Implement OpenShift AI integration for chat completion, embeddings, and reranking

server/src/main/resources/transport/definitions/referable/ml_inference_openshift_ai_added.csv

@@ -1 +1 @@
-9213000
+9216000

server/src/main/resources/transport/upper_bounds/9.3.csv

@@ -1 +1 @@
-ml_inference_openshift_ai_added,9213000
+ml_inference_openshift_ai_added,9216000

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/ServiceFields.java

@@ -14,6 +14,7 @@ public final class ServiceFields {
 
     public static final String SIMILARITY = "similarity";
     public static final String DIMENSIONS = "dimensions";
+    public static final String DIMENSIONS_SET_BY_USER = "dimensions_set_by_user";
     // Typically we use this to define the maximum tokens for the input text (text being sent to an integration)
     public static final String MAX_INPUT_TOKENS = "max_input_tokens";
     public static final String URL = "url";

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/openshiftai/OpenShiftAiServiceSettings.java

@@ -64,7 +64,7 @@ protected OpenShiftAiServiceSettings(StreamInput in) throws IOException {
      */
     protected OpenShiftAiServiceSettings(@Nullable String modelId, URI uri, @Nullable RateLimitSettings rateLimitSettings) {
         this.modelId = modelId;
-        this.uri = uri;
+        this.uri = Objects.requireNonNull(uri);
         this.rateLimitSettings = Objects.requireNonNullElse(rateLimitSettings, DEFAULT_RATE_LIMIT_SETTINGS);
     }
 

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/openshiftai/embeddings/OpenShiftAiEmbeddingsServiceSettings.java

@@ -27,6 +27,7 @@
 import java.util.Objects;
 
 import static org.elasticsearch.xpack.inference.services.ServiceFields.DIMENSIONS;
+import static org.elasticsearch.xpack.inference.services.ServiceFields.DIMENSIONS_SET_BY_USER;
 import static org.elasticsearch.xpack.inference.services.ServiceFields.MAX_INPUT_TOKENS;
 import static org.elasticsearch.xpack.inference.services.ServiceFields.SIMILARITY;
 import static org.elasticsearch.xpack.inference.services.ServiceUtils.createUri;
@@ -40,7 +41,6 @@
  */
 public class OpenShiftAiEmbeddingsServiceSettings extends OpenShiftAiServiceSettings {
     public static final String NAME = "openshift_ai_embeddings_service_settings";
-    static final String DIMENSIONS_SET_BY_USER = "dimensions_set_by_user";
 
     private final Integer dimensions;
     private final SimilarityMeasure similarity;
@@ -137,7 +137,7 @@ public OpenShiftAiEmbeddingsServiceSettings(
         this.dimensions = dimensions;
         this.similarity = similarity;
         this.maxInputTokens = maxInputTokens;
-        this.dimensionsSetByUser = dimensionsSetByUser;
+        this.dimensionsSetByUser = Objects.requireNonNull(dimensionsSetByUser);
     }
 
     /**

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/openshiftai/completion/OpenShiftAiChatCompletionModelTests.java

@@ -13,8 +13,15 @@
 import org.elasticsearch.xpack.inference.services.settings.DefaultSecretSettings;
 
 import static org.hamcrest.Matchers.is;
+import static org.hamcrest.Matchers.sameInstance;
 
 public class OpenShiftAiChatCompletionModelTests extends ESTestCase {
+
+    private static final String MODEL_VALUE = "model_name";
+    private static final String API_KEY_VALUE = "test_api_key";
+    private static final String URL_VALUE = "http://www.abc.com";
+    private static final String ALTERNATE_MODEL_VALUE = "different_model";
+
     public static OpenShiftAiChatCompletionModel createCompletionModel(String url, String apiKey, String modelName) {
         return createModelWithTaskType(url, apiKey, modelName, TaskType.COMPLETION);
     }
@@ -34,23 +41,30 @@ public static OpenShiftAiChatCompletionModel createModelWithTaskType(String url,
     }
 
     public void testOverrideWith_UnifiedCompletionRequest_KeepsSameModelId() {
-        var model = createCompletionModel("url", "api_key", "model_name");
-        var overriddenModel = OpenShiftAiChatCompletionModel.of(model, "model_name");
+        var model = createCompletionModel(URL_VALUE, API_KEY_VALUE, MODEL_VALUE);
+        var overriddenModel = OpenShiftAiChatCompletionModel.of(model, MODEL_VALUE);
 
-        assertThat(overriddenModel.getServiceSettings().modelId(), is("model_name"));
+        assertThat(overriddenModel, is(sameInstance(model)));
     }
 
     public void testOverrideWith_UnifiedCompletionRequest_OverridesExistingModelId() {
-        var model = createCompletionModel("url", "api_key", "model_name");
-        var overriddenModel = OpenShiftAiChatCompletionModel.of(model, "different_model");
+        var model = createCompletionModel(URL_VALUE, API_KEY_VALUE, MODEL_VALUE);
+        var overriddenModel = OpenShiftAiChatCompletionModel.of(model, ALTERNATE_MODEL_VALUE);
 
-        assertThat(overriddenModel.getServiceSettings().modelId(), is("different_model"));
+        assertThat(overriddenModel.getServiceSettings().modelId(), is(ALTERNATE_MODEL_VALUE));
     }
 
     public void testOverrideWith_UnifiedCompletionRequest_UsesModelFields_WhenRequestDoesNotOverride() {
-        var model = createCompletionModel("url", "api_key", "model_name");
+        var model = createCompletionModel(URL_VALUE, API_KEY_VALUE, MODEL_VALUE);
+        var overriddenModel = OpenShiftAiChatCompletionModel.of(model, null);
+
+        assertThat(overriddenModel, is(sameInstance(model)));
+    }
+
+    public void testOverrideWith_UnifiedCompletionRequest_KeepsNullIfNoModelIdProvided() {
+        var model = createCompletionModel(URL_VALUE, API_KEY_VALUE, null);
         var overriddenModel = OpenShiftAiChatCompletionModel.of(model, null);
 
-        assertThat(overriddenModel.getServiceSettings().modelId(), is("model_name"));
+        assertThat(overriddenModel, is(sameInstance(model)));
     }
 }

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/openshiftai/completion/OpenShiftAiChatCompletionResponseHandlerTests.java

@@ -31,6 +31,8 @@
 import static org.mockito.Mockito.when;
 
 public class OpenShiftAiChatCompletionResponseHandlerTests extends ESTestCase {
+    private static final String URL_VALUE = "http://www.abc.com";
+    private static final String INFERENCE_ID = "id";
     private final OpenShiftAiChatCompletionResponseHandler responseHandler = new OpenShiftAiChatCompletionResponseHandler(
         "chat completions",
         (a, b) -> mock()
@@ -49,19 +51,20 @@ public void testFailNotFound() throws IOException {
             {
               "error" : {
                 "code" : "not_found",
-                "message" : "Resource not found at [https://api.llama.ai/v1/chat/completions] for request from inference entity id [id] \
+                "message" : "Resource not found at [%s] for request from inference entity id [%s] \
             status [404]. Error message: [{\\"detail\\":\\"Not Found\\"}]",
                 "type" : "openshift_ai_error"
               }
-            }""")));
+            }""".formatted(URL_VALUE, INFERENCE_ID))));
     }
 
     public void testFailBadRequest() throws IOException {
         var responseJson = XContentHelper.stripWhitespace("""
             {
                 "object": "error",
-                "message": "[{'type': 'missing', 'loc': ('body', 'messages'), 'msg': 'Field required', 'input': {'model': 'llama-31-8b-ins\
-                truct', '1messages': [{'role': 'user', 'content': 'What is deep learning?'}], 'max_tokens': 2, 'stream': True}}]",
+                "message": "[{'type': 'missing', 'loc': ('body', 'messages'), 'msg': 'Field required', \
+            'input': {'model': 'llama-31-8b-instruct', 'messages': [{'role': 'user', 'content': 'What is deep learning?'}], \
+            'max_tokens': 2, 'stream': True}}]",
                 "type": "Bad Request",
                 "param": null,
                 "code": 400
@@ -74,14 +77,14 @@ public void testFailBadRequest() throws IOException {
             {
                 "error": {
                     "code": "bad_request",
-                    "message": "Received a bad request status code for request from inference entity id [id] status [400].\
-             Error message: [{\\"object\\":\\"error\\",\\"message\\":\\"[{'type': 'missing', 'loc': ('body', 'messages'), 'msg': 'Field r\
-            equired', 'input': {'model': 'llama-31-8b-ins    truct', '1messages': [{'role': 'user', 'content': 'What is deep learning?'}]\
-            , 'max_tokens': 2, 'stream': True}}]\\",\\"type\\":\\"Bad Request\\",\\"param\\":null,\\"code\\":400}]",
+                    "message": "Received a bad request status code for request from inference entity id [%s] status [400]. Error message: \
+            [{\\"object\\":\\"error\\",\\"message\\":\\"[{'type': 'missing', 'loc': ('body', 'messages'), 'msg': 'Field required', \
+            'input': {'model': 'llama-31-8b-instruct', 'messages': [{'role': 'user', 'content': 'What is deep learning?'}], \
+            'max_tokens': 2, 'stream': True}}]\\",\\"type\\":\\"Bad Request\\",\\"param\\":null,\\"code\\":400}]",
                     "type": "openshift_ai_error"
                 }
             }
-            """)));
+            """.formatted(INFERENCE_ID))));
     }
 
     public void testFailValidationWithInvalidJson() throws IOException {
@@ -95,12 +98,12 @@ public void testFailValidationWithInvalidJson() throws IOException {
             {
                 "error": {
                     "code": "bad_request",
-                    "message": "Received a server error status code for request from inference entity id [id] status [500]. Error message: \
-            [what? this isn't a json\\n]",
+                    "message": "Received a server error status code for request from inference entity id [%s] status [500]. \
+            Error message: [what? this isn't a json\\n]",
                     "type": "openshift_ai_error"
                 }
             }
-            """)));
+            """.formatted(INFERENCE_ID))));
     }
 
     private String invalidResponseJson(String responseJson, int statusCode) throws IOException {
@@ -124,9 +127,9 @@ private Exception invalidResponse(String responseJson, int statusCode) {
 
     private static Request mockRequest() throws URISyntaxException {
         var request = mock(Request.class);
-        when(request.getInferenceEntityId()).thenReturn("id");
+        when(request.getInferenceEntityId()).thenReturn(INFERENCE_ID);
         when(request.isStreaming()).thenReturn(true);
-        when(request.getURI()).thenReturn(new URI("https://api.llama.ai/v1/chat/completions"));
+        when(request.getURI()).thenReturn(new URI(URL_VALUE));
         return request;
     }
 

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/openshiftai/completion/OpenShiftAiChatCompletionServiceSettingsTests.java

@@ -33,18 +33,19 @@
 public class OpenShiftAiChatCompletionServiceSettingsTests extends AbstractBWCWireSerializationTestCase<
     OpenShiftAiChatCompletionServiceSettings> {
 
-    public static final String MODEL_ID = "some model";
-    public static final String CORRECT_URL = "https://www.elastic.co";
-    public static final int RATE_LIMIT = 2;
+    private static final String MODEL_VALUE = "some_model";
+    private static final String URL_VALUE = "http://www.abc.com";
+    private static final String INVALID_URL_VALUE = "^^^";
+    private static final int RATE_LIMIT = 2;
 
     public void testFromMap_AllFields_Success() {
         var serviceSettings = OpenShiftAiChatCompletionServiceSettings.fromMap(
             new HashMap<>(
                 Map.of(
                     ServiceFields.MODEL_ID,
-                    MODEL_ID,
+                    MODEL_VALUE,
                     ServiceFields.URL,
-                    CORRECT_URL,
+                    URL_VALUE,
                     RateLimitSettings.FIELD_NAME,
                     new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
                 )
@@ -54,7 +55,7 @@ public void testFromMap_AllFields_Success() {
 
         assertThat(
             serviceSettings,
-            is(new OpenShiftAiChatCompletionServiceSettings(MODEL_ID, CORRECT_URL, new RateLimitSettings(RATE_LIMIT)))
+            is(new OpenShiftAiChatCompletionServiceSettings(MODEL_VALUE, URL_VALUE, new RateLimitSettings(RATE_LIMIT)))
         );
     }
 
@@ -63,56 +64,85 @@ public void testFromMap_MissingModelId_Success() {
             new HashMap<>(
                 Map.of(
                     ServiceFields.URL,
-                    CORRECT_URL,
+                    URL_VALUE,
                     RateLimitSettings.FIELD_NAME,
                     new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
                 )
             ),
             ConfigurationParseContext.PERSISTENT
         );
 
-        assertThat(serviceSettings, is(new OpenShiftAiChatCompletionServiceSettings(null, CORRECT_URL, new RateLimitSettings(RATE_LIMIT))));
+        assertThat(serviceSettings, is(new OpenShiftAiChatCompletionServiceSettings(null, URL_VALUE, new RateLimitSettings(RATE_LIMIT))));
     }
 
     public void testFromMap_MissingUrl_ThrowsException() {
+        testFromMap_InvalidUrl(
+            Map.of(
+                ServiceFields.MODEL_ID,
+                MODEL_VALUE,
+                RateLimitSettings.FIELD_NAME,
+                new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
+            ),
+            "Validation Failed: 1: [service_settings] does not contain the required setting [url];"
+        );
+    }
+
+    public void testFromMap_InvalidUrl_ThrowsException() {
+        testFromMap_InvalidUrl(
+            Map.of(
+                ServiceFields.URL,
+                INVALID_URL_VALUE,
+                ServiceFields.MODEL_ID,
+                MODEL_VALUE,
+                RateLimitSettings.FIELD_NAME,
+                new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
+            ),
+            """
+                Validation Failed: 1: [service_settings] Invalid url [^^^] received for field [url]. \
+                Error: unable to parse url [^^^]. Reason: Illegal character in path;"""
+        );
+    }
+
+    public void testFromMap_EmptyUrl_ThrowsException() {
+        testFromMap_InvalidUrl(
+            Map.of(
+                ServiceFields.URL,
+                "",
+                ServiceFields.MODEL_ID,
+                MODEL_VALUE,
+                RateLimitSettings.FIELD_NAME,
+                new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
+            ),
+            "Validation Failed: 1: [service_settings] Invalid value empty string. [url] must be a non-empty string;"
+        );
+    }
+
+    private static void testFromMap_InvalidUrl(Map<String, Object> serviceSettingsMap, String expectedErrorMessage) {
         var thrownException = expectThrows(
             ValidationException.class,
-            () -> OpenShiftAiChatCompletionServiceSettings.fromMap(
-                new HashMap<>(
-                    Map.of(
-                        ServiceFields.MODEL_ID,
-                        MODEL_ID,
-                        RateLimitSettings.FIELD_NAME,
-                        new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
-                    )
-                ),
-                ConfigurationParseContext.PERSISTENT
-            )
+            () -> OpenShiftAiChatCompletionServiceSettings.fromMap(new HashMap<>(serviceSettingsMap), ConfigurationParseContext.PERSISTENT)
         );
 
-        assertThat(
-            thrownException.getMessage(),
-            containsString("Validation Failed: 1: [service_settings] does not contain the required setting [url];")
-        );
+        assertThat(thrownException.getMessage(), containsString(expectedErrorMessage));
     }
 
     public void testFromMap_MissingRateLimit_Success() {
         var serviceSettings = OpenShiftAiChatCompletionServiceSettings.fromMap(
-            new HashMap<>(Map.of(ServiceFields.MODEL_ID, MODEL_ID, ServiceFields.URL, CORRECT_URL)),
+            new HashMap<>(Map.of(ServiceFields.MODEL_ID, MODEL_VALUE, ServiceFields.URL, URL_VALUE)),
             ConfigurationParseContext.PERSISTENT
         );
 
-        assertThat(serviceSettings, is(new OpenShiftAiChatCompletionServiceSettings(MODEL_ID, CORRECT_URL, null)));
+        assertThat(serviceSettings, is(new OpenShiftAiChatCompletionServiceSettings(MODEL_VALUE, URL_VALUE, null)));
     }
 
     public void testToXContent_WritesAllValues() throws IOException {
         var serviceSettings = OpenShiftAiChatCompletionServiceSettings.fromMap(
             new HashMap<>(
                 Map.of(
                     ServiceFields.MODEL_ID,
-                    MODEL_ID,
+                    MODEL_VALUE,
                     ServiceFields.URL,
-                    CORRECT_URL,
+                    URL_VALUE,
                     RateLimitSettings.FIELD_NAME,
                     new HashMap<>(Map.of(RateLimitSettings.REQUESTS_PER_MINUTE_FIELD, RATE_LIMIT))
                 )
@@ -125,20 +155,20 @@ public void testToXContent_WritesAllValues() throws IOException {
         String xContentResult = Strings.toString(builder);
         var expected = XContentHelper.stripWhitespace("""
             {
-                "model_id": "some model",
-                "url": "https://www.elastic.co",
+                "model_id": "%s",
+                "url": "%s",
                 "rate_limit": {
                     "requests_per_minute": 2
                 }
             }
-            """);
+            """.formatted(MODEL_VALUE, URL_VALUE));
 
         assertThat(xContentResult, is(expected));
     }
 
     public void testToXContent_DoesNotWriteOptionalValues_DefaultRateLimit() throws IOException {
         var serviceSettings = OpenShiftAiChatCompletionServiceSettings.fromMap(
-            new HashMap<>(Map.of(ServiceFields.URL, CORRECT_URL)),
+            new HashMap<>(Map.of(ServiceFields.URL, URL_VALUE)),
             ConfigurationParseContext.PERSISTENT
         );
 
@@ -147,12 +177,12 @@ public void testToXContent_DoesNotWriteOptionalValues_DefaultRateLimit() throws
         String xContentResult = Strings.toString(builder);
         var expected = XContentHelper.stripWhitespace("""
             {
-                "url": "https://www.elastic.co",
+                "url": "%s",
                 "rate_limit": {
                     "requests_per_minute": 3000
                 }
             }
-            """);
+            """.formatted(URL_VALUE));
         assertThat(xContentResult, is(expected));
     }