Applad treats cloud provider resources as adapters you draw from when they make sense for a specific resource. You’re not locked into a platform — you use cloud services surgically.
You might:
- Run your core app on a Hetzner VPS (fast and cheap)
- Use S3 for storage (excellent pricing)
- Use SES for high-volume email at scale
- Spin up a cloud compute instance for heavy processing — pay for duration, then tear it down
Your application code never changes. Only the adapter config.
Storage Adapters
Switch between local filesystem, S3, R2, GCS, or any S3-compatible service without changing application code.
Local Filesystem
adapter: "filesystem"
config:
base_path: "/var/applad/storage"
max_file_size: "100MB"
AWS S3
adapter: "s3"
config:
region: "us-east-1"
bucket: ${S3_BUCKET}
access_key: ${AWS_ACCESS_KEY_ID}
secret_key: ${AWS_SECRET_ACCESS_KEY}
Cloudflare R2
adapter: "s3"
config:
endpoint: ${R2_ENDPOINT}
bucket: ${R2_BUCKET}
access_key: ${R2_ACCESS_KEY}
secret_key: ${R2_SECRET_KEY}
Google Cloud Storage
adapter: "gcs"
config:
project_id: ${GCP_PROJECT_ID}
bucket: ${GCS_BUCKET}
credentials: ${GCP_CREDENTIALS_JSON}
Backblaze B2
adapter: "s3"
config:
endpoint: "s3.us-west-004.backblazeb2.com"
bucket: ${B2_BUCKET}
access_key: ${B2_KEY_ID}
secret_key: ${B2_APPLICATION_KEY}
Applad uses Rclone internally, supporting 40+ storage backends. Any Rclone-compatible provider works as an adapter.
Database Adapters
Run SQLite locally, Postgres on VPS, then switch to managed RDS — same config structure, different adapter target.
PostgreSQL (Self-Hosted)
connections:
primary:
adapter: "postgres"
host: "localhost"
port: 5432
database: ${DB_NAME}
user: ${DB_USER}
password: ${DB_PASSWORD}
migrations:
dir: "database/migrations/primary"
AWS RDS (Managed PostgreSQL)
connections:
primary:
adapter: "postgres"
host: ${RDS_HOST}
port: 5432
database: ${RDS_DATABASE}
user: ${RDS_USER}
password: ${RDS_PASSWORD}
ssl: true
migrations:
dir: "database/migrations/primary"
Google Cloud SQL
connections:
primary:
adapter: "postgres"
host: ${CLOUD_SQL_HOST}
port: 5432
database: ${CLOUD_SQL_DATABASE}
user: ${CLOUD_SQL_USER}
password: ${CLOUD_SQL_PASSWORD}
ssl: true
migrations:
dir: "database/migrations/primary"
SQLite (Local Development)
connections:
primary:
adapter: "sqlite"
path: "/var/applad/data/app.db"
migrations:
dir: "database/migrations/primary"
Multiple Database Connections
Run different databases for different concerns:
connections:
primary:
adapter: "postgres"
host: ${PRIMARY_DB_HOST}
migrations:
dir: "database/migrations/primary"
analytics:
adapter: "postgres"
host: ${ANALYTICS_DB_HOST}
migrations:
dir: "database/migrations/analytics"
cache:
adapter: "redis"
host: ${REDIS_HOST}
port: 6379
database: field:
database/tables/users.yaml
name: "users"
database: "primary"
# ...
database/tables/events.yaml
name: "events"
database: "analytics"
# ...
Cross-database relations are detected at validation time with actionable errors. Handle joins at the application layer.
Messaging Adapters
Switch between SMTP, SendGrid, SES, Postmark, or Mailgun based on volume and cost.
SMTP (Self-Hosted)
email:
adapter: "smtp"
config:
host: ${SMTP_HOST}
port: 587
user: ${SMTP_USER}
password: ${SMTP_PASSWORD}
from: "[email protected]"
AWS SES (High-Volume)
email:
adapter: "ses"
config:
region: "us-east-1"
access_key: ${AWS_ACCESS_KEY_ID}
secret_key: ${AWS_SECRET_ACCESS_KEY}
from: "[email protected]"
SendGrid
email:
adapter: "sendgrid"
config:
api_key: ${SENDGRID_API_KEY}
from: "[email protected]"
Postmark
email:
adapter: "postmark"
config:
server_token: ${POSTMARK_SERVER_TOKEN}
from: "[email protected]"
SMS Adapters
sms:
adapter: "twilio"
config:
account_sid: ${TWILIO_ACCOUNT_SID}
auth_token: ${TWILIO_AUTH_TOKEN}
from: ${TWILIO_PHONE_NUMBER}
Compute Adapters
Run functions locally in Docker, burst to cloud compute for heavy workloads.
Local Docker (Default)
functions/process-video.yaml
name: "process-video"
runtime: "python:3.11"
memory: "2GB"
timeout: "300s"
source:
type: "local"
path: "./src/functions/process-video"
AWS Lambda (On-Demand)
functions/process-video.yaml
name: "process-video"
runtime: "python:3.11"
memory: "10GB"
timeout: "900s"
source:
type: "local"
path: "./src/functions/process-video"
compute:
adapter: "lambda"
region: "us-east-1"
Cloud Build Instances
For workloads that need specific hardware:
deployments/ios-production.yaml
name: "ios-production"
type: "ios"
build:
adapter: "aws-mac"
instance_type: "mac2.metal"
region: "us-east-1"
source:
repo: "github.com/acme/mobile-app"
branch: "main"
output:
target: "appstore"
CDN Adapters
Serve static assets through CDN:
Cloudflare
storage/buckets/public.yaml
name: "public"
public: true
cdn:
adapter: "cloudflare"
zone_id: ${CF_ZONE_ID}
api_token: ${CF_API_TOKEN}
domain: "cdn.example.com"
AWS CloudFront
storage/buckets/public.yaml
name: "public"
public: true
cdn:
adapter: "cloudfront"
distribution_id: ${CF_DISTRIBUTION_ID}
domain: "cdn.example.com"
Deployment Continuum
Move along the continuum as your needs grow:
Local dev (Docker Compose, SQLite, filesystem)
↓
VPS staging (Docker Compose, Postgres, filesystem)
↓
VPS production (Docker Compose, Postgres, filesystem)
↓
VPS + S3 storage adapter
↓
VPS + RDS database adapter
↓
VPS + SES email adapter
↓
Multi-VPS cluster + managed services
↓
Kubernetes on cloud
- Application code unchanged
- Only adapter config changes
applad up reconciles to new state
Switching Adapters
Preview the change:
applad up --env production --dry-run --diff
DIFF: storage/storage.yaml
- adapter: "filesystem"
+ adapter: "s3"
ACTIONS:
✓ Storage adapter will be changed to S3
✓ Existing files will be migrated
✓ Application code: no changes required
applad up --env production
Cost Optimization Strategies
Start Simple
# Local dev
storage → filesystem
database → sqlite
email → smtp (on VPS)
compute → docker (on VPS)
Add Adapters as Needed
# Growing production
storage → S3 (pay per GB)
database → Postgres (on VPS or RDS)
email → SES (pay per send)
compute → Docker (on VPS) + Lambda (for spikes)
Scale What Makes Sense
- S3 at scale: $0.023/GB/month — cheaper than VPS disk at high volumes
- SES for email: $0.10/1000 emails — cheaper than dedicated SMTP
- RDS when needed: Managed backups, high availability — worth it at scale
- Lambda for spikes: Pay only for execution, zero idle cost
The sweet spot is usually: VPS for core app + cloud adapters for specific resources.
Secrets for Cloud Adapters
Cloud credentials are secrets — never in config files:
# Set AWS credentials for production
applad secrets set AWS_ACCESS_KEY_ID --env production
applad secrets set AWS_SECRET_ACCESS_KEY --env production
# Set different credentials for staging
applad secrets set AWS_ACCESS_KEY_ID --env staging
applad secrets set AWS_SECRET_ACCESS_KEY --env staging
adapter: "s3"
config:
access_key: ${AWS_ACCESS_KEY_ID}
secret_key: ${AWS_SECRET_ACCESS_KEY}
Adapter Validation
Applad validates adapter configs before applying:
applad up --env production --dry-run
ERROR storage/storage.yaml line 3
S3 adapter requires "bucket" field
Add: bucket: ${S3_BUCKET}
ERROR Missing required variable S3_BUCKET
Referenced by: storage/storage.yaml line 3
Set it with: applad secrets set S3_BUCKET --env production
Next Steps