testing image format fix

Author: dixitaniket

tee_gateway/controllers/chat_controller.py

@@ -911,9 +911,7 @@ def _chat_request_to_dict(chat_request: CreateChatCompletionRequest) -> dict:
             messages.append(
                 {
                     "role": "user",
-                    "content": msg.content
-                    if isinstance(msg.content, str)
-                    else str(msg.content),
+                    "content": msg.content,
                 }
             )
         elif isinstance(msg, ChatCompletionRequestAssistantMessage):

tee_gateway/llm_backend.py

@@ -325,6 +325,17 @@ def generate_images(model: str, prompt: str, n: int = 1) -> tuple[list[str], int
     return images, len(images)
 
 
+def _normalize_user_content_parts(content: list) -> list:
+    """Preserve multimodal user content while tolerating primitive text parts."""
+    normalized = []
+    for part in content:
+        if isinstance(part, dict):
+            normalized.append(part)
+        else:
+            normalized.append({"type": "text", "text": str(part)})
+    return normalized
+
+
 def convert_messages(messages: list) -> List[Any]:
     """Convert OpenAI-format message objects or dicts to LangChain message objects."""
     langchain_messages: List[BaseMessage] = []
@@ -348,12 +359,8 @@ def convert_messages(messages: list) -> List[Any]:
             langchain_messages.append(SystemMessage(content=content))
 
         elif role == "user":
-            # content may be a string or a list of content parts; handle both
             if isinstance(content, list):
-                content = "".join(
-                    part.get("text", "") if isinstance(part, dict) else str(part)
-                    for part in content
-                )
+                content = _normalize_user_content_parts(content)
             langchain_messages.append(HumanMessage(content=content))
 
         elif role == "assistant":

tee_gateway/test/test_tee_core.py

@@ -569,20 +569,24 @@ def test_multi_turn_order_preserved(self):
         self.assertIsInstance(result[2], AIMessage)
 
     def test_user_content_as_list_of_parts(self):
-        """Multimodal content parts should be concatenated into a single string."""
+        """Multimodal content parts should be preserved for vision-capable models."""
+        content = [
+            {"type": "text", "text": "Hello world"},
+            {
+                "type": "image_url",
+                "image_url": {"url": "data:image/png;base64,abcd"},
+            },
+        ]
         result = convert_messages(
             [
                 {
                     "role": "user",
-                    "content": [
-                        {"type": "text", "text": "Hello "},
-                        {"type": "text", "text": "world"},
-                    ],
+                    "content": content,
                 }
             ]
         )
         self.assertIsInstance(result[0], HumanMessage)
-        self.assertEqual(result[0].content, "Hello world")
+        self.assertEqual(result[0].content, content)
 
     def test_full_tool_call_conversation(self):
         """End-to-end multi-turn with tool use: user → assistant (tool call) → tool result."""