Deployment Status

Understanding deployment status helps you build reliable integrations and debug issues quickly. The status field returned by the Deployments API uses the values below.

Status Reference

A deployment advances through these statuses from creation to ready:

Status Phase Meaning
requested Created Request received and queued; provisioning starts in the background.
provisioning_supervisor Provisioning The supervisor instance is being launched.
provisioning_worker Provisioning Worker instance(s) are being launched.
waiting_for_xinference Starting Waiting for the Xinference server to come online.
launching_model Starting Loading the model weights into GPU memory.
running Ready Model is loaded and the inference endpoint is accepting requests.

Termination

Status Meaning
terminating_model Unloading the model and shutting the deployment down.
releasing_instance Releasing the underlying cloud instances.
terminated Fully shut down. The deployment record is retained for history.

Failure States

Status Meaning
failed An unrecoverable error occurred.
failed_cleaned The deployment failed and its resources were cleaned up.
cleanup_failed The deployment failed and automatic cleanup did not complete.

When a deployment fails, failure_stage records the lifecycle status it was in when the error occurred (e.g. provisioning_worker or launching_model), and failure_reason holds the error detail.

Checking Status

Dashboard

The deployment card shows a color-coded status badge that updates in real time.

API

Fetch a deployment to read its current status. Requests are authenticated with the session cookie set at sign-in:

import requests

r = requests.get(
    "https://api.xinference.co/api/v1/model-deployments/dep_abc123",
    cookies={"session": "<your-session-cookie>"},
)
print(r.json()["status"])
print(r.json().get("failure_reason"))

Notifications

The platform records deployment state transitions and can send email notifications when a deployment changes state. Email delivery and preferences are managed through the notification settings (see the backend's notification-settings API).