AI rule node: support timeout for Vertex AI

application/src/main/java/org/thingsboard/server/service/ai/Langchain4jChatModelConfigurerImpl.java

@@ -15,10 +15,14 @@
  */
 package org.thingsboard.server.service.ai;
 
+import com.google.api.gax.core.FixedCredentialsProvider;
+import com.google.api.gax.retrying.RetrySettings;
 import com.google.auth.oauth2.ServiceAccountCredentials;
 import com.google.cloud.vertexai.Transport;
 import com.google.cloud.vertexai.VertexAI;
 import com.google.cloud.vertexai.api.GenerationConfig;
+import com.google.cloud.vertexai.api.PredictionServiceClient;
+import com.google.cloud.vertexai.api.PredictionServiceSettings;
 import com.google.cloud.vertexai.generativeai.GenerativeModel;
 import dev.langchain4j.model.bedrock.BedrockChatModel;
 import dev.langchain4j.model.chat.ChatModel;
@@ -96,17 +100,44 @@ class Langchain4jChatModelConfigurerImpl implements Langchain4jChatModelConfigur
         // construct service account credentials using service account key JSON
         ServiceAccountCredentials serviceAccountCredentials;
         try {
-            serviceAccountCredentials = ServiceAccountCredentials
+            serviceAccountCredentials = ServiceAccountCredentials.fromStream(
-                    .fromStream(new ByteArrayInputStream(JacksonUtil.writeValueAsBytes(providerConfig.serviceAccountKey())));
+                    new ByteArrayInputStream(JacksonUtil.writeValueAsBytes(providerConfig.serviceAccountKey()))
+            );
         } catch (IOException e) {
             throw new RuntimeException("Failed to parse service account key JSON", e);
         }
 
+        PredictionServiceSettings predictionServiceClientSettings;
+        try {
+            // create prediction service settings for REST transport with service account key credentials
+            PredictionServiceSettings.Builder settingsBuilder = PredictionServiceSettings.newHttpJsonBuilder()
+                    .setCredentialsProvider(FixedCredentialsProvider.create(serviceAccountCredentials));
+
+            // get the retry settings that control request timeout for generateContent RPC
+            RetrySettings.Builder retrySettings = settingsBuilder
+                    .generateContentSettings()
+                    .getRetrySettings()
+                    .toBuilder();
+
+            // set request timeout from model config
+            if (modelConfig.timeoutSeconds() != null) {
+                retrySettings.setTotalTimeout(org.threeten.bp.Duration.ofSeconds(modelConfig.timeoutSeconds()));
+            }
+
+            // set updated retry settings
+            settingsBuilder.generateContentSettings().setRetrySettings(retrySettings.build());
+
+            // build the client settings
+            predictionServiceClientSettings = settingsBuilder.build();
+        } catch (IOException e) {
+            throw new RuntimeException("Failed to create prediction service client settings", e);
+        }
+
         // construct Vertex AI instance
         var vertexAI = new VertexAI.Builder()
                 .setProjectId(providerConfig.projectId())
                 .setLocation(providerConfig.location())
-                .setCredentials(serviceAccountCredentials)
+                .setPredictionClientSupplier(() -> createPredictionServiceClient(predictionServiceClientSettings))
                 .setTransport(Transport.REST) // GRPC also possible, but likely does not work with service account keys
                 .build();
 
@@ -121,12 +152,19 @@ class Langchain4jChatModelConfigurerImpl implements Langchain4jChatModelConfigur
         var generationConfig = generationConfigBuilder.build();
 
         // construct generative model instance
-        var generativeModel = new GenerativeModel(modelConfig.modelId(), vertexAI)
+        var generativeModel = new GenerativeModel(modelConfig.modelId(), vertexAI).withGenerationConfig(generationConfig);
-                .withGenerationConfig(generationConfig);
 
         return new VertexAiGeminiChatModel(generativeModel, generationConfig, modelConfig.maxRetries());
     }
 
+    private static PredictionServiceClient createPredictionServiceClient(PredictionServiceSettings settings) {
+        try {
+            return PredictionServiceClient.create(settings);
+        } catch (IOException e) {
+            throw new RuntimeException("Failed to create prediction service client", e);
+        }
+    }
+
     @Override
     public ChatModel configureChatModel(MistralAiChatModel chatModel) {
         MistralAiChatModel.Config modelConfig = chatModel.modelConfig();

common/data/src/main/java/org/thingsboard/server/common/data/ai/model/chat/GoogleVertexAiGeminiChatModel.java

@@ -31,7 +31,7 @@ public record GoogleVertexAiGeminiChatModel(
             String modelId,
             Double temperature,
             Double topP,
-            Integer timeoutSeconds, // TODO: not supported by Vertex AI
+            Integer timeoutSeconds,
             Integer maxRetries
     ) implements AiChatModelConfig<GoogleVertexAiGeminiChatModel.Config> {}