fix(litellm): guard against Content with no parts in _content_to_message_param

types.Content(role="user") defaults parts to None, and the fallback path can
produce an empty user turn, so iterating content.parts in _content_to_message_param
raised TypeError: 'NoneType' object is not iterable during LiteLLM request
assembly, before any model call.

Normalize parts to content.parts or [] once (matching the existing idiom in this
module and the google_llm adapter, which skips contents without parts) and iterate
the normalized list in all three branches. A content with no parts now yields a
message with empty content instead of crashing.

Add regression tests covering user and assistant roles for both parts=None and
parts=[].

Author: anxkhn

src/google/adk/models/lite_llm.py

@@ -922,9 +922,15 @@ async def _content_to_message_param(
   """
   _ensure_litellm_imported()
 
+  # A types.Content may have parts=None (e.g. types.Content(role="user")) or an
+  # empty list (e.g. from _append_fallback_user_content_if_missing). Normalize to
+  # an iterable so the loops below do not raise, matching the google_llm adapter
+  # which skips contents without parts.
+  parts = content.parts or []
+
   tool_messages: list[Message] = []
   non_tool_parts: list[types.Part] = []
-  for part in content.parts:
+  for part in parts:
     if part.function_response:
       response = part.function_response.response
       response_content = (
@@ -965,7 +971,7 @@ async def _content_to_message_param(
   role = _to_litellm_role(content.role)
 
   if role == "user":
-    user_parts = [part for part in content.parts if not part.thought]
+    user_parts = [part for part in parts if not part.thought]
     message_content = (
         await _get_content(user_parts, provider=provider, model=model) or None
     )
@@ -974,7 +980,7 @@ async def _content_to_message_param(
     tool_calls = []
     content_parts: list[types.Part] = []
     reasoning_parts: list[types.Part] = []
-    for part in content.parts:
+    for part in parts:
       if part.function_call:
         tool_call_id = part.function_call.id or ""
         tool_call_dict: ChatCompletionAssistantToolCall = {

tests/unittests/models/test_litellm.py

@@ -1928,6 +1928,26 @@ async def test_content_to_message_param_user_message():
   assert message["content"] == "Test prompt"
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize("parts", [None, []])
+async def test_content_to_message_param_user_message_without_parts(parts):
+  # types.Content(role="user") defaults parts to None, so a content with no
+  # parts must not raise (mirrors the google_llm adapter which skips it).
+  content = types.Content(role="user", parts=parts)
+  message = await _content_to_message_param(content)
+  assert message["role"] == "user"
+  assert message["content"] is None
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("parts", [None, []])
+async def test_content_to_message_param_assistant_message_without_parts(parts):
+  content = types.Content(role="assistant", parts=parts)
+  message = await _content_to_message_param(content)
+  assert message["role"] == "assistant"
+  assert message["content"] is None
+
+
 @pytest.mark.asyncio
 @pytest.mark.parametrize("file_uri,mime_type", FILE_URI_TEST_CASES)
 async def test_content_to_message_param_user_message_with_file_uri(