fix(vertex_and_google_ai_studio_gemini.py): bubble up stream is not finished, even if stop reason is given

prevents early completion of stream

Closes https://github.com/BerriAI/litellm/issues/11549

litellm/litellm_core_utils/streaming_handler.py

@@ -439,7 +439,13 @@ class CustomStreamWrapper:
                 else:  # function/tool calling chunk - when content is None. in this case we just return the original chunk from openai
                     pass
                 if str_line.choices[0].finish_reason:
-                    is_finished = True
+                    is_finished = (
+                        True  # check if str_line._hidden_params["is_finished"] is True
+                    )
+                    if hasattr(
+                        str_line, "_hidden_params"
+                    ) and str_line._hidden_params.get("is_finished"):
+                        is_finished = str_line._hidden_params.get("is_finished")
                     finish_reason = str_line.choices[0].finish_reason
 
                 # checking for logprobs

litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py

@@ -1886,6 +1886,8 @@ class ModelResponseIterator:
                     ).web_search_requests = web_search_requests
 
             setattr(model_response, "usage", usage)  # type: ignore
+
+            model_response._hidden_params["is_finished"] = False
             return model_response
 
         except json.JSONDecodeError:

litellm/proxy/_new_secret_config.yaml

@@ -1,9 +1,4 @@
 model_list:
-  - model_name: "gemini-2.0-flash"
-    litellm_params:
-      model: vertex_ai/gemini-2.0-flash
-      vertex_project: my-project-id
-      vertex_location: us-central1
   - model_name: "gpt-4o-mini-openai"
     litellm_params:
       model: gpt-4o-mini
@@ -42,9 +37,9 @@ model_list:
     litellm_params:
       model: text-embedding-ada-002
       api_key: os.environ/OPENAI_API_KEY
-  - model_name: gemini/gemini-2.0-flash
+  - model_name: gemini/*
     litellm_params:
-      model: gemini/gemini-2.0-flash
+      model: gemini/*
   - model_name: llama-qwen
     litellm_params:
       model: ollama/qwen2:0.5b

tests/local_testing/test_amazing_vertex_completion.py

@@ -3942,3 +3942,20 @@ def test_vertex_ai_streaming_response_id():
         iterator = iter(iterator)
         first_chunk = next(iterator)
         assert first_chunk.id == "vertex_ai_response_stream_123"
+
+
+def test_vertex_ai_gemini_2_5_pro_streaming():
+    load_vertex_ai_credentials()
+    litellm._turn_on_debug()
+    response = completion(
+        model="vertex_ai/gemini-2.5-pro-preview-06-05",
+        messages=[{"role": "user", "content": "Hi!"}],
+        vertex_location="global",
+        stream=True,
+    )
+    has_real_content = False
+    for chunk in response:
+        print(chunk)
+        if chunk.choices[0].delta.content is not None and len(chunk.choices[0].delta.content) > 0:
+            has_real_content = True
+    assert has_real_content