Presets

Endpoint presets are reusable configurations that bundle compute, replica counts, and runtime selections. Instead of filling out the full endpoint form every time, pick a preset and the deployment form pre-fills around it.

The presets list

The list view shows every endpoint preset in your workspace. Each row displays:

• Name — the preset's identifier.
• Description — what this preset is for.
• CPU / Memory / GPU — the compute shape.
• Min / Max replicas — autoscaling bounds.
• Runtimes — which inference runtimes this preset supports.

Click a preset to see its full configuration, then click Use this preset to open the endpoint deployment form with the preset's values pre-filled.

Creating a preset

Click Create preset to define a new endpoint configuration. You'll specify:

Field	Required	What it does
Name	required	Unique identifier for the preset.
Description	optional	What this preset is for.
CPU	required	CPU allocation (e.g. 2, 4000m).
Memory	required	Memory allocation (e.g. 8Gi).
GPU count	optional	Number of GPUs. Zero for CPU-only.
Compute pool	optional	Target compute pool.
Min replicas	required	Minimum pod count. Set to 0 for scale-to-zero.
Max replicas	required	Maximum pod count.
Runtimes	optional	Which inference runtimes this preset supports.
Configurations	optional	Pod defaults to attach (env vars, secrets, init scripts).

Next steps

• Deploying an endpoint — how to use a preset when deploying
• Runtimes — the serving environments presets reference