fix: Deployments should be cached for non-model-information

packages/gen-ai-hub/src/utils/deployment-cache.test.ts

@@ -28,42 +28,65 @@ describe('deployment cache', () => {
     });
   });
 
-  it('should cache all deployments independent of potentially given models', () => {
-    const opts = {
-      scenarioId: 'foundation-models',
-      model: { name: 'gpt-4o', version: 'latest' }
-    };
+  describe('should cache all deployments independent of potentially given models', () => {
+    beforeEach(() => {
+      deploymentCache.setAll(
+        {
+          scenarioId: 'foundation-models',
+          model: { name: 'gpt-4o', version: 'latest' }
+        },
+        [
+          mockAiDeployment('deployment-id1', {
+            name: 'gpt-35-turbo',
+            version: 'latest'
+          }),
+          mockAiDeployment('deployment-id2', {
+            name: 'gpt-35-turbo',
+            version: '123'
+          })
+        ]
+      );
+    });
 
-    deploymentCache.setAll(opts, [
-      mockAiDeployment('deployment-id1', {
-        name: 'gpt-35-turbo',
-        version: 'latest'
-      }),
-      mockAiDeployment('deployment-id2', {
-        name: 'gpt-35-turbo',
-        version: '123'
-      })
-    ]);
+    it('cache nothing for unlisted model names', () => {
+      expect(
+        deploymentCache.get({
+          scenarioId: 'foundation-models',
+          model: { name: 'gpt-4o', version: 'latest' }
+        })
+      ).toBeUndefined();
+    });
 
-    expect(deploymentCache.get(opts)).toBeUndefined();
-    expect(
-      deploymentCache.get({
-        ...opts,
-        model: {
-          name: 'gpt-35-turbo',
-          version: 'latest'
-        }
-      })?.id
-    ).toEqual('deployment-id1');
-    expect(
-      deploymentCache.get({
-        ...opts,
-        model: {
-          name: 'gpt-35-turbo',
-          version: '123'
-        }
-      })?.id
-    ).toEqual('deployment-id2');
+    it('retrieve the deployment matching the model name and version', () => {
+      expect(
+        deploymentCache.get({
+          scenarioId: 'foundation-models',
+          model: {
+            name: 'gpt-35-turbo',
+            version: '123'
+          }
+        })?.id
+      ).toEqual('deployment-id2');
+    });
+
+    it('retrieve the first deployment matching the model name when version is missing', () => {
+      expect(
+        deploymentCache.get({
+          scenarioId: 'foundation-models',
+          model: {
+            name: 'gpt-35-turbo'
+          }
+        })?.id
+      ).toEqual('deployment-id1');
+    });
+
+    it('retrieve the deployment when model is missing', () => {
+      expect(
+        deploymentCache.get({
+          scenarioId: 'foundation-models'
+        })?.id
+      ).toEqual('deployment-id1');
+    });
   });
 
   it('should cache only the first deployments for equal models and versions', () => {

packages/gen-ai-hub/src/utils/deployment-cache.ts

@@ -9,7 +9,7 @@ function getCacheKey({
   model,
   resourceGroup = 'default'
 }: DeploymentResolutionOptions) {
-  return `${scenarioId}-${executableId}-${model?.name ?? ''}-${model?.version ?? ''}-${resourceGroup}`;
+  return `${scenarioId}:${executableId}:${model?.name ?? ''}:${model?.version ?? ''}:${resourceGroup}`;
 }
 
 interface Deployment {
@@ -51,15 +51,25 @@ function createDeploymentCache(cache: Cache<Deployment>) {
      * @param deployments - Deployments to retrieve the IDs and models from.
      */
     setAll: (
-      opts: Omit<DeploymentResolutionOptions, 'model'>,
+      opts: DeploymentResolutionOptions,
       deployments: AiDeployment[]
     ): void => {
       // go backwards to cache the first deployment ID for each model
-      [...deployments].reverse().forEach(deployment => {
-        cache.set(getCacheKey({ ...opts, model: extractModel(deployment) }), {
-          entry: transformDeploymentForCache(deployment)
+      [...deployments]
+        .reverse()
+        .map(deployment => transformDeploymentForCache(deployment))
+        .flatMap(entry => [
+          entry,
+          { id: entry.id },
+          ...(entry.model
+            ? [{ id: entry.id, model: { name: entry.model.name } }]
+            : [])
+        ])
+        .forEach(entry => {
+          cache.set(getCacheKey({ ...opts, model: entry.model }), {
+            entry
+          });
         });
-      });
     },
     clear: () => cache.clear()
   };

packages/gen-ai-hub/src/utils/deployment-resolver.test.ts

@@ -29,23 +29,23 @@ describe('deployment resolver', () => {
       const id = await resolveDeploymentId({
         scenarioId: 'foundation-models'
       });
-      expect(id).toBe('1');
+      expect(id).toEqual('1');
     });
 
     it('should return the first deployment with the correct model name', async () => {
       const id = await resolveDeploymentId({
         scenarioId: 'foundation-models',
         model: { name: 'gpt-4o' }
       });
-      expect(id).toBe('1');
+      expect(id).toEqual('1');
     });
 
     it('should return the deployment with the correct model name and version', async () => {
       const id = await resolveDeploymentId({
         scenarioId: 'foundation-models',
         model: { name: 'gpt-4o', version: '0613' }
       });
-      expect(id).toBe('2');
+      expect(id).toEqual('2');
     });
 
     it('should retrieve deployment from cache if available', async () => {
@@ -55,8 +55,8 @@ describe('deployment resolver', () => {
       };
       deploymentCache.set(opts, { id: '1' } as AiDeployment);
       const id = await resolveDeploymentId(opts);
-      expect(id).toBe('1');
-      expect(nock.isDone()).toBe(false);
+      expect(id).toEqual('1');
+      expect(nock.isDone()).toEqual(false);
     });
 
     it('should throw in case no deployment with the given model name is found', async () => {
@@ -131,7 +131,7 @@ describe('deployment resolver', () => {
       resourceGroup: 'otherId'
     });
 
-    expect(id).toBe('5');
+    expect(id).toEqual('5');
   });
 });