app.py

import os
import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

# Use the base (untrained) model from Hugging Face Hub
model_id = "mistralai/Mistral-7B-Instruct-v0.3"
api_key = os.environ.get("HF_KEY")  # Your Hugging Face token

tokenizer = AutoTokenizer.from_pretrained(model_id, token = api_key)
model = AutoModelForCausalLM.from_pretrained(model_id, token = api_key)

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=512,
    do_sample=True,
)

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Combine history and system message into a prompt
    prompt = system_message.strip() + "\n"
    for user, assistant in history:
        if user:
            prompt += f"User: {user}\n"
        if assistant:
            prompt += f"Assistant: {assistant}\n"
    prompt += f"User: {message}\nAssistant:"

    outputs = pipe(
        prompt,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        pad_token_id=tokenizer.eos_token_id,
    )
    response = outputs[0]["generated_text"][len(prompt):]
    yield response.strip()

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a professional AI coach helping people build skills.", 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=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
    ],
)

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