Matthias' review

Author: CharlesDuboisSAP

docs/guides/ORCHESTRATION_CHAT_COMPLETION.md

@@ -214,14 +214,16 @@ In this example, the input will be masked before the call to the LLM. Note that
 
 ### Set model parameters
 
-Change your LLM module configuration to add model parameters:
+Change your LLM configuration to add model parameters:
 
 ```java
-LLMModuleConfig llmConfig =
-    OrchestrationAiModel.GPT_4O.modelParams(
-        Map.of(
-            "max_tokens", 50,
-            "temperature", 0.1,
-            "frequency_penalty", 0,
-            "presence_penalty", 0));
+OrchestrationAiModel customGPT4O =
+    OrchestrationAiModel.GPT_4O
+        .withModelParams(
+            Map.of(
+                "max_tokens", 50,
+                "temperature", 0.1,
+                "frequency_penalty", 0,
+                "presence_penalty", 0))
+        .withModelVersion("2024-05-13");
 ```

orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationAiModel.java

@@ -1,18 +1,36 @@
 package com.sap.ai.sdk.orchestration;
 
+import com.sap.ai.sdk.core.AiModel;
 import com.sap.ai.sdk.orchestration.client.model.LLMModuleConfig;
 import java.util.Map;
 import javax.annotation.Nonnull;
+import javax.annotation.Nullable;
 import lombok.AccessLevel;
-import lombok.Getter;
-import lombok.RequiredArgsConstructor;
+import lombok.AllArgsConstructor;
+import lombok.With;
 
 /** Large language models available in Orchestration. */
-// https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/models-and-scenarios-in-generative-ai-hub
-@Getter
-@RequiredArgsConstructor(access = AccessLevel.PRIVATE)
-public class OrchestrationAiModel {
-  private final LLMModuleConfig config;
+@With
+@AllArgsConstructor(access = AccessLevel.PRIVATE)
+public class OrchestrationAiModel implements AiModel {
+  /** The name of the model */
+  private String modelName;
+
+  /** The version of the model, defaults to "latest". */
+  private String modelVersion = "latest";
+
+  /**
+   * Optional parameters on this model.
+   *
+   * <pre>{@code
+   * Map.of(
+   *     "max_tokens", 50,
+   *     "temperature", 0.1,
+   *     "frequency_penalty", 0,
+   *     "presence_penalty", 0)
+   * }</pre>
+   */
+  private Map<String, Object> modelParams;
 
   /** IBM Granite 13B chat completions model */
   public static final OrchestrationAiModel IBM_GRANITE_13B_CHAT =
@@ -50,10 +68,6 @@ public class OrchestrationAiModel {
   public static final OrchestrationAiModel CLAUDE_3_5_SONNET =
       new OrchestrationAiModel("anthropic--claude-3.5-sonnet");
 
-  /** Amazon Titan Embed Text model */
-  public static final OrchestrationAiModel TITAN_EMBED_TEXT =
-      new OrchestrationAiModel("amazon--titan-embed-text");
-
   /** Amazon Titan Text Lite model */
   public static final OrchestrationAiModel TITAN_TEXT_LITE =
       new OrchestrationAiModel("amazon--titan-text-lite");
@@ -75,38 +89,12 @@ public class OrchestrationAiModel {
   /** Azure OpenAI GPT-4-32k chat completions model */
   public static final OrchestrationAiModel GPT_4_32K = new OrchestrationAiModel("gpt-4-32k");
 
-  /** Azure OpenAI Text Embedding Ada 002 model */
-  public static final OrchestrationAiModel TEXT_EMBEDDING_ADA_002 =
-      new OrchestrationAiModel("text-embedding-ada-002");
-
-  /** Azure OpenAI Text Embedding 3 Small model */
-  public static final OrchestrationAiModel TEXT_EMBEDDING_3_SMALL =
-      new OrchestrationAiModel("text-embedding-3-small");
-
-  /** Azure OpenAI Text Embedding 3 Large model */
-  public static final OrchestrationAiModel TEXT_EMBEDDING_3_LARGE =
-      new OrchestrationAiModel("text-embedding-3-large");
-
   /** Azure OpenAI GPT-4o chat completions model */
   public static final OrchestrationAiModel GPT_4O = new OrchestrationAiModel("gpt-4o");
 
   /** Azure OpenAI GPT-4o-mini chat completions model */
   public static final OrchestrationAiModel GPT_4O_MINI = new OrchestrationAiModel("gpt-4o-mini");
 
-  /** Google Cloud Platform Text Bison model */
-  public static final OrchestrationAiModel TEXT_BISON = new OrchestrationAiModel("text-bison");
-
-  /** Google Cloud Platform Chat Bison model */
-  public static final OrchestrationAiModel CHAT_BISON = new OrchestrationAiModel("chat-bison");
-
-  /** Google Cloud Platform Text Embedding Gecko model */
-  public static final OrchestrationAiModel TEXT_EMBEDDING_GECKO =
-      new OrchestrationAiModel("textembedding-gecko");
-
-  /** Google Cloud Platform Text Embedding Gecko Multilingual model */
-  public static final OrchestrationAiModel TEXT_EMBEDDING_GECKO_MULTILINGUAL =
-      new OrchestrationAiModel("textembedding-gecko-multilingual");
-
   /** Google Cloud Platform Gemini 1.0 Pro model */
   public static final OrchestrationAiModel GEMINI_1_0_PRO =
       new OrchestrationAiModel("gemini-1.0-pro");
@@ -120,50 +108,28 @@ public class OrchestrationAiModel {
       new OrchestrationAiModel("gemini-1.5-flash");
 
   OrchestrationAiModel(@Nonnull final String modelName) {
-    config = new LLMModuleConfig().modelName(modelName).modelParams(Map.of());
+    this.modelName = modelName;
   }
 
-  /**
-   * Set model version on this model.
-   *
-   * <pre>{@code
-   * .modelVersion("latest")
-   * }</pre>
-   *
-   * @param version The new version.
-   * @return New instance of this class with new version.
-   */
   @Nonnull
-  public OrchestrationAiModel modelVersion(@Nonnull final String version) {
-    return new OrchestrationAiModel(
-        new LLMModuleConfig()
-            .modelVersion(version)
-            .modelParams(config.getModelParams())
-            .modelName(config.getModelName()));
+  LLMModuleConfig createConfig() {
+    return new LLMModuleConfig()
+        .modelName(modelName)
+        .modelParams(modelParams)
+        .modelVersion(modelVersion);
   }
 
-  /**
-   * Set model parameters on this model.
-   *
-   * <pre>{@code
-   * .modelParams(
-   *    Map.of(
-   *        "max_tokens", 50,
-   *        "temperature", 0.1,
-   *        "frequency_penalty", 0,
-   *        "presence_penalty", 0));
-   * }</pre>
-   *
-   * @param modelParams Map of parameters.
-   * @return New instance of this class.
-   */
+  /** {@inheritDoc} */
   @Nonnull
-  public OrchestrationAiModel modelParams(
-      @Nonnull final Map<String, ? extends Number> modelParams) {
-    return new OrchestrationAiModel(
-        new LLMModuleConfig()
-            .modelVersion(config.getModelVersion())
-            .modelParams(modelParams)
-            .modelName(config.getModelName()));
+  @Override
+  public String name() {
+    return modelName;
+  }
+
+  /** {@inheritDoc} */
+  @Nullable
+  @Override
+  public String version() {
+    return modelVersion;
   }
 }

orchestration/src/main/java/com/sap/ai/sdk/orchestration/OrchestrationModuleConfig.java

@@ -4,6 +4,7 @@
 import com.sap.ai.sdk.orchestration.client.model.LLMModuleConfig;
 import com.sap.ai.sdk.orchestration.client.model.MaskingModuleConfig;
 import com.sap.ai.sdk.orchestration.client.model.TemplatingModuleConfig;
+import javax.annotation.Nonnull;
 import javax.annotation.Nullable;
 import lombok.AccessLevel;
 import lombok.AllArgsConstructor;
@@ -28,7 +29,6 @@
  * </ul>
  */
 @Value
-@With
 @AllArgsConstructor(access = AccessLevel.PRIVATE)
 @NoArgsConstructor(force = true)
 public class OrchestrationModuleConfig {
@@ -41,11 +41,23 @@ public class OrchestrationModuleConfig {
    * A template to be populated with input parameters. Upon request execution, this template will be
    * enhanced with any messages and parameter values from {@link OrchestrationPrompt}.
    */
-  @Nullable TemplatingModuleConfig templateConfig;
+  @With @Nullable TemplatingModuleConfig templateConfig;
 
   /** A masking configuration to pseudonymous or anonymize sensitive data in the input. */
-  @Nullable MaskingModuleConfig maskingConfig;
+  @With @Nullable MaskingModuleConfig maskingConfig;
 
   /** A content filter to filter the prompt. */
-  @Nullable FilteringModuleConfig filteringConfig;
+  @With @Nullable FilteringModuleConfig filteringConfig;
+
+  /**
+   * Creates a new configuration with the given LLM configuration.
+   *
+   * @param aiModel The LLM configuration to use.
+   * @return A new configuration with the given LLM configuration.
+   */
+  @Nonnull
+  public OrchestrationModuleConfig withLlmConfig(@Nonnull final OrchestrationAiModel aiModel) {
+    return new OrchestrationModuleConfig(
+        aiModel.createConfig(), templateConfig, maskingConfig, filteringConfig);
+  }
 }

orchestration/src/test/java/com/sap/ai/sdk/orchestration/ConfigToRequestTransformerTest.java

@@ -1,6 +1,6 @@
 package com.sap.ai.sdk.orchestration;
 
-import static com.sap.ai.sdk.orchestration.OrchestrationUnitTest.LLM_CONFIG;
+import static com.sap.ai.sdk.orchestration.OrchestrationUnitTest.CUSTOM_GPT_35;
 import static org.assertj.core.api.Assertions.assertThat;
 import static org.assertj.core.api.Assertions.assertThatThrownBy;
 
@@ -71,7 +71,7 @@ void testMessagesHistory() {
     var prompt = new OrchestrationPrompt("bar").messageHistory(List.of(systemMessage));
     var actual =
         ConfigToRequestTransformer.toCompletionPostRequest(
-            prompt, new OrchestrationModuleConfig().withLlmConfig(LLM_CONFIG));
+            prompt, new OrchestrationModuleConfig().withLlmConfig(CUSTOM_GPT_35));
 
     assertThat(actual.getMessagesHistory()).containsExactly(systemMessage);
   }

orchestration/src/test/java/com/sap/ai/sdk/orchestration/OrchestrationUnitTest.java

@@ -16,6 +16,7 @@
 import static com.github.tomakehurst.wiremock.client.WireMock.stubFor;
 import static com.github.tomakehurst.wiremock.client.WireMock.urlPathEqualTo;
 import static com.github.tomakehurst.wiremock.client.WireMock.verify;
+import static com.sap.ai.sdk.orchestration.OrchestrationAiModel.GPT_35_TURBO_16K;
 import static com.sap.ai.sdk.orchestration.client.model.AzureThreshold.NUMBER_0;
 import static com.sap.ai.sdk.orchestration.client.model.AzureThreshold.NUMBER_4;
 import static org.apache.hc.core5.http.HttpStatus.SC_BAD_REQUEST;
@@ -39,7 +40,6 @@
 import com.sap.ai.sdk.orchestration.client.model.FilteringModuleConfig;
 import com.sap.ai.sdk.orchestration.client.model.GenericModuleResult;
 import com.sap.ai.sdk.orchestration.client.model.InputFilteringConfig;
-import com.sap.ai.sdk.orchestration.client.model.LLMModuleConfig;
 import com.sap.ai.sdk.orchestration.client.model.LLMModuleResultSynchronous;
 import com.sap.ai.sdk.orchestration.client.model.MaskingModuleConfig;
 import com.sap.ai.sdk.orchestration.client.model.OutputFilteringConfig;
@@ -62,16 +62,15 @@
  */
 @WireMockTest
 class OrchestrationUnitTest {
-  static final LLMModuleConfig LLM_CONFIG =
-      OrchestrationAiModel.GPT_35_TURBO_16K
-          .modelParams(
+  static final OrchestrationAiModel CUSTOM_GPT_35 =
+      GPT_35_TURBO_16K
+          .withModelParams(
               Map.of(
                   "max_tokens", 50,
                   "temperature", 0.1,
                   "frequency_penalty", 0,
                   "presence_penalty", 0))
-          .modelVersion("latest")
-          .getConfig();
+          .withModelVersion("latest");
   private final Function<String, InputStream> fileLoader =
       filename -> Objects.requireNonNull(getClass().getClassLoader().getResourceAsStream(filename));
 
@@ -107,7 +106,7 @@ void setup(WireMockRuntimeInfo server) {
             .forDeploymentByScenario("orchestration")
             .withResourceGroup("my-resource-group");
     client = new OrchestrationClient(deployment);
-    config = new OrchestrationModuleConfig().withLlmConfig(LLM_CONFIG);
+    config = new OrchestrationModuleConfig().withLlmConfig(CUSTOM_GPT_35);
     prompt = new OrchestrationPrompt("Hello World! Why is this phrase so famous?");
   }
 

sample-code/spring-app/src/main/java/com/sap/ai/sdk/app/controllers/OrchestrationController.java

@@ -32,8 +32,7 @@
 @RequestMapping("/orchestration")
 class OrchestrationController {
   private final OrchestrationClient client = new OrchestrationClient();
-  OrchestrationModuleConfig config =
-      new OrchestrationModuleConfig().withLlmConfig(GPT_35_TURBO.getConfig());
+  OrchestrationModuleConfig config = new OrchestrationModuleConfig().withLlmConfig(GPT_35_TURBO);
 
   /**
    * Chat request to OpenAI through the Orchestration service with a simple prompt.