[ML] Trained Models: Hide Adaptive Allocations Toggle in Serverless (#224097)

Resolves https://github.com/elastic/kibana/issues/221894
Removes the adaptive allocations toggle in serverless (always enabled).
Adjusts the `min_allocations` param:
* Observability - min: 0
* Search - min: 0
* Security - min: 1

config/serverless.es.yml

@@ -85,18 +85,18 @@ xpack.ml.dfa.enabled: false
 xpack.ml.nlp:
   enabled: true
   modelDeployment:
-    allowStaticAllocations: true
+    allowStaticAllocations: false
     vCPURange:
       low:
         min: 0
         max: 2
         static: 2
       medium:
-        min: 1
+        min: 0
         max: 32
         static: 32
       high:
-        min: 1
+        min: 0
         max: 512
         static: 512
 xpack.ml.compatibleModuleType: 'search'

config/serverless.oblt.yml

@@ -119,10 +119,10 @@ xpack.ml.nlp:
         min: 0
         max: 2
       medium:
-        min: 1
+        min: 0
         max: 32
       high:
-        min: 1
+        min: 0
         max: 128
 xpack.ml.compatibleModuleType: 'observability'
 

x-pack/platform/plugins/shared/ml/public/application/model_management/deployment_params_mapper.test.ts

@@ -48,12 +48,12 @@ describe('DeploymentParamsMapper', () => {
           static: 16,
         });
         expect(mapper.getVCURange('medium')).toEqual({
-          min: 8,
+          min: 0,
           max: 256,
           static: 256,
         });
         expect(mapper.getVCURange('high')).toEqual({
-          min: 8,
+          min: 0,
           max: 4096,
           static: 4096,
         });
@@ -93,6 +93,48 @@ describe('DeploymentParamsMapper', () => {
             number_of_allocations: 2,
           },
         });
+
+        expect(
+          mapper.mapUiToApiDeploymentParams(modelId, {
+            deploymentId: 'test-deployment',
+            optimized: 'optimizedForSearch',
+            adaptiveResources: true,
+            vCPUUsage: 'medium',
+          })
+        ).toEqual({
+          modelId: 'test-model',
+          deploymentParams: {
+            deployment_id: 'test-deployment',
+            priority: 'normal',
+            threads_per_allocation: 16,
+          },
+          adaptiveAllocationsParams: {
+            enabled: true,
+            max_number_of_allocations: 2,
+            min_number_of_allocations: 0,
+          },
+        });
+
+        expect(
+          mapper.mapUiToApiDeploymentParams(modelId, {
+            deploymentId: 'test-deployment',
+            optimized: 'optimizedForIngest',
+            adaptiveResources: true,
+            vCPUUsage: 'high',
+          })
+        ).toEqual({
+          modelId: 'test-model',
+          deploymentParams: {
+            deployment_id: 'test-deployment',
+            priority: 'normal',
+            threads_per_allocation: 1,
+          },
+          adaptiveAllocationsParams: {
+            enabled: true,
+            max_number_of_allocations: 512,
+            min_number_of_allocations: 0,
+          },
+        });
       });
 
       it('overrides vCPUs levels and enforces adaptive allocations if static support is not configured', () => {

x-pack/platform/plugins/shared/ml/public/application/model_management/deployment_params_mapper.ts

@@ -56,8 +56,8 @@ export class DeploymentParamsMapper {
    */
   private readonly serverlessVCPUBreakpoints: VCPUBreakpoints = {
     low: { min: this.minAllowedNumberOfAllocation, max: 2, static: 2 },
-    medium: { min: 1, max: 32, static: 32 },
-    high: { min: 1, max: 512, static: 512 },
+    medium: { min: 0, max: 32, static: 32 },
+    high: { min: 0, max: 512, static: 512 },
   };
 
   /**
@@ -152,8 +152,11 @@ export class DeploymentParamsMapper {
       number_of_allocations: maxValue,
       min_number_of_allocations:
         Math.floor(levelValues.min / threadsPerAllocation) ||
-        // in any env, allow scale down to 0 only for "low" vCPU usage
-        (params.vCPUUsage === 'low' ? this.minAllowedNumberOfAllocation : 1),
+        // For serverless env, always allow scale down to 0
+        // For other envs, allow scale down to 0 only for "low" vCPU usage
+        (this.showNodeInfo === false || params.vCPUUsage === 'low'
+          ? this.minAllowedNumberOfAllocation
+          : 1),
       max_number_of_allocations: maxValue,
     };
   }

x-pack/platform/plugins/shared/ml/public/application/model_management/deployment_setup.tsx

@@ -818,7 +818,7 @@ export const StartUpdateDeploymentModal: FC<StartDeploymentModalProps> = ({
           errors={errors}
           isUpdate={isUpdate}
           disableAdaptiveResourcesControl={
-            showNodeInfo ? false : !nlpSettings.modelDeployment.allowStaticAllocations
+            !showNodeInfo || !nlpSettings.modelDeployment.allowStaticAllocations
           }
           deploymentsParams={
             isModelNotDownloaded || !isNLPModelItem(model)

x-pack/test_serverless/functional/test_suites/search/ml/trained_models_list.ts

@@ -60,15 +60,9 @@ export default function ({ getService, getPageObjects, getPageObject }: FtrProvi
           'Your model will scale up to a maximum of 4,096 VCUs per hour based on your search or ingest load. It will automatically scale down when demand decreases, and you only pay for the resources you use.'
         );
 
-        // Adaptive resources switch should be checked by default
-        await ml.trainedModelsTable.assertAdaptiveResourcesSwitchChecked(true);
-
-        // Static allocations should be allowed for search projects
-        await ml.trainedModelsTable.toggleAdaptiveResourcesSwitch(false);
-
-        await ml.trainedModelsTable.assertVCPUHelperText(
-          'Your model will consume 4,096 VCUs, even when not in use.'
-        );
+        // Adaptive resources switch should be hidden
+        // always use adaptive resources for serverless projects
+        await ml.trainedModelsTable.assertAdaptiveResourcesSwitchExists(false);
       });
     });
   });