langgenius/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-07-21 20:15:31 -04:00
Code Issues 775 Packages Projects Releases 10 Wiki Activity

Some qwen llm can not invoke function-call #4731

New Issue
Closed
opened 2026-02-21 18:07:46 -05:00 by yindo · 4 comments
yindo commented 2026-02-21 18:07:46 -05:00
Owner
Copy Link
Copy Source

Originally created by @Dofine-dufei on GitHub (Jul 23, 2024).

Originally assigned to: @laipz8200, @Dofine-dufei on GitHub.

Self Checks

  • This is only for bug report, if you would like to ask a question, please head to Discussions.
  • I have searched for existing issues search for existing issues, including closed ones.
  • I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
  • 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
  • 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.
6fc774a5c3857d1cd156496fc140f6f

✔️ Expected Behavior

I hope it invoke fc correctly
d5d4eed8762fc1bd2d1f7ab531f2338

❌ Actual Behavior

However, qwen-max can not invoke correctly

Originally created by @Dofine-dufei on GitHub (Jul 23, 2024). Originally assigned to: @laipz8200, @Dofine-dufei on GitHub. ### Self Checks - [X] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [X] I have searched for existing issues [search for existing issues](https://github.com/langgenius/dify/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](https://github.com/langgenius/dify/issues/1542)). - [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. ![6fc774a5c3857d1cd156496fc140f6f](https://github.com/user-attachments/assets/2bf81fc3-5f86-4213-8056-871bb4ecaa2a) ### ✔️ Expected Behavior I hope it invoke fc correctly ![d5d4eed8762fc1bd2d1f7ab531f2338](https://github.com/user-attachments/assets/60050782-4caf-4720-bcee-b4df83b7567b) ### ❌ Actual Behavior However, qwen-max can not invoke correctly
yindo added the 🐞 bug label 2026-02-21 18:07:46 -05:00
yindo closed this issue 2026-02-21 18:07:46 -05:00
yindo commented 2026-02-21 18:07:46 -05:00
Author
Owner
Copy Link
Copy Source

@Dofine-dufei commented on GitHub (Jul 23, 2024):

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\"."}

@Dofine-dufei commented on GitHub (Jul 23, 2024): **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\"."} `
yindo commented 2026-02-21 18:07:46 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Jul 23, 2024):

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.

@dosubot[bot] commented on GitHub (Jul 23, 2024): <!-- Answer --> 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: ```python 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: ```python 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 <sup>[[1]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/tongyi/llm/llm.py)</sup><sup>[[2]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/localai/llm/llm.py)</sup><sup>[[3]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/zhipuai/llm/llm.py)</sup><sup>[[4]](https://github.com/langgenius/dify/blob/main/api/core/tools/tool_engine.py)</sup><sup>[[5]](https://github.com/langgenius/dify/blob/main/api/core/model_runtime/model_providers/azure_openai/llm/llm.py)</sup>. <!-- Dosu Comment Footer --> *To continue talking to [Dosu](https://dosu.dev), mention @dosu.*
yindo commented 2026-02-21 18:07:46 -05:00
Author
Owner
Copy Link
Copy Source

@Dofine-dufei commented on GitHub (Jul 24, 2024):

I fixed it！！！

@Dofine-dufei commented on GitHub (Jul 24, 2024): I fixed it！！！
yindo commented 2026-02-21 18:07:46 -05:00
Author
Owner
Copy Link
Copy Source

@laipz8200 commented on GitHub (Jul 24, 2024):

I fixed it！！！

Congrats! Are you interested in making a PR?

@laipz8200 commented on GitHub (Jul 24, 2024): > I fixed it！！！ Congrats! Are you interested in making a PR?
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Knowledge Pipeline
Proposal
WIP
cloud
dependencies
feat:webapp
good first issue
outdated
performance
pull-request
Mirrored from GitHub Pull Request
 python:uv
refactor
size:L
stale
status: accepting prs
status: blocked
status: in discussion
stt & tts
vulnerablity
web
wont-fix
🌊 feat:workflow
🌚 invalid
🐞 bug
👻 feat:rag
💪 enhancement
📚 documentation
🔨 feat:tools
🙋‍♂️ question
🙏 help wanted
🤔 cant-reproduce
🤖 feat:agent
No Label 🐞 bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#4731
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?