fix: java fallback

Author: sinedied

docs/sections/java-quarkus/06-chat-api.md

@@ -25,22 +25,29 @@ Before we can create the clients, we need to retrieve the credentials to access
 Add this code under the `TODO:` to retrieve the token to build the `AzureOpenAIChatModel`:
 
 ```java
-AzureOpenAiChatModel model;
-
-try {
-  // Use the current user identity to authenticate with Azure OpenAI.
-  // (no secrets needed, just use `az login` or `azd auth login` locally, and managed identity when deployed on Azure).
-  model = AzureOpenAiChatModel.builder()
-    .tokenCredential(new DefaultAzureCredentialBuilder().build())
-    .endpoint(azureOpenAiEndpoint)
-    .deploymentName(azureOpenAiDeploymentName)
-    .timeout(ofSeconds(60))
-    .logRequestsAndResponses(true)
-    .build();
-} catch (Exception e) {
-  // default value for local execution
-  log.info("### Using fallback configuration for OpenAI");
-}
+    AzureOpenAiChatModel model;
+
+    try {
+      // Use the current user identity to authenticate with Azure OpenAI.
+      // (no secrets needed, just use `az login` or `azd auth login` locally, and managed identity when deployed on Azure).
+      DefaultAzureCredential credentials = new DefaultAzureCredentialBuilder().build();
+
+      // Try requesting a token, so we can fallback to the other builder if it doesn't work
+      TokenRequestContext request = new TokenRequestContext();
+      request.addScopes("https://cognitiveservices.azure.com/.default");
+      credentials.getTokenSync(request);
+
+      model = AzureOpenAiChatModel.builder()
+        .tokenCredential(credentials)
+        .endpoint(azureOpenAiEndpoint)
+        .deploymentName(azureOpenAiDeploymentName)
+        .timeout(ofSeconds(60))
+        .logRequestsAndResponses(true)
+        .build();
+    } catch (Exception e) {
+      // Default value for local execution
+      // ...
+    }
 ```
 
 This will use the current user identity to authenticate with Azure OpenAI and AI Search. We don't need to provide any secrets, just use `az login` (or `azd auth login`) locally, and [managed identity](https://learn.microsoft.com/entra/identity/managed-identities-azure-resources/overview) when deployed on Azure.
@@ -58,7 +65,7 @@ To use the fallback, add the following code in the catch statement and return th
 
 ```java
   } catch (Exception e) {
-    // default value for local execution
+    // Default value for local execution
     log.info("### Using fallback configuration for OpenAI");
     model = AzureOpenAiChatModel.builder()
       .apiKey("__dummy")

scripts/repo/create-github-template.sh

@@ -268,6 +268,9 @@ import dev.langchain4j.model.azure.AzureOpenAiChatModel;
 import dev.langchain4j.model.chat.ChatLanguageModel;
 import jakarta.enterprise.inject.Produces;
 import org.eclipse.microprofile.config.inject.ConfigProperty;
+
+import com.azure.core.credential.TokenRequestContext;
+import com.azure.identity.DefaultAzureCredential;
 import com.azure.identity.DefaultAzureCredentialBuilder;
 
 import static java.time.Duration.ofSeconds;

src/backend-java-quarkus/src/main/java/ai/azure/openai/rag/workshop/backend/configuration/ChatLanguageModelAzureOpenAiProducer.java

@@ -4,6 +4,9 @@
 import dev.langchain4j.model.chat.ChatLanguageModel;
 import jakarta.enterprise.inject.Produces;
 import org.eclipse.microprofile.config.inject.ConfigProperty;
+
+import com.azure.core.credential.TokenRequestContext;
+import com.azure.identity.DefaultAzureCredential;
 import com.azure.identity.DefaultAzureCredentialBuilder;
 
 import static java.time.Duration.ofSeconds;
@@ -22,27 +25,36 @@ public class ChatLanguageModelAzureOpenAiProducer {
 
   @Produces
   public ChatLanguageModel chatLanguageModel() {
-
+    // initialize chat model here
     AzureOpenAiChatModel model;
+
     try {
-      // Instantiate the DefaultAzureCredential using managed identities
+      // Use the current user identity to authenticate with Azure OpenAI.
+      // (no secrets needed, just use `az login` or `azd auth login` locally, and managed identity when deployed on Azure).
+      DefaultAzureCredential credentials = new DefaultAzureCredentialBuilder().build();
+
+      // Try requesting a token, so we can fallback to the other builder if it doesn't work
+      TokenRequestContext request = new TokenRequestContext();
+      request.addScopes("https://cognitiveservices.azure.com/.default");
+      credentials.getTokenSync(request);
+
       model = AzureOpenAiChatModel.builder()
-      .tokenCredential(new DefaultAzureCredentialBuilder().build())
-      .endpoint(azureOpenAiEndpoint)
-      .deploymentName(azureOpenAiDeploymentName)
-      .timeout(ofSeconds(60))
-      .logRequestsAndResponses(true)
-      .build();
+        .tokenCredential(credentials)
+        .endpoint(azureOpenAiEndpoint)
+        .deploymentName(azureOpenAiDeploymentName)
+        .timeout(ofSeconds(60))
+        .logRequestsAndResponses(true)
+        .build();
     } catch (Exception e) {
-      // default value for local execution
+      // Default value for local execution
       log.info("### Using fallback configuration for OpenAI");
       model = AzureOpenAiChatModel.builder()
-      .apiKey("__dummy")
-      .endpoint(azureOpenAiEndpoint)
-      .deploymentName(azureOpenAiDeploymentName)
-      .timeout(ofSeconds(60))
-      .logRequestsAndResponses(true)
-      .build();
+        .apiKey("__dummy")
+        .endpoint(azureOpenAiEndpoint)
+        .deploymentName(azureOpenAiDeploymentName)
+        .timeout(ofSeconds(60))
+        .logRequestsAndResponses(true)
+        .build();
     }
 
     log.info("### Producing ChatLanguageModel with AzureOpenAiChatModel");