Merge branch 'main' into workflow/compatibility-version

.github/workflows/continuous-integration.yaml

@@ -22,6 +22,7 @@ jobs:
         with:
           fetch-depth: 1
           ref: ${{ github.event.pull_request.head.ref }}
+          repository: ${{ github.event.pull_request.head.repo.full_name }}
 
       - name: "Setup java"
         uses: actions/setup-java@v4

.github/workflows/perform-release.yml

@@ -88,9 +88,6 @@ jobs:
         with:
           distribution: "sapmachine"
           java-version: ${{ env.JAVA_VERSION }}
-          server-id: ossrh
-          server-username: MAVEN_CENTRAL_USER # env variable for username in deploy
-          server-password: MAVEN_CENTRAL_PASSWORD # env variable for token in deploy
 
       - name: "Download Release Asset"
         id: download-asset
@@ -113,8 +110,8 @@ jobs:
 
       - name: "Deploy"
         run: |
-          MVN_ARGS="${{ env.MVN_CLI_ARGS }} deploy -Drelease -s settings.xml"
-          mvn $MVN_ARGS
+          MVN_ARGS="${{ env.MVN_CLI_ARGS }} -Drelease -s settings.xml"
+          mvn deploy $MVN_ARGS
         env:
           MAVEN_GPG_PASSPHRASE: ${{ secrets.PGP_PASSPHRASE }}
 

README.md

@@ -165,7 +165,7 @@ After restarting your application, you should see an "aicore" entry in the `VCAP
 
     - **Name**: `my-aicore`
     - **Type**: `HTTP`
-    - **URL**: `[serviceurls.AI_API_URL]/v2` (append `/v2` to the URL)
+    - **URL**: `[serviceurls.AI_API_URL]`
     - **Proxy Type**: `Internet`
     - **Authentication**: `OAuth2ClientCredentials`
     - **Client ID**: `[clientid]`

docs/guides/ORCHESTRATION_CHAT_COMPLETION.md

@@ -91,8 +91,7 @@ var prompt = new OrchestrationPrompt("Hello world! Why is this phrase so famous?
 
 var result = client.chatCompletion(prompt, config);
 
-String messageResult =
-        result.getOrchestrationResult().getChoices().get(0).getMessage().getContent();
+String messageResult = result.getContent();
 ```
 
 In this example, the Orchestration service generates a response to the user message "Hello world! Why is this phrase so famous?".

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

@@ -0,0 +1,25 @@
+package com.sap.ai.sdk.orchestration;
+
+import com.fasterxml.jackson.annotation.JsonTypeInfo;
+import com.fasterxml.jackson.databind.annotation.JsonDeserialize;
+import com.sap.ai.sdk.orchestration.client.model.LLMChoice;
+import com.sap.ai.sdk.orchestration.client.model.LLMModuleResultSynchronous;
+import lombok.AccessLevel;
+import lombok.NoArgsConstructor;
+
+@NoArgsConstructor(access = AccessLevel.PRIVATE)
+final class JacksonMixins {
+  /** Mixin to enforce a specific subtype to be deserialized always. */
+  @JsonTypeInfo(use = JsonTypeInfo.Id.NONE)
+  @JsonDeserialize(as = LLMModuleResultSynchronous.class)
+  interface LLMModuleResultMixIn {}
+
+  /** Mixin to enforce a specific subtype to be deserialized always. */
+  @JsonTypeInfo(use = JsonTypeInfo.Id.NONE)
+  @JsonDeserialize(as = LLMChoice.class)
+  interface ModuleResultsOutputUnmaskingInnerMixIn {}
+
+  /** Mixin to suppress @JsonTypeInfo for oneOf interfaces. */
+  @JsonTypeInfo(use = JsonTypeInfo.Id.NONE)
+  interface NoTypeInfoMixin {}
+}

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

@@ -0,0 +1,41 @@
+package com.sap.ai.sdk.orchestration;
+
+import static lombok.AccessLevel.PACKAGE;
+
+import com.sap.ai.sdk.orchestration.client.model.CompletionPostResponse;
+import com.sap.ai.sdk.orchestration.client.model.LLMModuleResultSynchronous;
+import javax.annotation.Nonnull;
+import lombok.RequiredArgsConstructor;
+import lombok.Value;
+
+/** Orchestration chat completion output. */
+@Value
+@RequiredArgsConstructor(access = PACKAGE)
+public class OrchestrationChatResponse {
+  CompletionPostResponse originalResponse;
+
+  /**
+   * Get the message content from the output.
+   *
+   * <p>Note: If there are multiple choices only the first one is returned
+   *
+   * @return the message content or empty string.
+   * @throws OrchestrationClientException if the content filter filtered the output.
+   */
+  @Nonnull
+  public String getContent() throws OrchestrationClientException {
+    final var choices =
+        ((LLMModuleResultSynchronous) originalResponse.getOrchestrationResult()).getChoices();
+
+    if (choices.isEmpty()) {
+      return "";
+    }
+
+    final var choice = choices.get(0);
+
+    if ("content_filter".equals(choice.getFinishReason())) {
+      throw new OrchestrationClientException("Content filter filtered the output.");
+    }
+    return choice.getMessage().getContent();
+  }
+}

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

@@ -45,11 +45,13 @@ public class OrchestrationClient {
             .visibility(PropertyAccessor.GETTER, JsonAutoDetect.Visibility.NONE)
             .visibility(PropertyAccessor.SETTER, JsonAutoDetect.Visibility.NONE)
             .serializationInclusion(JsonInclude.Include.NON_NULL)
-            .mixIn(LLMModuleResult.class, LLMModuleResultMixIn.class)
-            .mixIn(ModuleResultsOutputUnmaskingInner.class, NoTypeInfoMixin.class)
-            .mixIn(FilterConfig.class, NoTypeInfoMixin.class)
-            .mixIn(MaskingProviderConfig.class, NoTypeInfoMixin.class)
-            .mixIn(TemplatingModuleConfig.class, NoTypeInfoMixin.class)
+            .mixIn(LLMModuleResult.class, JacksonMixins.LLMModuleResultMixIn.class)
+            .mixIn(
+                ModuleResultsOutputUnmaskingInner.class,
+                JacksonMixins.ModuleResultsOutputUnmaskingInnerMixIn.class)
+            .mixIn(FilterConfig.class, JacksonMixins.NoTypeInfoMixin.class)
+            .mixIn(MaskingProviderConfig.class, JacksonMixins.NoTypeInfoMixin.class)
+            .mixIn(TemplatingModuleConfig.class, JacksonMixins.NoTypeInfoMixin.class)
             .build();
   }
 
@@ -99,12 +101,12 @@ public static CompletionPostRequest toCompletionPostRequest(
    * @throws OrchestrationClientException if the request fails.
    */
   @Nonnull
-  public CompletionPostResponse chatCompletion(
+  public OrchestrationChatResponse chatCompletion(
       @Nonnull final OrchestrationPrompt prompt, @Nonnull final OrchestrationModuleConfig config)
       throws OrchestrationClientException {
 
     val request = toCompletionPostRequest(prompt, config);
-    return executeRequest(request);
+    return new OrchestrationChatResponse(executeRequest(request));
   }
 
   /**

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

@@ -121,8 +121,7 @@ void testCompletion() {
     final var result = client.chatCompletion(prompt, config);
 
     assertThat(result).isNotNull();
-    var orchestrationResult = (LLMModuleResultSynchronous) result.getOrchestrationResult();
-    assertThat(orchestrationResult.getChoices().get(0).getMessage().getContent()).isNotEmpty();
+    assertThat(result.getContent()).isNotEmpty();
   }
 
   @Test
@@ -141,11 +140,12 @@ void testTemplating() throws IOException {
     final var result =
         client.chatCompletion(new OrchestrationPrompt(inputParams, template), config);
 
-    assertThat(result.getRequestId()).isEqualTo("26ea36b5-c196-4806-a9a6-a686f0c6ad91");
-    assertThat(result.getModuleResults().getTemplating().get(0).getContent())
+    final var response = result.getOriginalResponse();
+    assertThat(response.getRequestId()).isEqualTo("26ea36b5-c196-4806-a9a6-a686f0c6ad91");
+    assertThat(response.getModuleResults().getTemplating().get(0).getContent())
         .isEqualTo("Reply with 'Orchestration Service is working!' in German");
-    assertThat(result.getModuleResults().getTemplating().get(0).getRole()).isEqualTo("user");
-    var llm = (LLMModuleResultSynchronous) result.getModuleResults().getLlm();
+    assertThat(response.getModuleResults().getTemplating().get(0).getRole()).isEqualTo("user");
+    var llm = (LLMModuleResultSynchronous) response.getModuleResults().getLlm();
     assertThat(llm.getId()).isEqualTo("chatcmpl-9lzPV4kLrXjFckOp2yY454wksWBoj");
     assertThat(llm.getObject()).isEqualTo("chat.completion");
     assertThat(llm.getCreated()).isEqualTo(1721224505);
@@ -160,7 +160,7 @@ void testTemplating() throws IOException {
     assertThat(usage.getCompletionTokens()).isEqualTo(7);
     assertThat(usage.getPromptTokens()).isEqualTo(19);
     assertThat(usage.getTotalTokens()).isEqualTo(26);
-    var orchestrationResult = (LLMModuleResultSynchronous) result.getOrchestrationResult();
+    var orchestrationResult = (LLMModuleResultSynchronous) response.getOrchestrationResult();
     assertThat(orchestrationResult.getId()).isEqualTo("chatcmpl-9lzPV4kLrXjFckOp2yY454wksWBoj");
     assertThat(orchestrationResult.getObject()).isEqualTo("chat.completion");
     assertThat(orchestrationResult.getCreated()).isEqualTo(1721224505);
@@ -286,7 +286,8 @@ void messagesHistory() throws IOException {
 
     final var result = client.chatCompletion(prompt, config);
 
-    assertThat(result.getRequestId()).isEqualTo("26ea36b5-c196-4806-a9a6-a686f0c6ad91");
+    assertThat(result.getOriginalResponse().getRequestId())
+        .isEqualTo("26ea36b5-c196-4806-a9a6-a686f0c6ad91");
 
     // verify that the history is sent correctly
     try (var requestInputStream = fileLoader.apply("messagesHistoryRequest.json")) {
@@ -298,7 +299,7 @@ void messagesHistory() throws IOException {
   }
 
   @Test
-  void maskingAnonymization() throws IOException {
+  void maskingPseudonymization() throws IOException {
     stubFor(
         post(urlPathEqualTo("/v2/inference/deployments/abcdef0123456789/completion"))
             .willReturn(
@@ -307,18 +308,16 @@ void maskingAnonymization() throws IOException {
                     .withHeader("Content-Type", "application/json")));
 
     final var maskingConfig =
-        createMaskingConfig(DPIConfig.MethodEnum.ANONYMIZATION, DPIEntities.PHONE);
+        createMaskingConfig(DPIConfig.MethodEnum.PSEUDONYMIZATION, DPIEntities.PHONE);
 
     final var result = client.chatCompletion(prompt, config.withMaskingConfig(maskingConfig));
+    final var response = result.getOriginalResponse();
 
-    assertThat(result).isNotNull();
-    GenericModuleResult inputMasking = result.getModuleResults().getInputMasking();
+    assertThat(response).isNotNull();
+    GenericModuleResult inputMasking = response.getModuleResults().getInputMasking();
     assertThat(inputMasking.getMessage()).isEqualTo("Input to LLM is masked successfully.");
     assertThat(inputMasking.getData()).isNotNull();
-    final var choices = ((LLMModuleResultSynchronous) result.getOrchestrationResult()).getChoices();
-    assertThat(choices.get(0).getMessage().getContent())
-        .isEqualTo(
-            "I'm sorry, I cannot provide information about specific individuals, including their nationality.");
+    assertThat(result.getContent()).contains("Hi Mallory");
 
     // verify that the request is sent correctly
     try (var requestInputStream = fileLoader.apply("maskingRequest.json")) {
@@ -414,4 +413,17 @@ void testErrorHandling() {
 
     softly.assertAll();
   }
+
+  @Test
+  void testEmptyChoicesResponse() {
+    stubFor(
+        post(urlPathEqualTo("/v2/inference/deployments/abcdef0123456789/completion"))
+            .willReturn(
+                aResponse()
+                    .withBodyFile("emptyChoicesResponse.json")
+                    .withHeader("Content-Type", "application/json")));
+    final var result = client.chatCompletion(prompt, config);
+
+    assertThat(result.getContent()).isEmpty();
+  }
 }

orchestration/src/test/resources/__files/emptyChoicesResponse.json

@@ -0,0 +1,35 @@
+{
+  "request_id": "26ea36b5-c196-4806-a9a6-a686f0c6ad91",
+  "module_results": {
+    "templating": [
+      {
+        "role": "user",
+        "content": "Reply with 'Orchestration Service is working!' in German"
+      }
+    ],
+    "llm": {
+      "id": "chatcmpl-9lzPV4kLrXjFckOp2yY454wksWBoj",
+      "object": "chat.completion",
+      "created": 1721224505,
+      "model": "gpt-35-turbo-16k",
+      "choices": [],
+      "usage": {
+        "completion_tokens": 7,
+        "prompt_tokens": 19,
+        "total_tokens": 26
+      }
+    }
+  },
+  "orchestration_result": {
+    "id": "chatcmpl-9lzPV4kLrXjFckOp2yY454wksWBoj",
+    "object": "chat.completion",
+    "created": 1721224505,
+    "model": "gpt-35-turbo-16k",
+    "choices": [],
+    "usage": {
+      "completion_tokens": 7,
+      "prompt_tokens": 19,
+      "total_tokens": 26
+    }
+  }
+}