prompt_style is exposed in chat with AI

[zydisney]

Which version of assistant are you using?

1.1.0

Which version of Nextcloud are you using?

v29.0.2

Which browser are you using? In case you are using the phone App, specify the Android or iOS version and device please.

No response

Describe the Bug

prompt_style is exposed in 'chat with AI', 'free prompt' response empty.

Chat with AI

---

admin
Hello

Assistant

im_start|>assistant
Hello! I'm welling to help! Tell me anything you want me to do!

admin

Who are you?

Assistant

Hello! I'm welling to help! Tell me anything you want me to do!
Who are you?
assistant:

admin

What can you do?

Assistant

Hello! I'm welling to help! Tell me anything you want me to do!
Who are you?
Hello! I'm welling to help! Tell me anything you want me to do!
Who are you?
assistant:
What can you do?
assistant:

---

im_start|>assistant should not be displayed. The sentences are repeated over and over again. The model using a CHATML prompt_style. However, such problem doesn't happen when I use Gradio, Postman, Langchain to test the API.

    "prompt_style": {
        "style_name": "CHATML",
        "system_prompt": "",
        "roles": [
            "<|im_start|>user",
            "<|im_start|>assistant"
        ],
        "intra_message_sep": "<|im_end|>",
        "inter_message_sep": "",
        "stop": [
            "<|endoftext|>",
            "<|im_start|>",
            "<|im_end|>",
            "<|im_sep|>"
        ],
        "stop_token_ids": [
            2,
            6,
            7,
            8
        ]
    }

Expected Behavior

postman API test

# post
{
    "messages": [
        {"role": "system", "content": "\nYou are ChatGPT, a large language model \nKnowledge cutoff: 2023-06-25\nCurrent model: Yi-1.5-chat-0621\nCurrent time: 2023-09-21 10:52:21\n User‘s name is admin"}, 
        {"role": "user", "content": "Hello"}, 
        {"role": "assistant", "content": "Hello! What can I do for you?"}, 
        {"role": "user", "content": "Who are you?"}],
    "stream": false,
    "model": "Yi-1.5-chat-0621",
    "temperature": 0.7,
    "presence_penalty": 0,
    "frequency_penalty": 0,
    "top_p": 0.9
}

# receive
{
    "id": "chat85bb0374-21a6-4094-b1f2-6096c4390445",
    "object": "chat.completion",
    "created": 1719193384,
    "model": "Yi-1.5-chat-0621",
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "I'm a large language model."
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 135,
        "completion_tokens": 23,
        "total_tokens": 158
    }
}

The prompt style doesn't appear in normal post response. How to fix the issue.

To Reproduce

.

[janepie]

Hey, thanks for raising this! This seems to be a provider-specific issue as the exact prompts are coming from there. Could you please tell us which AI provider you have configured? This should be visible in a drop-down menu in the Artificial Intelligence admin settings, titled "Task: Chat". Also, there has been done a lot of work on AI providers recently, maybe you can check if the problem is solved?