Get a VM with its GPU/node/template info and price

compute-fabric capability tier (e.g. "pod"); empty for a plain VM

compute-fabric purchase option (on_demand; pods created before the rename carry the legacy reserved or interruptible values); empty for a plain VM

Guest RAM (MiB) this VM is running with, stamped at placement. Absent on a VM that has never been placed.

Guest vCPU this VM is running with, stamped by the same placement write as guestMemMb.

pods only: free disk on the pod's host node, shared across the pods on that node. Pods have no per-pod quota, and pod storage does not persist across terminate. Absent when the node reports no live disk signal.

pods only: total disk on the pod's host node, shared across the pods on that node. Pods have no per-pod quota, and pod storage does not persist across terminate. Absent when the node reports no live disk signal.

RFC3339 time the workload became interrupted. Cleared when it recovers; kept on a workload that was stopped or deleted because of the interruption.

This VM's usage is recorded but never charged: a provider org consuming GPUs on a node it owns under its own reservation. pricePerHourCents still reports the real rate (the measurement is kept, the charge is waived) so a client showing a rate MUST also show this or it will display a bill that never arrives.

bundled hourly rate. When the VM has an open usage session this is the rate LOCKED at session open (what settlement charges); otherwise the rate a session WOULD lock: the fabric admission snapshot for a pod, else the current catalog rate for the VM's shape.

true = open endpoint (no data-plane auth); false = require an org API key

Guest RAM (MiB) this VM asked for, or the value auto resolved to at admission. Placement honours it, so a co-tenant landing afterwards cannot shrink the VM below it. The exception is the node the VM's disk lives on: when a stopped VM is started, or re-dispatched after its job is lost, and that node's per-GPU share times gpuCount is below this value, and that share still clears the floorPerGpu RAM times gpuCount, the VM starts at the share and this field is updated to it.

managed-service tag; empty for a plain VM

ready-to-paste ssh command (ssh <handle>@<host>)

RunPod-style SSH handle (first label of endpointUrl); used as the SSH username

data-plane SSH gateway host, e.g. ssh.run.beta.openrelay.inc

true when SSH should work now. Observed: a node-side probe saw the guest sshd answer. Falls back to the running-status guess on nodes not yet reporting probes (sshState=unverified).

why sshReady is what it is. booting = the VM is running but its guest OS has not started sshd yet (GPU VMs take minutes); unverified = no probe signal for this node, sshReady is a guess; offline = the VM is not running.

human-readable explanation of sshState, safe to show to the customer

login user the gateway lands as inside the VM (e.g. ubuntu)

Cause of the current status; see VmSummary.statusReason for the machine-readable values.

RFC3339 time the VM was terminated. Set only once status is "terminated" (a stopped/failed VM can restart, so it has none); absent while the VM is live. On a terminated VM, healthy is false and provisioningStage is absent.

pods only: size in GB of the pod's persistent volume at volumeMountPath. Absent on a pod created before volumes existed.

pods only: where the persistent volume is mounted inside the container (/workspace).

pods only: RFC3339 time of the volumeUsedBytes report.

pods only: bytes the volume currently holds, as of volumeUsedAt (the node reports daily). Absent until the node has reported.