Create a calculation instance

Create a compute instance in the current workspace.

POST https://moi.matrixorigin.cn/newmoi/compute-resources

Preparation before calling

First list the computing specifications to obtain the available spec_id; when you need to specify the image, then list the Worker image]. Prepare a personal access token and target workspace ID that has access to the target workspace.

The example below uses:

• $AI_STUDIO_API_KEY: The actual personal access token, passed through the X-API-Key Header.

• $WORKSPACE_ID: Workspace ID to create the compute instance, passed through the X-Workspace-ID Header.

• $SPEC_ID: The compute spec ID to use, taken from the response of List compute specs, specifying the instance spec in the request body.

Request example

curl -X POST "https://moi.matrixorigin.cn/newmoi/compute-resources" \
  -H "X-API-Key: $AI_STUDIO_API_KEY" \
  -H "X-Workspace-ID: $WORKSPACE_ID" \
  -H 'Content-Type: application/json' \
  -d '{
    "name": "data-processing",
    "spec_id": "'"$SPEC_ID"'",
    "min_replicas": 0,
    "max_replicas": 1,
    "worker_images": [
      {
        "worker_type": "go-worker",
        "image_id": "$GO_WORKER_IMAGE_ID",
        "platform": "linux/amd64"
      },
      {
        "worker_type": "python-worker",
        "image_id": "$PYTHON_WORKER_IMAGE_ID",
        "platform": "linux/amd64"
      },
      {
        "worker_type": "java-worker",
        "image_id": "$JAVA_WORKER_IMAGE_ID",
        "platform": "linux/amd64"
      },
      {
        "worker_type": "custom-tool-worker",
        "image_id": "$CUSTOM_TOOL_WORKER_IMAGE_ID",
        "platform": "linux/amd64"
      }
    ]
  }'

Request body

Field	Type	Required	Description
name	string	Yes	Compute instance name.
spec_id	string	Yes	Calculated specification ID.
kind	string	No	Compatibility prompt; the instance type shall be subject to specifications.
description	string	No	Usage description.
cpu	integer	No	Compatible input; the actual capacity is determined by the corresponding specifications of spec_id.
memory_gib	integer	No	Compatible input; the actual capacity is determined by the corresponding specifications of spec_id.
gpu	integer	No	Compatible input; the actual capacity is determined by the corresponding specifications of spec_id.
min_replicas	integer	No	Serverless instances only accept 0, or 0 when omitted.
max_replicas	integer	No	The upper limit of copies, must be greater than 0; when omitted, it is 1.
auto_suspend_minutes	integer	No	Automatic pause waiting time.
go_worker_image_id	string	No	Compatible Go Worker image ID.
python_worker_image_id	string	No	Compatible Python Worker image ID.
platform	string	No	Compatible operating platform.
worker_images	object[]	Yes	Worker image selection, must cover all currently active worker_type in the image list.
worker_images[].worker_type	string	Yes	Worker type, taken from worker_type in the mirror list.
worker_images[].image_id	string	Yes	Image ID, taken from id in the image list.
worker_images[].platform	string	Yes	Image running platform, taken from platform in the image list.
worker_images[].kind	string	No	Compatible alias; should be consistent with worker_type when passed in.

In this document, [] after a type means an array; for example, object[] is an array of objects. In field paths, [] means each item in an array; for example, worker_images[].worker_type is the worker_type field of each item in worker_images.

worker_type, image_id and platform in worker_images should come from the mirror list. Each creation requires re-reading the mirror list and submitting one entry for each active Worker type; you cannot submit just one of them or rely on server-side auto-completion.

Successful response

Returns 200 on success. Save data.id. The new Serverless instance is initially IDLE, and the Worker will not be deployed immediately; subsequent workload requirements will activate the required Worker type.

{
  "code": "OK",
  "msg": "OK",
  "data": {
    "id": "resource-123",
    "name": "data-processing",
    "spec_id": "task-standard-small",
    "status": "IDLE",
    "status_message": "",
    "min_replicas": 0,
    "max_replicas": 1,
    "current_replicas": 0
  }
}

The response fields are as follows.

Field	Type	Description
code	string	OK on success.
msg	string	OK on success.
data.id	string	ID of the new computing instance.
data.workspace_id	string	ID of the workspace to which it belongs.
data.name	string	Compute instance name.
data.description	string	Example description; returned when there is a value.
data.spec_id	string	The compute specification ID to use.
data.kind	string	Instance type.
data.cpu	integer	CPU configuration value.
data.memory_gib	integer	Configuration value expressed in GiB of memory.
data.gpu	integer	GPU configuration value.
data.cpu_milli	integer	CPU configuration value in millicores.
data.memory_mib	integer	Configuration value expressed in MiB of memory.
data.gpu_count	integer	Number of GPUs.
data.gpu_memory_mib	integer	MiB of video memory for a single GPU.
data.gpu_cores	integer	Number of GPU cores.
data.min_replicas	integer	Serverless instances are fixed to 0.
data.max_replicas	integer	Maximum number of copies.
data.desired_replicas	integer	Number of target copies.
data.current_replicas	integer	The current number of copies.
data.go_worker_image_id	string	Go Worker image ID; returned if there is a value.
data.python_worker_image_id	string	Python Worker image ID; returned if there is a value.
data.worker_images	object[]	Worker image selection; returned when there is a value.
data.worker_images[].worker_type	string	Worker type.
data.worker_images[].image_id	string	Image ID.
data.worker_images[].platform	string	Running platform.
data.platform	string	Instance running platform; returned if there is a value.
data.auto_suspend_minutes	integer	Automatic pause waiting time.
data.is_default	boolean	Whether it is the default computing instance of the workspace.
data.status	string	Current life cycle status.
data.status_message	string	Supplementary status information; returned if there is a value.
data.scale_reason	string	The scaling reason provided by the server; returned if there is a value.
data.last_activation_at	string	Last activation time; returned if there is a value.
data.last_active_at	string	Last active time; returned if there is a value.
data.created_by	string	Creator ID; returned if there is a value.
data.created_at	string	Creation time; returned if there is a value.
data.updated_at	string	Last updated time; returned if there is a value.

Error response

{
  "code": "INVALID_PARAMS",
  "msg": "请求参数无效",
  "data": null
}

Common HTTP errors

HTTP status code	error code	Common causes	Recommended actions
400	INVALID_PARAMS	The request body cannot be parsed, or the required name and spec_id are missing.	Check request fields and calculation specifications.
401	—	Missing or invalid access credentials.	Check API Key and Workspace Header.
403	ErrForbidden	The current identity does not have permission to create instances.	Use valid credentials, or contact your administrator for authorization.
409	COMPUTE_RESOURCE_NAME_EXISTS	An instance with the same name already exists in the workspace.	Replace name and try again.
500	CREATE_FAILED	The server failed to create an instance.	Record the request time and error message and try again.

Follow-up operations

Query calculation instance details until the status meets the task requirements.

Last updated on Sep 3, 2026