Sign in

v1

latestOpenAPI 3.1.02026-07-13897404.3 KB
Serverless

create workergroup

Creates a new workergroup configuration that manages worker instances for a serverless endpoint.

CLI Usage: vastai create workergroup --template_hash <hash> --endpoint_name <name> [options]

post/api/v0/workergroups/

Request body

endpoint_namestring

Name of the endpoint group

endpoint_idinteger

ID of existing endpoint group (alternative to endpoint_name)

template_hashstring

Hash ID of template to use for worker instances

template_idinteger

ID of template (alternative to template_hash)

search_paramsstring

Search query for finding worker instances (alternative to template)

launch_argsstring

Additional launch arguments for worker instances

min_loadnumber

Minimum load threshold for scaling

target_utilnumber

Target GPU utilization

cold_multnumber

Cold start multiplier

cold_workersinteger

Number of cold workers to maintain

max_workersinteger

Maximum number of worker instances

test_workersinteger

Number of test workers

gpu_raminteger

Minimum GPU RAM in GB

Example request

{
  "endpoint_name": "vLLM-Qwen3-8B",
  "endpoint_id": 123,
  "template_hash": "abc123def456",
  "template_id": 456,
  "search_params": "gpu_name=RTX_3090 rentable=true",
  "launch_args": "--env VAR=value",
  "min_load": 1,
  "target_util": 0.9,
  "cold_mult": 3,
  "cold_workers": 3,
  "max_workers": 20,
  "test_workers": 3,
  "gpu_ram": 24
}

Response

Successfully created workergroup

successboolean
idinteger

ID of created autoscaling job

Example response

{
  "success": true,
  "id": 789
}