Deployment options

Yabanin runs where you put it. The gateway controls configured traffic; where inference runs depends on the model endpoints you configure.

Local-only configuration

Use local inference (e.g. llama.cpp behind the gateway) when the workload must stay inside your chosen environment. Verify network rules and the full dependency chain — including tools, telemetry, logs and backups.

See: air-gap checklist.

Approved-cloud configuration

Run Yabanin in your VPC or on premises and configure approved external inference routes (OpenAI-compatible providers). Review provider terms and data settings for each enabled service and feature.

See: install guide.

The boundary

Self-hosting the gateway does not turn external inference into local inference. Both configurations require secure installation and ongoing operation. For the data-handling questions to answer before expanding, see the agent readiness checklist.