gz
cd medos-airgap-thailand-*

# Load Docker images (no internet needed)
gunzip -c medos-images.tar.gz | docker load
# Takes 5-10 minutes

3c. Configure

# Generate SSL cert
bash scripts/generate-ssl-cert.sh medos.local

# Create env file
cp .env.onpremise .env.onpremise.thailand
nano .env.onpremise.thailand
# Set: BASE_URL, TZ=Asia/Bangkok, JWT_SECRET, SOCKET_IO_TOKEN

# Add to /etc/hosts
echo "127.0.0.1 medos.local" | sudo tee -a /etc/hosts

3d. Start Everything

docker compose \
  --env-file .env.onpremise.thailand \
  -f docker-compose-onpremise.yml \
  up -d

# Watch startup
docker compose -f docker-compose-onpremise.yml logs -f
# Wait for "Gateway started" (~60-90 seconds)

3e. Load Seed Data

# Thailand (default seeds in web/scripts/ and web/supabase/migrations/)
# These need Supabase configured — run against your Supabase instance

# Japan seeds (additional)
for sql in market-packs/medos-japan/seed-*.sql; do
  echo "Loading: $sql"
  psql "$SUPABASE_DB_URL" -f "$sql"
done

# Philippines seeds (additional)
for sql in market-packs/medos-philippines/seed-*.sql; do
  echo "Loading: $sql"
  psql "$SUPABASE_DB_URL" -f "$sql"
done

# English seeds (additional)
for sql in market-packs/medos-english/seed-*.sql; do
  echo "Loading: $sql"
  psql "$SUPABASE_DB_URL" -f "$sql"
done

# Blood donor seeds (TypeScript — need Node on server)
cd /path/to/web
cp .env.onpremise.thailand .env
npx tsx ../infrastructure/market-packs/medos-japan/seed-blood-donors.ts
npx tsx ../infrastructure/market-packs/medos-philippines/seed-blood-donors.ts
npx tsx ../infrastructure/market-packs/medos-english/seed-blood-donors.ts

3f. Verify

# All containers running
docker compose -f docker-compose-onpremise.yml ps

# API health
curl -k https://medos.local/api/health

# Open browser
open https://medos.local

---

Deployment Modes Summary

Mode	Build Where	Run Where	Internet Needed
Cloud (AWS)	GitHub Actions / local	ECS Fargate	Yes (always)
EC2 Demo	On the EC2 itself	Docker Compose on EC2	Yes (for build)
On-Premise (connected)	On the server	Docker Compose	Yes (for initial build)
Air-Gapped	Internet machine → USB	Docker Compose	No (uses pre-built images)

---

Seed Data by Region

All seeds are additive and idempotent. You can load multiple regions into the same database.

Market Pack	IDs Prefix	Locale	Facilities
Thailand (default)	(none)	th	Hospital + Clinic
Japan	jp-*	ja	Hospital + 特別養護老人ホーム
Philippines	ph-*	fil	Hospital + Nursing Home
English	en-*	en	Hospital + Skilled Nursing

---

Updating an Existing Installation

# On internet machine: rebuild
yarn install && yarn build:platform && yarn build:api
docker build -t medos-api:latest -f docker/api/Dockerfile --network=host .
docker save medos-api:latest | gzip > medos-api-update.tar.gz

# Transfer to server
scp medos-api-update.tar.gz user@server:/opt/medos/

# On server: load new image and restart
gunzip -c medos-api-update.tar.gz | docker load
docker compose -f docker-compose-onpremise.yml up -d
# Data volumes are preserved — no data loss

---

CI/CD (GitHub Actions)

For automated deployment to the demo EC2:

# .github/workflows/deploy.yml
name: Deploy to Demo
on:
  push:
    branches: [main]
jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: yarn install && yarn build:platform && yarn build:api
      - run: |
          ssh user@${{ secrets.DEMO_SERVER_IP }} "
            cd /opt/medos/medOS-ultra
            git pull
            yarn install && yarn build:platform && yarn build:api
            docker compose -f infrastructure/docker-compose-onpremise.yml restart
          "

---

Quick Reference

# Build (internet machine)
yarn install && yarn build:platform && yarn build:api

# Package for air-gap
cd infrastructure && ./scripts/package-airgap.sh thailand

# Deploy (target server)
gunzip -c medos-images.tar.gz | docker load
docker compose --env-file .env.onpremise.thailand -f docker-compose-onpremise.yml up -d

# Load all seeds
./deploy.sh thailand seed
./deploy.sh japan seed
./deploy.sh philippines seed

# Start/Stop/Restart
docker compose -f docker-compose-onpremise.yml up -d
docker compose -f docker-compose-onpremise.yml down
docker compose -f docker-compose-onpremise.yml restart

# Backup
./backup.sh

# Update (no data loss)
docker load < new-images.tar
docker compose -f docker-compose-onpremise.yml up -d