fix(gemini): update test_message_with_empty_content to match filtering behavior

Update test expectations to align with the implementation that filters out
assistant messages with empty parts. Empty content assistant messages are
now correctly expected to be removed rather than kept with empty parts lists.

Author: WH-2099

models/gemini/models/tests/test_llm.py

@@ -614,10 +614,10 @@ def test_empty_message_list(self):
     def test_message_with_empty_content(self):
         """Test handling of messages with empty content
 
-        In Gemini message history, both empty parts list and empty string text parts
-        for model messages are valid and won't cause communication errors.
+        Assistant messages with empty parts are filtered out to avoid invalid requests.
+        This ensures clean message history for Gemini API calls.
         """
-        # Test case 1: Assistant message with empty string content
+        # Test case 1: Assistant message with empty string content is filtered out
         messages = [
             AssistantPromptMessage(content=""),
             UserPromptMessage(content="Hello"),
@@ -629,14 +629,12 @@ def test_message_with_empty_content(self):
             config=self.mock_config,
         )
 
-        # Empty string content creates an empty parts list, which is valid for Gemini
-        assert len(contents) == 2
-        assert contents[0].role == "model"
-        assert contents[0].parts == []  # Empty parts list is valid
-        assert contents[1].role == "user"
-        assert contents[1].parts[0].text == "Hello"
+        # Empty assistant message is filtered out, only user message remains
+        assert len(contents) == 1
+        assert contents[0].role == "user"
+        assert contents[0].parts[0].text == "Hello"
 
-        # Test case 2: Assistant message with only thinking tags (results in empty content after removal)
+        # Test case 2: Assistant message with only thinking tags is filtered out
         messages = [
             AssistantPromptMessage(content="<think>Internal thoughts only</think>"),
             UserPromptMessage(content="Hi"),
@@ -648,14 +646,12 @@ def test_message_with_empty_content(self):
             config=self.mock_config,
         )
 
-        # After removing thinking tags, content is empty - creates empty parts list
-        assert len(contents) == 2
-        assert contents[0].role == "model"
-        assert contents[0].parts == []  # Empty parts list after thinking tag removal
-        assert contents[1].role == "user"
-        assert contents[1].parts[0].text == "Hi"
+        # After removing thinking tags, content is empty and filtered out
+        assert len(contents) == 1
+        assert contents[0].role == "user"
+        assert contents[0].parts[0].text == "Hi"
 
-        # Test case 3: Mixed empty and non-empty assistant messages
+        # Test case 3: Empty assistant message is filtered, non-empty one is kept
         messages = [
             UserPromptMessage(content="Question"),
             AssistantPromptMessage(content=""),
@@ -669,14 +665,12 @@ def test_message_with_empty_content(self):
             config=self.mock_config,
         )
 
-        # Should have 3 contents - consecutive assistant messages are merged
+        # Empty assistant message filtered out, remaining messages properly ordered
         assert len(contents) == 3
         assert contents[0].role == "user"
         assert contents[0].parts[0].text == "Question"
         assert contents[1].role == "model"
-        assert (
-            len(contents[1].parts) == 1
-        )  # Empty content doesn't add parts, only "Actual response" does
+        assert len(contents[1].parts) == 1
         assert contents[1].parts[0].text == "Actual response"
         assert contents[2].role == "user"
         assert contents[2].parts[0].text == "Thanks"