Some qwen llm can not invoke function-call

[Dofine-dufei]

Self Checks

• [X] This is only for bug report, if you would like to ask a question, please head to Discussions.
• [X] I have searched for existing issues search for existing issues, including closed ones.
• [X] I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [X] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• [X] Please do not modify this template :) and fill in all the required fields.

Dify version

0.6.14

Cloud or Self Hosted

Self Hosted (Source)

Steps to reproduce

on 2024.07.23，I can not invoke fc with some qwen llm，like qwen-max... But actually,it work last week.

✔️ Expected Behavior

I hope it invoke fc correctly

❌ Actual Behavior

However, qwen-max can not invoke correctly

[Dofine-dufei]

LOG data: {"event": "error", "conversation_id": "5d74e70a-bcf9-4a76-bc6f-ad91db9b9b2f", "message_id": "aeccf4a0-399f-499b-a1ea-49c7afceb939", "created_at": 1721690363, "code": "completion_request_error", "status": 400, "message": "[tongyi] Server Unavailable Error, Failed to invoke model qwen-max, status code: 500, message: messages with role \"tool\" must be a response to a preceeding message with \"tool_calls\"."}

[dosubot[bot]]

To resolve the issue with invoking function calls using the qwen-max model in your self-hosted Dify version 0.6.14, you need to ensure that messages with the role 'tool' are correctly preceded by messages with 'tool_calls'. This is a requirement for the model to process tool calls correctly.

From the provided code, it appears that the _convert_prompt_messages_to_tongyi_messages method is responsible for converting prompt messages to the format expected by the model. Ensure that the tool calls are correctly formatted and included in the prompt messages.

Here is a snippet of the relevant code that handles tool calls:

def _convert_prompt_messages_to_tongyi_messages(self, prompt_messages: list[PromptMessage],
                                                rich_content: bool = False) -> list[dict]:
    """
    Convert prompt messages to tongyi messages
    :param prompt_messages: prompt messages
    :return: tongyi messages
    """
    tongyi_messages = []
    for prompt_message in prompt_messages:
        if isinstance(prompt_message, SystemPromptMessage):
            tongyi_messages.append({
                'role': 'system',
                'content': prompt_message.content if not rich_content else [{"text": prompt_message.content}],
            })
        elif isinstance(prompt_message, UserPromptMessage):
            if isinstance(prompt_message.content, str):
                tongyi_messages.append({
                    'role': 'user',
                    'content': prompt_message.content if not rich_content else [{"text": prompt_message.content}],
                })
            else:
                sub_messages = []
                for message_content in prompt_message.content:
                    if message_content.type == PromptMessageContentType.TEXT:
                        message_content = cast(TextPromptMessageContent, message_content)
                        sub_message_dict = {
                            "text": message_content.data
                        }
                        sub_messages.append(sub_message_dict)
                    elif message_content.type == PromptMessageContentType.IMAGE:
                        message_content = cast(ImagePromptMessageContent, message_content)

                        image_url = message_content.data
                        if message_content.data.startswith("data:"):
                            # convert image base64 data to file in /tmp
                            image_url = self._save_base64_image_to_file(message_content.data)

                        sub_message_dict = {
                            "image": image_url
                        }
                        sub_messages.append(sub_message_dict)
                # resort sub_messages to ensure text is always at last
                sub_messages = sorted(sub_messages, key=lambda x: 'text' in x)

                tongyi_messages.append({
                    'role': 'user',
                    'content': sub_messages
                })
        elif isinstance(prompt_message, AssistantPromptMessage):
            content = prompt_message.content
            if not content:
                content = ' '
            tongyi_messages.append({
                'role': 'assistant',
                'content': content if not rich_content else [{"text": content}],
            })
        elif isinstance(prompt_message, ToolPromptMessage):
            tongyi_messages.append({
                "role": "tool",
                "content": prompt_message.content,
                "name": prompt_message.tool_call_id
            })
        else:
            raise ValueError(f"Got unknown type {prompt_message}")

    return tongyi_messages

Additionally, ensure that the _convert_prompt_message_to_dict method correctly handles the conversion of ToolPromptMessage to the expected format:

def _convert_prompt_message_to_dict(self, message: PromptMessage) -> dict:
    """
    Convert PromptMessage to dict for OpenAI Compatibility API
    """
    if isinstance(message, UserPromptMessage):
        message = cast(UserPromptMessage, message)
        if isinstance(message.content, str):
            message_dict = {"role": "user", "content": message.content}
        else:
            raise ValueError("User message content must be str")
    elif isinstance(message, AssistantPromptMessage):
        message = cast(AssistantPromptMessage, message)
        message_dict = {"role": "assistant", "content": message.content}
        if message.tool_calls and len(message.tool_calls) > 0:
            message_dict["function_call"] = {
                "name": message.tool_calls[0].function.name,
                "arguments": message.tool_calls[0].function.arguments
            }
    elif isinstance(message, SystemPromptMessage):
        message = cast(SystemPromptMessage, message)
        message_dict = {"role": "system", "content": message.content}
    elif isinstance(message, ToolPromptMessage):
        message = cast(ToolPromptMessage, message)
        message_dict = {
            "role": "user",
            "content": [{
                "type": "tool_result",
                "tool_use_id": message.tool_call_id,
                "content": message.content
            }]
        }
    else:
        raise ValueError(f"Unknown message type {type(message)}")

    return message_dict

Ensure that your prompt messages are structured correctly, with the appropriate 'tool_calls' message preceding any 'tool' messages. This should resolve the "messages with role 'tool' must be a response to a preceding message with 'tool_calls'" error ^{[1][2][3][4][5]}.

To continue talking to Dosu, mention @dosu.

[Dofine-dufei]

I fixed it！！！

[laipz8200]

I fixed it！！！

Congrats! Are you interested in making a PR?