Add Nvidia integration for Completion and Chat Completion

Author: Jan-Kazlouski-elastic

server/src/main/java/org/elasticsearch/TransportVersions.java

@@ -353,6 +353,7 @@ static TransportVersion def(int id) {
     public static final TransportVersion NODE_WEIGHTS_ADDED_TO_NODE_BALANCE_STATS = def(9_129_0_00);
     public static final TransportVersion RERANK_SNIPPETS = def(9_130_0_00);
     public static final TransportVersion PIPELINE_TRACKING_INFO = def(9_131_0_00);
+    public static final TransportVersion ML_INFERENCE_NVIDIA_ADDED = def(9_132_0_00);
 
     /*
      * STOP! READ THIS FIRST! No, really,

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferenceNamedWriteablesProvider.java

@@ -110,6 +110,7 @@
 import org.elasticsearch.xpack.inference.services.llama.embeddings.LlamaEmbeddingsServiceSettings;
 import org.elasticsearch.xpack.inference.services.mistral.completion.MistralChatCompletionServiceSettings;
 import org.elasticsearch.xpack.inference.services.mistral.embeddings.MistralEmbeddingsServiceSettings;
+import org.elasticsearch.xpack.inference.services.nvidia.completion.NvidiaChatCompletionServiceSettings;
 import org.elasticsearch.xpack.inference.services.openai.completion.OpenAiChatCompletionServiceSettings;
 import org.elasticsearch.xpack.inference.services.openai.completion.OpenAiChatCompletionTaskSettings;
 import org.elasticsearch.xpack.inference.services.openai.embeddings.OpenAiEmbeddingsServiceSettings;
@@ -178,6 +179,7 @@ public static List<NamedWriteableRegistry.Entry> getNamedWriteables() {
         addVoyageAINamedWriteables(namedWriteables);
         addCustomNamedWriteables(namedWriteables);
         addLlamaNamedWriteables(namedWriteables);
+        addNvidiaNamedWriteables(namedWriteables);
 
         addUnifiedNamedWriteables(namedWriteables);
 
@@ -298,6 +300,17 @@ private static void addLlamaNamedWriteables(List<NamedWriteableRegistry.Entry> n
         // no task settings for Llama
     }
 
+    private static void addNvidiaNamedWriteables(List<NamedWriteableRegistry.Entry> namedWriteables) {
+        namedWriteables.add(
+            new NamedWriteableRegistry.Entry(
+                ServiceSettings.class,
+                NvidiaChatCompletionServiceSettings.NAME,
+                NvidiaChatCompletionServiceSettings::new
+            )
+        );
+        // no task settings for Nvidia
+    }
+
     private static void addAzureAiStudioNamedWriteables(List<NamedWriteableRegistry.Entry> namedWriteables) {
         namedWriteables.add(
             new NamedWriteableRegistry.Entry(

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/InferencePlugin.java

@@ -135,6 +135,7 @@
 import org.elasticsearch.xpack.inference.services.jinaai.JinaAIService;
 import org.elasticsearch.xpack.inference.services.llama.LlamaService;
 import org.elasticsearch.xpack.inference.services.mistral.MistralService;
+import org.elasticsearch.xpack.inference.services.nvidia.NvidiaService;
 import org.elasticsearch.xpack.inference.services.openai.OpenAiService;
 import org.elasticsearch.xpack.inference.services.sagemaker.SageMakerClient;
 import org.elasticsearch.xpack.inference.services.sagemaker.SageMakerService;
@@ -413,6 +414,7 @@ public List<InferenceServiceExtension.Factory> getInferenceServiceFactories() {
             context -> new VoyageAIService(httpFactory.get(), serviceComponents.get(), context),
             context -> new DeepSeekService(httpFactory.get(), serviceComponents.get(), context),
             context -> new LlamaService(httpFactory.get(), serviceComponents.get(), context),
+            context -> new NvidiaService(httpFactory.get(), serviceComponents.get(), context),
             ElasticsearchInternalService::new,
             context -> new CustomService(httpFactory.get(), serviceComponents.get(), context)
         );

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/llama/LlamaModel.java

@@ -76,7 +76,7 @@ public void setURI(String newUri) {
     /**
      * Retrieves the secret settings from the provided map of secrets.
      * If the map is null or empty, it returns an instance of EmptySecretSettings.
-     * Caused by the fact that Llama model doesn't have out of the box security settings and can be used witout authentication.
+     * Caused by the fact that Llama model doesn't have out of the box security settings and can be used without authentication.
      *
      * @param secrets the map containing secret settings
      * @return an instance of SecretSettings

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/nvidia/NvidiaModel.java

@@ -0,0 +1,54 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the Elastic License
+ * 2.0; you may not use this file except in compliance with the Elastic License
+ * 2.0.
+ */
+
+package org.elasticsearch.xpack.inference.services.nvidia;
+
+import org.elasticsearch.inference.ModelConfigurations;
+import org.elasticsearch.inference.ModelSecrets;
+import org.elasticsearch.inference.ServiceSettings;
+import org.elasticsearch.xpack.inference.external.action.ExecutableAction;
+import org.elasticsearch.xpack.inference.services.RateLimitGroupingModel;
+import org.elasticsearch.xpack.inference.services.nvidia.action.NvidiaActionVisitor;
+import org.elasticsearch.xpack.inference.services.settings.DefaultSecretSettings;
+
+/**
+ * Abstract class representing an Nvidia model for inference.
+ * This class extends RateLimitGroupingModel and provides common functionality for Nvidia models.
+ */
+public abstract class NvidiaModel extends RateLimitGroupingModel {
+
+    /**
+     * Constructor for creating a NvidiaModel with specified configurations and secrets.
+     *
+     * @param configurations the model configurations
+     * @param secrets the secret settings for the model
+     */
+    protected NvidiaModel(ModelConfigurations configurations, ModelSecrets secrets) {
+        super(configurations, secrets);
+    }
+
+    /**
+     * Constructor for creating a NvidiaModel with specified model, service settings, and secret settings.
+     * @param model the model configurations
+     * @param serviceSettings the settings for the inference service
+     */
+    protected NvidiaModel(RateLimitGroupingModel model, ServiceSettings serviceSettings) {
+        super(model, serviceSettings);
+    }
+
+    @Override
+    public DefaultSecretSettings getSecretSettings() {
+        return (DefaultSecretSettings) super.getSecretSettings();
+    }
+
+    /**
+     * Accepts a visitor to create an executable action for the Nvidia model.
+     * @param creator the visitor that creates the executable action
+     * @return an executable action for the Nvidia model
+     */
+    protected abstract ExecutableAction accept(NvidiaActionVisitor creator);
+}