update vertex2

Author: ionmincu

src/uipath_llamaindex/llms/vertex.py

@@ -319,14 +319,59 @@ async def gen() -> CompletionResponseAsyncGen:
 
         return gen()
 
+    def _create_sync_client(self) -> google.genai.Client:
+        """Create a fresh google.genai client for sync operations."""
+        uipath_url = self._uipath_url
+        headers = self._build_headers_static(self._uipath_token)
+
+        http_options = genai_types.HttpOptions(
+            httpx_client=httpx.Client(
+                transport=_UrlRewriteTransport(uipath_url),
+                headers=headers,
+                follow_redirects=True,
+            ),
+        )
+
+        return google.genai.Client(
+            api_key="uipath-gateway",
+            http_options=http_options,
+        )
+
     @llm_chat_callback()
     def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
-        """Chat endpoint - delegates to parent's async chat via sync wrapper."""
-        import asyncio
+        """Chat endpoint using sync client."""
+        from llama_index.llms.google_genai.utils import chat_from_gemini_response
 
-        return asyncio.get_event_loop().run_until_complete(
-            self.achat(messages, **kwargs)
-        )
+        generation_config = {
+            **(self._generation_config or {}),
+            **kwargs.pop("generation_config", {}),
+        }
+
+        # Create a fresh sync client to avoid event loop issues
+        sync_client = self._create_sync_client()
+
+        try:
+            # Convert messages to Gemini format (sync version - no file uploads)
+            contents = []
+            for message in messages:
+                content = genai_types.Content(
+                    role="user" if message.role.value == "user" else "model",
+                    parts=[genai_types.Part(text=message.content or "")],
+                )
+                contents.append(content)
+
+            # Use sync client to send message
+            chat = sync_client.chats.create(
+                model=self.model,
+                config=generation_config,
+                history=contents[:-1] if len(contents) > 1 else None,
+            )
+            response = chat.send_message(contents[-1].parts if contents else [])
+
+            return chat_from_gemini_response(response, [])
+        finally:
+            # Clean up the sync client
+            pass
 
     @llm_chat_callback()
     def stream_chat(