app.py

import gradio as gr
import requests
import json
import os
from dotenv import load_dotenv

load_dotenv()

API_URL = os.getenv("API_URL")
API_TOKEN = os.getenv("API_TOKEN")

if not API_URL or not API_TOKEN:
    raise ValueError("invalid API_URL | API_TOKEN")

print(f"[INFO] starting:")
print(f"[INFO] API_URL: {API_URL[:6]}...{API_URL[-12:]}")
print(f"[INFO] API_TOKEN: {API_TOKEN[:10]}...{API_TOKEN[-10:]}")

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""

def respond(
    message,
    history: list[dict],
    system_message,
    max_tokens,
    temperature,
    top_p,
    with_think,
):
    messages = [{"role": "system", "content": system_message}]

    messages.extend(history)
    
    if with_think:
        message = message + " /think"
    else:
        message = message + " /no_think"
    
    messages.append({"role": "user", "content": message})

    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {API_TOKEN}"
    }

    # print(f"[INFO] process user msg...")
    # print(f"[INFO] userMsg: {message}")

    data = {
        "model": "/data/DMind-1-mini",
        "stream": True,
        "messages": messages,
        "temperature": temperature,
        "top_p": top_p,
        "top_k": 20,
        "min_p": 0.1,
        "max_tokens": 16384
    }

    try:
        with requests.post(API_URL, headers=headers, json=data, stream=True) as r:
            if r.status_code == 200:
                current_response = ""
                for line in r.iter_lines():
                    if line:
                        line = line.decode('utf-8')
                        if line.startswith('data: '):
                            try:
                                json_response = json.loads(line[6:])
                                if 'choices' in json_response and len(json_response['choices']) > 0:
                                    delta = json_response['choices'][0].get('delta', {})
                                    if 'content' in delta:
                                        content = delta['content']
                                        if content:
                                            current_response += content
                                            
                                            if len(current_response) > 21:
                                                if with_think:
                                                    if '<think>' in current_response:
                                                        current_response = current_response.replace('<think>', '<details open><summary>Thinking</summary>\n\n```')
                                                    if '</think>' in current_response:
                                                        current_response = current_response.replace('</think>', '```\n\n</details>')
                                                    if '**Final Answer**' in current_response:
                                                        current_response = current_response.replace('**Final Answer**', '')
                                                    
                                                    formatted_response = current_response[:-16]
                                                    
                                                    formatted_response = formatted_response.replace('<', '&lt;').replace('>', '&gt;')
                                                    formatted_response = formatted_response.replace('&lt;details open&gt;', '<details open>')
                                                    formatted_response = formatted_response.replace('&lt;/details&gt;', '</details>')
                                                    formatted_response = formatted_response.replace('&lt;summary&gt;', '<summary>')
                                                    formatted_response = formatted_response.replace('&lt;/summary&gt;', '</summary>')
                                                    formatted_response = formatted_response.replace('*', '\\*')
                                                    yield formatted_response
                                                else:
                                                    if '<think>' in current_response and '</think>\n' in current_response:
                                                        start = current_response.find('<think>')
                                                        end = current_response.find('</think>\n') + len('</think>\n')
                                                        current_response = current_response[:start] + current_response[end:]
                                                    yield current_response
                            except json.JSONDecodeError:
                                continue
                if current_response:
                    # current_response = current_response.replace('<details open>', '<details>')
                    if '**Final Answer**' in current_response:
                        current_response = current_response.replace('**Final Answer**', '')
                    
                    formatted_response = current_response
                    formatted_response = formatted_response.replace('<', '&lt;').replace('>', '&gt;')
                    formatted_response = formatted_response.replace('&lt;details open&gt;', '<details open>')
                    # formatted_response = formatted_response.replace('&lt;details&gt;', '<details>')
                    formatted_response = formatted_response.replace('&lt;/details&gt;', '</details>')
                    formatted_response = formatted_response.replace('&lt;summary&gt;', '<summary>')
                    formatted_response = formatted_response.replace('&lt;/summary&gt;', '</summary>')
                    formatted_response = formatted_response.replace('*', '\\*')
                    yield formatted_response
            else:
                print(f"[ERROR] Bad status code: {r.status_code}, response: {r.text}")
                yield "Service temporarily unavailable"
    except Exception as e:
        print(f"[ERROR] Request error: {e}")
        yield "Service error occurred"


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are DMind AI Assistant, built by DMind.AI. Never present as Qwen, ChatGPT, Tongyi, OpenAI, etc.", label="System message"),
        gr.Slider(minimum=1, maximum=32768, value=16384, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.6, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
        gr.Checkbox(value=False, label="With Think"),
    ],
    type="messages",
    css="""
    .prose pre {
        white-space: pre-wrap !important;
        word-wrap: break-word !important;
        overflow-wrap: break-word !important;
        max-width: 100% !important;
        margin-bottom: 1.5em !important;
    }
    .prose code {
        white-space: pre-wrap !important;
        word-wrap: break-word !important;
        overflow-wrap: break-word !important;
        max-width: 100% !important;
    }
    .prose pre code {
        white-space: pre-wrap !important;
        word-wrap: break-word !important;
        overflow-wrap: break-word !important;
        max-width: 100% !important;
    }
    .accordion {
        margin: 0 !important;
        border: none !important;
    }
    .accordion-header {
        background: #f0f0f0 !important;
        padding: 8px !important;
        cursor: pointer !important;
    }
    .accordion-content {
        padding: 8px !important;
    }
    """
)


if __name__ == "__main__":
    demo.launch()