use previewFeatures to determine which pro model to use for A2A (#15131)

packages/a2a-server/src/config/config.ts

@@ -22,6 +22,7 @@ import {
   DEFAULT_GEMINI_MODEL,
   type ExtensionLoader,
   startupProfiler,
+  PREVIEW_GEMINI_MODEL,
 } from '@google/gemini-cli-core';
 
 import { logger } from '../utils/logger.js';
@@ -38,7 +39,9 @@ export async function loadConfig(
 
   const configParams: ConfigParameters = {
     sessionId: taskId,
-    model: DEFAULT_GEMINI_MODEL,
+    model: settings.general?.previewFeatures
+      ? PREVIEW_GEMINI_MODEL
+      : DEFAULT_GEMINI_MODEL,
     embeddingModel: DEFAULT_GEMINI_EMBEDDING_MODEL,
     sandbox: undefined, // Sandbox might not be relevant for a server-side agent
     targetDir: workspaceDir, // Or a specific directory the agent operates on