Skip to content

README.md

Latest commit

 

History

History
136 lines (99 loc) · 5.29 KB

README.md

File metadata and controls

136 lines (99 loc) · 5.29 KB
Raw

Deployment Guide

Deploy the backend pattern (FastAPI + Pixeltable) from this monorepo. Batch and serving patterns have their own deploy configs — see batch/deploy/ and serving/deploy/.

Storage

All options below should set PIXELTABLE_HOME=/data/pixeltable to a persistent volume. For large media workloads:

PIXELTABLE_INPUT_MEDIA_DEST=s3://your-bucket/input    # or gs:// or az://
PIXELTABLE_OUTPUT_MEDIA_DEST=s3://your-bucket/output

See Pixeltable Configuration.

Docker Compose

Local or single-server deployment:

cp .env.example .env          # add API keys
docker compose up --build     # http://localhost:8000

Pixeltable data persists via named Docker volumes: pixeltable-data (catalog + blobs at /data/pixeltable) and uploads (raw files at /app/data). To reset: docker compose down -v.

Platform Quick Reference

Platform Quick start Details
Fly.io cp deploy/fly/fly.toml . → fly launch → create volume → set secrets → fly deploy fly/README.md
Render cp deploy/render/render.yaml . → push → Blueprint Instance in dashboard render/README.md
Railway Deploy from GitHub; config path /deploy/railway/railway.json; volume at /data/pixeltable railway/README.md
DigitalOcean doctl apps create --spec deploy/digitalocean/app.yaml digitalocean/README.md
Vercel Frontend only — cp deploy/vercel/vercel.json frontend/ → npx vercel vercel/README.md
Helm Build/push image → helm install pixeltable-starter ./deploy/helm/pixeltable-starter helm/README.md
Terraform cd deploy/terraform-{k8s,gke,aks} → terraform apply terraform-k8s/README.md
AWS CDK cd deploy/aws-cdk && pip install -r requirements.txt && cdk deploy aws-cdk/README.md

Fly.io

cp deploy/fly/fly.toml .
fly launch --no-deploy
fly volumes create pxt_data --size 10 --region iad
fly secrets set OPENAI_API_KEY=sk-... ANTHROPIC_API_KEY=sk-ant-...
fly deploy

Render

cp deploy/render/render.yaml .
git add render.yaml && git commit -m "add render blueprint" && git push
# Then: Render dashboard → New → Blueprint Instance → connect repo

Railway

  1. railway.app/new → Deploy from GitHub repo
  2. Service → Settings → set config path to /deploy/railway/railway.json
  3. Set PIXELTABLE_HOME=/data/pixeltable, OPENAI_API_KEY, ANTHROPIC_API_KEY in Variables
  4. Add a Volume mounted at /data/pixeltable

DigitalOcean

doctl apps create --spec deploy/digitalocean/app.yaml

App Platform doesn't have native persistent volumes. See digitalocean/README.md for persistence options.

Vercel (frontend only)

cp deploy/vercel/vercel.json frontend/
cd frontend && npx vercel --yes
# Set BACKEND_URL=https://your-backend.fly.dev in Vercel dashboard

Deploys the React frontend on Vercel's edge CDN with /api proxied to your backend.

Helm (any existing Kubernetes cluster)

docker build -t <your-registry>/pixeltable-starter:latest .
docker push <your-registry>/pixeltable-starter:latest
helm install pixeltable-starter ./deploy/helm/pixeltable-starter \
  --set image.repository=<your-registry>/pixeltable-starter \
  --set secrets.OPENAI_API_KEY=sk-... \
  --set secrets.ANTHROPIC_API_KEY=sk-ant-...

Local testing with minikube:

minikube start --cpus=4 --memory=6144
docker build -t pixeltable-starter:latest .
minikube image load pixeltable-starter:latest
helm install pixeltable-starter ./deploy/helm/pixeltable-starter \
  --set image.pullPolicy=Never --set service.type=NodePort \
  --set secrets.OPENAI_API_KEY=$OPENAI_API_KEY \
  --set secrets.ANTHROPIC_API_KEY=$ANTHROPIC_API_KEY
kubectl port-forward svc/pixeltable-starter 9000:8000

Terraform (provision cluster from scratch)

cd deploy/terraform-k8s && terraform init && terraform apply   # AWS EKS
cd deploy/terraform-gke && terraform init && terraform apply   # GCP GKE
cd deploy/terraform-aks && terraform init && terraform apply   # Azure AKS

Each creates a managed K8s cluster with a 50Gi persistent volume.

AWS CDK (ECS Fargate)

cd deploy/aws-cdk && pip install -r requirements.txt && cdk deploy

Serverless containers with EFS for persistent storage and ALB for load balancing.

Batch processing deploy

For cron jobs, queue workers, and event-driven pipelines (no HTTP server):

Platform Config Best for
Cloud Run Jobs cloudbuild.yaml GCP, cron/Pub/Sub triggers
Kubernetes Job job.yaml, cronjob.yaml Any K8s, queue-driven scaling
ECS Fargate task-definition.json AWS Spot pricing
Lambda Dockerfile, handler.py Small batches, up to 15 min

See batch/README.md for details.