Allow more flexible assistant and system response (#1256)

python/sglang/srt/conversation.py

@@ -386,7 +386,16 @@ def generate_chat_conv(
     for message in request.messages:
         msg_role = message.role
         if msg_role == "system":
-            conv.system_message = message.content
+            if isinstance(message.content, str):
+                conv.system_message = message.content
+            elif isinstance(message.content, list):
+                if (
+                    len(message.content) != 1
+                    or getattr(message.content[0], "type", None) != "text"
+                ):
+                    raise ValueError("The system message should be a single text.")
+                else:
+                    conv.system_message = getattr(message.content[0], "text", "")
         elif msg_role == "user":
             # Handle the various types of Chat Request content types here.
             role = conv.roles[0]
@@ -414,7 +423,20 @@ def generate_chat_conv(
                         conv.append_image(content.image_url.url)
                 conv.append_message(conv.roles[0], real_content)
         elif msg_role == "assistant":
-            conv.append_message(conv.roles[1], message.content)
+            parsed_content = ""
+            if isinstance(message.content, str):
+                parsed_content = message.content
+            elif isinstance(message.content, list):
+                if (
+                    len(message.content) != 1
+                    or getattr(message.content[0], "type", None) != "text"
+                ):
+                    raise ValueError(
+                        "The assistant's response should be a single text."
+                    )
+                else:
+                    parsed_content = getattr(message.content[0], "text", "")
+            conv.append_message(conv.roles[1], parsed_content)
         else:
             raise ValueError(f"Unknown role: {msg_role}")
 

python/sglang/srt/openai_api/adapter.py

@@ -844,8 +844,23 @@ def v1_chat_generate_request(
         if not isinstance(request.messages, str):
             # Apply chat template and its stop strings.
             if chat_template_name is None:
+                openai_compatible_messages = []
+                for message in request.messages:
+                    if isinstance(message.content, str):
+                        openai_compatible_messages.append(
+                            {"role": message.role, "content": message.content}
+                        )
+                    else:
+                        content_list = message.dict()["content"]
+                        for content in content_list:
+                            if content["type"] == "text":
+                                openai_compatible_messages.append(
+                                    {"role": message.role, "content": content["text"]}
+                                )
                 prompt_ids = tokenizer_manager.tokenizer.apply_chat_template(
-                    request.messages, tokenize=True, add_generation_prompt=True
+                    openai_compatible_messages,
+                    tokenize=True,
+                    add_generation_prompt=True,
                 )
                 stop = request.stop
                 image_data = None

python/sglang/srt/openai_api/protocol.py

@@ -200,11 +200,6 @@ class CompletionStreamResponse(BaseModel):
     usage: Optional[UsageInfo] = None
 
 
-class ChatCompletionMessageGenericParam(BaseModel):
-    role: Literal["system", "assistant"]
-    content: str
-
-
 class ChatCompletionMessageContentTextPart(BaseModel):
     type: Literal["text"]
     text: str
@@ -225,6 +220,11 @@ ChatCompletionMessageContentPart = Union[
 ]
 
 
+class ChatCompletionMessageGenericParam(BaseModel):
+    role: Literal["system", "assistant"]
+    content: Union[str, List[ChatCompletionMessageContentTextPart]]
+
+
 class ChatCompletionMessageUserParam(BaseModel):
     role: Literal["user"]
     content: Union[str, List[ChatCompletionMessageContentPart]]