app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("jhangmez/CHATPRG-v1.2-Phi-3.5-mini-instruct-GGUF")

PLACEHOLDER = """
<div style="padding: 50px; text-align: center; display: flex; flex-direction: column; align-items: center;">
   <img src="https://avatars.githubusercontent.com/u/60937214?v=4" style="width: 80%; max-width: 550px; height: auto; opacity: 0.80;  ">
   <h1 style="font-size: 36px; margin-bottom: 15px; opacity: 0.85; font-family: 'Source Sans Pro', sans-serif;">CHATPRG</h1>
</div>
"""

custom_css = """
@import url('https://fonts.googleapis.com/css2?family=Source+Sans+Pro:wght@300;400;700&display=swap');

body, .gradio-container, .gr-button, .gr-input, .gr-slider, .gr-dropdown, .gr-markdown {
    font-family: 'Source Sans Pro', sans-serif !important;
    margin: 0;
    padding: 0;
    box-sizing: border-box;
    font-size: 18px;
}
._button {
    font-size: 20px;
    padding: 12px 24px;
}
pre, code {
    direction: ltr !important;
    unicode-bidi: plaintext !important;
    font-size: 16px;
}
"""

def format_message(role, content):
    return f"<|im_start|>{role} {content}<|im_end|> "

def respond(message, history, system_message, max_tokens, temperature, top_p):
    formatted_messages = [format_message("system", system_message)]

    for human, assistant in history:
        if human:
            formatted_messages.append(format_message("user", human))
        if assistant:
            formatted_messages.append(format_message("assistant", assistant))

    formatted_messages.append(format_message("user", message))
    formatted_messages.append("<|im_start|>assistant ")

    full_prompt = "".join(formatted_messages)

    response = ""
    for chunk in client.text_generation(
        full_prompt,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        stream=True,
    ):
        response += chunk
        yield response.strip()

chatbot = gr.Chatbot(placeholder=PLACEHOLDER, scale=1, show_copy_button=True, height="68%", rtl=False,  elem_classes=["chatbot"])
chat_input = gr.Textbox(show_label=True, rtl=False, placeholder="Input", show_copy_button=True, scale=5)
submit_btn = gr.Button(variant="primary", value="Submit", size="sm", scale=1, elem_classes=["_button"])


demo = gr.ChatInterface(
    respond,
    additional_inputs_accordion=gr.Accordion(label="Configuration", open=False),
    additional_inputs=[
        gr.Textbox(value="You are a helpful AI assistant.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=1.5, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
    ],
    stop_btn="Stop",
    chatbot=chatbot,
    textbox=chat_input,
    retry_btn="🔄 Reintentar",
    submit_btn=submit_btn,
    undo_btn="↩️ Retroceder",
    clear_btn="🗑️ Borrar",
    title="CHATPRG v1.2 Chatbot",
    description="Modelo pre-entrenado para dar a conocer a estudiantes y personas externas, los reglamentos de la Universidad nacional Pedro Ruiz Gallo de Lambayeque, Perú",
)

if __name__ == "__main__":
    demo.launch(debug=True, share=True, inline=False)