Refactor Llama model classes to implement accept method for action visitors

Author: Jan-Kazlouski-elastic

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

@@ -12,7 +12,9 @@
 import org.elasticsearch.inference.ModelSecrets;
 import org.elasticsearch.inference.SecretSettings;
 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.llama.action.LlamaActionVisitor;
 import org.elasticsearch.xpack.inference.services.settings.DefaultSecretSettings;
 import org.elasticsearch.xpack.inference.services.settings.RateLimitSettings;
 
@@ -87,4 +89,6 @@ public void setURI(String newUri) {
     protected static SecretSettings retrieveSecretSettings(Map<String, Object> secrets) {
         return (secrets != null && secrets.isEmpty()) ? EmptySecretSettings.INSTANCE : DefaultSecretSettings.fromMap(secrets);
     }
+
+    protected abstract ExecutableAction accept(LlamaActionVisitor creator);
 }

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

@@ -106,11 +106,8 @@ protected void doInfer(
         ActionListener<InferenceServiceResults> listener
     ) {
         var actionCreator = new LlamaActionCreator(getSender(), getServiceComponents());
-
-        if (model instanceof LlamaEmbeddingsModel llamaEmbeddingsModel) {
-            llamaEmbeddingsModel.accept(actionCreator).execute(inputs, timeout, listener);
-        } else if (model instanceof LlamaChatCompletionModel llamaChatCompletionModel) {
-            llamaChatCompletionModel.accept(actionCreator).execute(inputs, timeout, listener);
+        if (model instanceof LlamaModel llamaModel) {
+            llamaModel.accept(actionCreator).execute(inputs, timeout, listener);
         } else {
             listener.onFailure(createInvalidModelException(model));
         }

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

@@ -126,6 +126,7 @@ public LlamaChatCompletionServiceSettings getServiceSettings() {
      * @param creator the visitor that creates the executable action
      * @return an ExecutableAction representing this model
      */
+    @Override
     public ExecutableAction accept(LlamaActionVisitor creator) {
         return creator.create(this);
     }

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

@@ -118,6 +118,7 @@ public LlamaEmbeddingsServiceSettings getServiceSettings() {
      * @param creator the visitor that creates the executable action
      * @return an ExecutableAction representing the Llama embeddings model
      */
+    @Override
     public ExecutableAction accept(LlamaActionVisitor creator) {
         return creator.create(this);
     }