fix: auto-normalize array of strings in message content to openai multi-modal spec

litellm/utils.py

@@ -7941,6 +7941,16 @@ def validate_and_fix_openai_messages(messages: List):
         if message.get("tool_calls"):
             message["tool_calls"] = jsonify_tools(tools=message["tool_calls"])
 
+        content = message.get("content")
+        if isinstance(content, list):
+            normalized_content = []
+            for item in content:
+                if isinstance(item, str):
+                    normalized_content.append({"type": "text", "text": item})
+                else:
+                    normalized_content.append(item)
+            message["content"] = normalized_content
+
         convert_msg_to_dict = cast(AllMessageValues, convert_to_dict(message))
         cleaned_message = cleanup_none_field_in_message(message=convert_msg_to_dict)
         new_messages.append(cleaned_message)

tests/test_litellm/test_utils.py

@@ -4107,3 +4107,18 @@ class TestValidateAndFixThinkingParam:
         validate_and_fix_thinking_param(thinking=thinking)
         assert "budgetTokens" in thinking
         assert "budget_tokens" not in thinking
+
+
+def test_normalize_array_of_strings_in_content():
+    from litellm.utils import validate_and_fix_openai_messages
+
+    messages = [
+        {
+            "role": "user",
+            "content": ["what is the capital of France?"],
+        }
+    ]
+    result = validate_and_fix_openai_messages(messages=messages)
+    assert result[0]["content"] == [
+        {"type": "text", "text": "what is the capital of France?"},
+    ]