Create a fleet

Create a new warm pool of pre-initialized compute nodes

POST

/auth/v1/seed/{namespace}/fleets

Path parameters

namespace string (required)

The namespace (user or organization) to create the fleet in.

name string (required)

A unique name for the fleet within the namespace. Allowed characters: alphanumeric, hyphens, underscores.

cloud string (required)

Cloud provider. One of `aws`, `azure`, `gcp`.

region string (required)

Cloud region to launch nodes in (e.g., `us-east-1`, `westus2`, `europe-west1`).

zone string

Availability zone within the region (e.g., `us-east-1a`). If omitted, the provider selects automatically.

min_nodes integer

Minimum number of warm nodes to keep running at all times.

default: 1

max_nodes integer

Maximum number of nodes the fleet can scale to.

default: 10

instance_type string

Specific instance type to use (e.g., `p4d.24xlarge`). If omitted, Seed selects based on resource requirements.

cpus string

CPU requirement. Use an exact value like `"4"` or a minimum like `"4+"`.

memory string

Memory requirement (e.g., `"16"` for 16 GB or `"16+"` for at least 16 GB).

gpus string

GPU requirement in the format `"GPU_TYPE:COUNT"` (e.g., `"A100:4"`, `"H100:8"`).

disk_size_gb integer

Disk size in gigabytes for each node.

disk_tier string

Disk performance tier. One of `standard`, `balanced`, `performance`.

default: standard

spot boolean

Use spot/preemptible instances. Significantly cheaper but may be interrupted.

default: false

idle_timeout_secs integer

Seconds a node can remain idle before being terminated (down to `min_nodes`).

default: 300

labels object

Key-value labels for fleet organization and job targeting.