Skip to main content

Custom models enable private deployments

Custom models are the foundation of a private deployment—wire up your own endpoints with settings.json, then apply the additional steps on this page to keep Zencoder entirely inside your boundary.

​
What Is a Private Deployment?

The standard Zencoder cloud experience is already secure for most teams: code runs ephemerally, data is encrypted in transit and at rest, access is gated by role-based controls, and Zencoder maintains ISO 27001/42001 plus SOC 2 Type II certifications. If you are comfortable with those guarantees, you can keep using the default service without extra configuration. A private deployment gives you deeper control by running agents, orchestration, and inference on infrastructure you operate. Typical benefits include:
  • Full isolation inside your own network perimeter.
  • Support for customer-hosted models and runtimes.
  • Zero dependency on external endpoints for sensitive workloads.
Private deployments are recommended for regulated industries, strict data-residency mandates, air-gapped environments, and teams with proprietary models they must keep on-premise.

​
Configure a Private Deployment

A full private deployment is a layered approach. Follow these steps to keep Zencoder inside infrastructure you manage:
  1. Configure custom models from local or VPC endpoints. Use the Custom Models configuration guide to declare every model you plan to expose to your users. This is how agents call the runtimes that live within your network boundary.
  2. Hide the managed catalog. Inside settings.json, set "useDefaultProviders": false so the selector only lists the providers you declared. This prevents accidental routing to Zencoder-hosted models after you stand up custom ones.
  3. Deactivate code completion. Code completion today runs on Zencoder-managed models. In VS Code open the Zencoder menu → Settings and uncheck Zencoder Code Completion. In JetBrains go to Tools → Zencoder → Settings and uncheck Enable code completion (see Code Completion for screenshots). This keeps autocomplete traffic from leaving your environment while the team codes against private models.
After applying these steps, all interactive chat and agent runs are served by your endpoints.
We are actively building additional controls to make fully-private deployments even more seamless. Stay tuned for updates in the changelog.