fix: [Core] Deployment resolution selects running targets

core/src/main/java/com/sap/ai/sdk/core/DeploymentResolver.java

@@ -2,13 +2,15 @@
 
 import com.sap.ai.sdk.core.client.DeploymentApi;
 import com.sap.ai.sdk.core.model.AiDeployment;
+import com.sap.ai.sdk.core.model.AiDeploymentStatus;
 import java.util.HashSet;
 import java.util.Map;
 import java.util.Objects;
 import java.util.Optional;
 import java.util.Set;
 import java.util.concurrent.ConcurrentHashMap;
 import java.util.function.Predicate;
+import java.util.stream.Collectors;
 import javax.annotation.Nonnull;
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
@@ -27,15 +29,15 @@ class DeploymentResolver {
 
   @Nonnull private final AiCoreService service;
 
-  /** Cache for deployment ids. The key is the model name and the value is the deployment id. */
+  /** Cache for deployments. The key is the resource group and the value a set of deployments. */
   @Nonnull private final Map<String, Set<AiDeployment>> cache;
 
   DeploymentResolver(@Nonnull final AiCoreService service) {
     this(service, DEFAULT_GLOBAL_CACHE);
   }
 
   /**
-   * Remove all entries from the cache then load all deployments into the cache.
+   * Remove cached deployments for the resource group and reload running deployments into the cache.
    *
    * <p><b>Call this whenever a deployment is deleted.</b>
    *
@@ -46,8 +48,16 @@ void reloadDeployments(@Nonnull final String resourceGroup) {
     try {
       val apiClient = new DeploymentApi(service);
       val deployments = new HashSet<>(apiClient.query(resourceGroup).getResources());
-      log.info("Found {} deployments in resource group '{}'.", deployments.size(), resourceGroup);
-      cache.put(resourceGroup, deployments);
+      val runningDeployments =
+          deployments.stream()
+              .filter(deployment -> AiDeploymentStatus.RUNNING.equals(deployment.getStatus()))
+              .collect(Collectors.toSet());
+      log.info(
+          "Found {} of {} deployments running in resource group '{}'.",
+          runningDeployments.size(),
+          deployments.size(),
+          resourceGroup);
+      cache.put(resourceGroup, runningDeployments);
     } catch (final RuntimeException e) {
       throw new DeploymentResolutionException(
           "Failed to load deployments for resource group " + resourceGroup, e);

core/src/test/java/com/sap/ai/sdk/core/AiCoreServiceTest.java

@@ -81,7 +81,8 @@ void testGetInferenceDestination() {
     assertThat(destination.getHeaders()).containsExactly(new Header("AI-Resource-Group", "foo"));
 
     // scenario-based destination
-    val d = "{\"count\":1,\"resources\":[{\"id\":\"0123456789abcdef\",\"scenarioId\":\"foobar\"}]}";
+    val d =
+        "{\"count\":1,\"resources\":[{\"id\":\"0123456789abcdef\",\"scenarioId\":\"foobar\", \"status\":\"RUNNING\"}]}";
     val server = new WireMockServer(wireMockConfig().dynamicPort());
     server.start();
     server.stubFor(get(urlEqualTo("/v2/lm/deployments")).willReturn(okJson(d)));

core/src/test/java/com/sap/ai/sdk/core/DeploymentResolverTest.java

@@ -10,6 +10,7 @@
 import static com.sap.ai.sdk.core.AiCoreService.DEFAULT_RESOURCE_GROUP;
 import static org.assertj.core.api.Assertions.assertThat;
 import static org.assertj.core.api.Assertions.assertThatThrownBy;
+import static org.assertj.core.api.Assertions.tuple;
 
 import com.sap.ai.sdk.core.client.WireMockTestServer;
 import com.sap.ai.sdk.core.model.AiDeployment;
@@ -169,6 +170,24 @@ void testIsDeploymentOfModel() {
     assertThat(DeploymentResolver.isDeploymentOfModel(gpt4VersionLatest, deployment)).isTrue();
   }
 
+  @Test
+  void testCacheContainsRunningDeployments() {
+    wireMockServer.stubFor(
+        get(anyUrl())
+            .willReturn(
+                aResponse()
+                    .withBodyFile("hasStoppedDeployment.json")
+                    .withHeader("content-type", "application/json")));
+
+    final var deployment =
+        resolver.getDeploymentIdByScenario(DEFAULT_RESOURCE_GROUP, "orchestration");
+
+    assertThat(deployment).isEqualTo("d4b1396b84c1944d");
+    assertThat(cache.get(DEFAULT_RESOURCE_GROUP))
+        .extracting(AiDeployment::getId, AiDeployment::getStatus)
+        .containsExactly(tuple("d4b1396b84c1944d", AiDeploymentStatus.RUNNING));
+  }
+
   private record TestModel(String name, String version) implements AiModel {}
 
   private static void stubResponse(String resourceGroup, String fileName) {

core/src/test/resources/__files/hasStoppedDeployment.json

@@ -0,0 +1,54 @@
+{
+  "count": 8,
+  "resources": [
+    {
+      "id": "d2a491b5010620b0",
+      "createdAt": "2026-06-08T14:56:25Z",
+      "modifiedAt": "2026-06-08T15:10:18Z",
+      "status": "STOPPED",
+      "details": {
+        "scaling": {
+          "backendDetails": {},
+          "backend_details": {}
+        },
+        "resources": {
+          "backendDetails": {},
+          "backend_details": {}
+        }
+      },
+      "scenarioId": "orchestration",
+      "configurationId": "0f0fb4c1-cf2c-441f-98d1-1f983d1b1756",
+      "targetStatus": "STOPPED",
+      "submissionTime": "2026-06-08T15:03:35Z",
+      "startTime": "2026-06-08T15:05:27Z",
+      "completionTime": "2026-06-08T15:27:23Z",
+      "configurationName": "orchestration-config",
+      "deploymentUrl": "https://api.ai.intprod-eu12.eu-central-1.aws.ml.hana.ondemand.com/v2/inference/deployments/d2a491b5010620b0"
+    },
+    {
+      "id": "d4b1396b84c1944d",
+      "createdAt": "2026-04-16T10:03:39Z",
+      "modifiedAt": "2026-04-16T10:03:39Z",
+      "status": "RUNNING",
+      "details": {
+        "scaling": {
+          "backendDetails": {},
+          "backend_details": {}
+        },
+        "resources": {
+          "backendDetails": {},
+          "backend_details": {}
+        }
+      },
+      "scenarioId": "orchestration",
+      "configurationId": "e88f7dba-fd9c-4068-ab18-3306f2aa46bd",
+      "latestRunningConfigurationId": "e88f7dba-fd9c-4068-ab18-3306f2aa46bd",
+      "lastOperation": "CREATE",
+      "targetStatus": "RUNNING",
+      "submissionTime": "2026-04-17T08:32:01Z",
+      "startTime": "2026-04-17T08:32:59Z",
+      "configurationName": "orchestration-config",
+      "deploymentUrl": "https://api.ai.intprod-eu12.eu-central-1.aws.ml.hana.ondemand.com/v2/inference/deployments/d4b1396b84c1944d"
+    }
+  ]
+}