Deployment Platforms

Rule Of Thumb

Peek is a single-primary service. It stores metadata in SQLite under PEEK_DATA, stores the generated signing/encryption key there when PEEK_SECRET is not set, and stores uploaded HTML there when PEEK_STORAGE=file.

For Docker and cloud platforms, mount durable storage at /data and set PEEK_DATA=/data. A Docker VOLUME declaration does not make a managed platform preserve files by itself; the platform must attach persistent storage at the same path.

Use one running instance against one data directory. S3-compatible storage can move uploaded HTML bytes out of /data, but /data/peek.db and /data/secret.key still need to survive restarts and deploys.

Base URL

Set PEEK_BASE_URL before the first real deployment. Peek uses it for generated share links, first-run setup URLs, OAuth callbacks, CLI login approval URLs, secure cookie behavior, and redirects.

The simplest production path is a custom domain:

PEEK_BASE_URL=https://peek.example.com

On platforms that assign a default hostname, avoid a deploy-redeploy loop by choosing the service name first and setting PEEK_BASE_URL to the expected platform URL before the first deploy. If the platform assigns a different URL, update PEEK_BASE_URL and redeploy once before inviting users or configuring OAuth.

Platform Matrix

Render

Create a Docker web service from the Peek image or repository. In the service settings, add a persistent disk with mount path /data. Only files written under that mount path survive deploys and restarts.

Set environment variables:

PEEK_DATA=/data
PEEK_BASE_URL=https://peek.example.com
PEEK_TRUSTED_PROXY=true

Use your custom domain if you have one. If you are using Render’s default domain, choose the service name first and set PEEK_BASE_URL to the expected https://<service-name>.onrender.com URL before the first deploy. If Render shows a different public URL after creation, update the value and redeploy before configuring OAuth or sharing links.

Set the health check path to /healthz. Use one instance because SQLite is the metadata store.

Fly.io

Create a volume and mount it at /data. In fly.toml, the mount destination should be /data, and the app should set PEEK_DATA=/data.

Keep one primary Machine for the app. Fly Volumes are local to Machines, so scaling to multiple Machines does not create one shared SQLite data directory.

Railway

Add a volume to the service and set its mount path to /data. Set PEEK_DATA=/data and PEEK_BASE_URL to the Railway domain or custom domain you will use.

Keep one replica. If you scale the service horizontally, multiple containers will not safely share one SQLite writer.

Kubernetes

Use a Deployment or StatefulSet with one replica and mount a PersistentVolumeClaim at /data.

Minimum storage shape:

volumeMounts:
  - name: peek-data
    mountPath: /data
volumes:
  - name: peek-data
    persistentVolumeClaim:
      claimName: peek-data

Use a ReadWriteOnce claim unless your storage class and SQLite operating model have been deliberately tested. Set PEEK_BASE_URL to the public Ingress URL and protect /metrics at the Ingress or network layer.

Sources

Platform behavior changes over time, so check the current provider docs before committing to a production setup: Render persistent disks, Render web services, Fly Volumes, Railway volumes, Kubernetes Persistent Volumes, Heroku dynos, DigitalOcean App Platform storage, AWS App Runner application code, and Cloud Run ephemeral disk.