app.py

import gradio as gr
from llama_cpp import Llama

# Initialize the model
model = Llama(
    model_path="Crazy_Reasoning_1.5B.Q4_K_M.gguf",  # Replace with your model path
    n_ctx=32768,  # Context window
    n_threads=2  # Number of CPU threads to use
)

def format_response(text):
    # Обрабатываем переносы строк для HTML
    return text.replace('\n', '<br>')

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Format the conversation history
    messages = [{"role": "system", "content": system_message}]
    for user_msg, assistant_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if assistant_msg:
            messages.append({"role": "assistant", "content": assistant_msg})
    messages.append({"role": "user", "content": message})
    
    # Generate response
    response = ""
    stream = model.create_chat_completion(
        messages=messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p
    )
    
    for chunk in stream:
        # Extract content from the chunk
        if 'choices' in chunk and len(chunk['choices']) > 0:
            if 'text' in chunk['choices'][0]:
                content = chunk['choices'][0]['text']
            elif 'delta' in chunk['choices'][0] and 'content' in chunk['choices'][0]['delta']:
                content = chunk['choices'][0]['delta']['content']
            else:
                continue
                
            if content is not None:
                response += content
                yield format_response(response)

# Create the Gradio interface with HTML formatting
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="", label="System prompt"),
        gr.Slider(minimum=1, maximum=2048, value=1024, step=1, label="Maximum new tokens"),
        gr.Slider(minimum=0.1, maximum=2.0, value=1, step=0.01, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p",
        ),
    ],
)

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