From b6cba411baafab408b571b55f1389b3291ce4395 Mon Sep 17 00:00:00 2001
From: Praveen11558 <44603409+Praveen11558@users.noreply.github.com>
Date: Sun, 21 Jun 2026 16:58:31 +0530
Subject: [PATCH] fix: Refactor HTTP handler initialization with client support

---
 .../vertex_gemma_models/transformation.py     | 29 +++++++++++++++----
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py b/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
index 35cd54d65f6..b0c7c4ead04 100644
--- a/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
+++ b/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
@@ -158,6 +158,7 @@ class VertexGemmaConfig(OpenAIGPTConfig):
                 logging_obj=logging_obj,
                 optional_params=optional_params,
                 litellm_params=litellm_params,
+                client= client,
                 timeout=timeout,
                 encoding=encoding,
             )
@@ -172,6 +173,7 @@ class VertexGemmaConfig(OpenAIGPTConfig):
                 logging_obj=logging_obj,
                 optional_params=optional_params,
                 litellm_params=litellm_params,
+                client= client,
                 timeout=timeout,
                 encoding=encoding,
             )
@@ -187,11 +189,12 @@ class VertexGemmaConfig(OpenAIGPTConfig):
         logging_obj: Any,
         optional_params: dict,
         litellm_params: dict,
+        client= Optional[Any],
         timeout: Optional[Union[float, httpx.Timeout]],
         encoding: Any,
     ):
         """Synchronous completion request"""
-        from litellm.llms.custom_httpx.http_handler import HTTPHandler
+        from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler
         from litellm.utils import convert_to_model_response_object
 
         # Check if streaming is requested (will be faked)
@@ -223,7 +226,12 @@ class VertexGemmaConfig(OpenAIGPTConfig):
         )
 
         # Make the HTTP request
-        http_handler = HTTPHandler(concurrent_limit=1)
+        if client is None or isinstance(client, (httpx.AsyncClient, AsyncHTTPHandler)):
+            http_handler = HTTPHandler(concurrent_limit=1)
+        elif isinstance(client, httpx.Client):
+            http_handler = HTTPHandler(client=client)
+        else:
+            http_handler = client
         response = http_handler.post(
             url=api_base,
             headers=headers,
@@ -279,11 +287,16 @@ class VertexGemmaConfig(OpenAIGPTConfig):
         logging_obj: Any,
         optional_params: dict,
         litellm_params: dict,
+        client: Optional[Any],
         timeout: Optional[Union[float, httpx.Timeout]],
         encoding: Any,
     ):
         """Asynchronous completion request"""
-        from litellm.llms.custom_httpx.http_handler import get_async_httpx_client
+        from litellm.llms.custom_httpx.http_handler import (
+            AsyncHTTPHandler,
+            HTTPHandler,
+            get_async_httpx_client,
+        )
         from litellm.types.utils import LlmProviders
         from litellm.utils import convert_to_model_response_object
 
@@ -316,9 +329,13 @@ class VertexGemmaConfig(OpenAIGPTConfig):
         )
 
         # Make the HTTP request
-        http_handler = get_async_httpx_client(
-            llm_provider=LlmProviders.VERTEX_AI,
-        )
+        if client is None or isinstance(client, (httpx.Client, HTTPHandler)):
+            http_handler = get_async_httpx_client(
+                llm_provider=LlmProviders.VERTEX_AI,
+            )
+        else:
+            http_handler = client
+            
         response = await http_handler.post(
             url=api_base,
             headers=headers,