Skip to main content
POST
create workergroup

Authorizations

Authorization
string
header
required

API key must be provided in the Authorization header

Body

application/json
search_params
string
required

Search filters for the offers that workers are recruited from, as a query string or JSON object. Combined with any search filters defined by the template; where both set the same filter, this value wins.

Example:

"gpu_name=RTX_3090 rentable=true"

endpoint_id
integer

ID of the endpoint this workergroup belongs to. Either endpoint_id or endpoint_name is required.

Example:

123

endpoint_name
string

Name of the endpoint this workergroup belongs to. Either endpoint_id or endpoint_name is required. If you have no endpoint with this name, one is created with default parameters.

Example:

"vLLM-Qwen3-8B"

template_hash
string

Hash ID of template to use for worker instances

Example:

"abc123def456"

template_id
integer

ID of template (alternative to template_hash)

Example:

456

launch_args
string

Additional launch arguments for worker instances

Example:

"--env VAR=value"

gpu_ram
number
default:8

Estimated GPU memory, in GB, that the model needs. Used to estimate worker load times; to filter offers by GPU memory, use search_params.

Example:

24

Response

Successfully created workergroup

success
boolean
Example:

true

id
integer

ID of created autoscaling job

Example:

789