Add Azure OpenAI structured outputs support

langchain4j-azure-open-ai-spring-boot-starter/src/main/java/dev/langchain4j/azure/openai/spring/AutoConfig.java

@@ -4,6 +4,7 @@
 import com.azure.core.util.Configuration;
 import dev.langchain4j.model.Tokenizer;
 import dev.langchain4j.model.azure.*;
+import dev.langchain4j.model.chat.request.ResponseFormat;
 import org.springframework.boot.autoconfigure.AutoConfiguration;
 import org.springframework.boot.autoconfigure.condition.ConditionalOnMissingBean;
 import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
@@ -46,6 +47,8 @@ AzureOpenAiChatModel openAiChatModel(Properties properties) {
                 .presencePenalty(chatModelProperties.presencePenalty())
                 .frequencyPenalty(chatModelProperties.frequencyPenalty())
                 .seed(chatModelProperties.seed())
+                .responseFormat(chatModelProperties.responseFormat() != null && chatModelProperties.responseFormat().equalsIgnoreCase("json") ? ResponseFormat.JSON : ResponseFormat.TEXT)
+                .strictJsonSchema(chatModelProperties.strictJsonSchema())
                 .timeout(Duration.ofSeconds(chatModelProperties.timeout() == null ? 0 : chatModelProperties.timeout()))
                 .maxRetries(chatModelProperties.maxRetries())
                 .proxyOptions(ProxyOptions.fromConfiguration(Configuration.getGlobalConfiguration()))

langchain4j-azure-open-ai-spring-boot-starter/src/main/java/dev/langchain4j/azure/openai/spring/ChatModelProperties.java

@@ -19,6 +19,7 @@ record ChatModelProperties(
     Double frequencyPenalty,
     Long seed,
     String responseFormat,
+    Boolean strictJsonSchema,
     Integer timeout, // TODO use Duration instead
     Integer maxRetries,
     Boolean logRequestsAndResponses,

langchain4j-azure-open-ai-spring-boot-starter/src/test/java/dev/langchain4j/azure/openai/spring/AutoConfigIT.java

@@ -8,6 +8,12 @@
 import dev.langchain4j.model.azure.AzureOpenAiStreamingChatModel;
 import dev.langchain4j.model.chat.ChatLanguageModel;
 import dev.langchain4j.model.chat.StreamingChatLanguageModel;
+import dev.langchain4j.model.chat.request.ChatRequest;
+import dev.langchain4j.model.chat.request.ResponseFormat;
+import dev.langchain4j.model.chat.request.json.JsonArraySchema;
+import dev.langchain4j.model.chat.request.json.JsonObjectSchema;
+import dev.langchain4j.model.chat.request.json.JsonSchema;
+import dev.langchain4j.model.chat.request.json.JsonStringSchema;
 import dev.langchain4j.model.embedding.EmbeddingModel;
 import dev.langchain4j.model.image.ImageModel;
 import dev.langchain4j.model.output.Response;
@@ -17,8 +23,12 @@
 import org.springframework.boot.autoconfigure.AutoConfigurations;
 import org.springframework.boot.test.context.runner.ApplicationContextRunner;
 
+import java.util.List;
 import java.util.concurrent.CompletableFuture;
 
+import static dev.langchain4j.data.message.UserMessage.userMessage;
+import static dev.langchain4j.model.chat.request.ResponseFormatType.JSON;
+import static java.util.Collections.singletonList;
 import static java.util.concurrent.TimeUnit.SECONDS;
 import static org.assertj.core.api.Assertions.assertThat;
 
@@ -53,6 +63,52 @@ void should_provide_chat_model(String deploymentName) {
                 });
     }
 
+    class Person {
+
+        String name;
+        List<String> favouriteColors;
+    }
+
+    @ParameterizedTest(name = "Deployment name: {0}")
+    @CsvSource({
+            "gpt-4o-mini"
+    })
+    void should_provide_chat_model_with_json_schema(String deploymentName) {
+        contextRunner
+                .withPropertyValues(
+                        "langchain4j.azure-open-ai.chat-model.api-key=" + AZURE_OPENAI_KEY,
+                        "langchain4j.azure-open-ai.chat-model.endpoint=" + AZURE_OPENAI_ENDPOINT,
+                        "langchain4j.azure-open-ai.chat-model.deployment-name=" + deploymentName,
+                        "langchain4j.azure-open-ai.chat-model.strict-json-schema=true"
+                )
+                .run(context -> {
+
+                    ChatLanguageModel chatLanguageModel = context.getBean(ChatLanguageModel.class);
+
+                    ChatRequest chatRequest = ChatRequest.builder()
+                            .messages(singletonList(userMessage("Julien likes blue, white and red")))
+                            .responseFormat(ResponseFormat.builder()
+                                    .type(JSON)
+                                    .jsonSchema(JsonSchema.builder()
+                                            .name("Person")
+                                            .rootElement(JsonObjectSchema.builder()
+                                                    .addStringProperty("name")
+                                                    .addProperty("favouriteColors", JsonArraySchema.builder()
+                                                            .items(new JsonStringSchema())
+                                                            .build())
+                                                    .required("name", "favouriteColors")
+                                                    .build())
+                                            .build())
+                                    .build())
+                            .build();
+
+                    assertThat(chatLanguageModel).isInstanceOf(AzureOpenAiChatModel.class);
+                    AiMessage aiMessage = chatLanguageModel.chat(chatRequest).aiMessage();
+                    assertThat(aiMessage.text()).contains("{\"name\":\"Julien\",\"favouriteColors\":[\"blue\",\"white\",\"red\"]}");
+                    assertThat(context.getBean(AzureOpenAiChatModel.class)).isSameAs(chatLanguageModel);
+                });
+    }
+
     @ParameterizedTest(name = "Deployment name: {0}")
     @CsvSource({
             "gpt-3.5-turbo"