POST /mltk/register_model

Description. Register a new SageMaker endpoint as an MLTK model. The credentials go to Splunk's password store (realm sagemaker); the model metadata becomes a __mlspl_{model_name}.mlmodel lookup. If lookup creation fails the credentials are rolled back.

Body (JSON).

{
  "model_name":   "my_sagemaker_model",
  "endpoint_name":"my-endpoint-123",
  "aws_credentials": {
    "region":            "us-east-1",
    "access_key_id":     "AKIA...",
    "secret_access_key": "...",
    "role_arn":          "arn:aws:iam::123456789012:role/SageMakerRole"
  },
  "input_feature_map":     { "cpu": "instances[*].cpu", "memory": "instances[*].memory" },
  "output_prediction_map": { "predictions[*]": "result" },
  "openapi_spec":          { /* OpenAPI 3.0 schema with /invocations POST */ },
  "batch_size": 1
}

batch_size. Integer 1–10000.

Validations.

• model_name is a valid Python identifier (no path traversal).
• Feature/output maps validated against the OpenAPI request/response schema. Empty maps allowed only when the OpenAPI spec uses text/csv content.
• AWS endpoint must exist and be InService.

Response (200).

{
  "success": true,
  "message": "Model 'my_sagemaker_model' registered successfully",
  "model_name":   "my_sagemaker_model",
  "endpoint_name":"my-endpoint-123",
  "warnings": []
}

Example.

curl -sk -u "$SPLUNK_USER:$SPLUNK_PASSWORD" -X POST 
  -H "Content-Type: application/json" 
  --data @register.json 
  "$SPLUNK_HOST/servicesNS/nobody/Splunk_ML_Toolkit/mltk/register_model?output_mode=json"