app.py

import gradio as gr
import time
from transformers import pipeline, TFAutoModelForSeq2SeqLM, AutoTokenizer
import torch, os
if torch.backends.mps.is_available():
    device = torch.device("mps")
elif torch.cuda.is_available():
    device = torch.device("cuda")
else:
    device = torch.device("cpu")
print(device)



modelName = "Qwen/Qwen2.5-1.5B-Instruct" #Qwen/Qwen2.5-1.5B-Instruct 
chatbot_model = pipeline("text-generation", model=modelName, device=device, batch_size=16)
sentiment_model = pipeline("sentiment-analysis", device=device)


def echo(message, history, system_prompt, tokens):
    conversation_history = []
    for item in history:
        conversation_history.append(
        {"role": "user", "content": item[0]},
        )
        conversation_history.append(
            {"role": "assistant", "content": item[1]},
        )
    sentiment = sentiment_model(message)[0]
    if sentiment['label'] == 'NEGATIVE':
        prompt = f"The user is upset, respond with empathy and support: {message}"
    else:
        prompt = f"Respond to the following query: {message}"
    
    conversation_history.append(
        {"role": "user", "content": prompt},
        )
    print(conversation_history)
    result = chatbot_model(conversation_history, num_return_sequences=1, max_new_tokens=tokens)
    
    return result[0]['generated_text'][-1]['content'] 




demo = gr.ChatInterface(
	echo,
	additional_inputs=[
    	gr.Textbox(value="You are a friendly Chatbot.", label="System prompt"),
    	gr.Slider(minimum=1, maximum=500, value=225, step=1, label="Max new tokens"),
	],
    title="Scotts Chatbot",
    description="A friendly, AI-powered chatbot designed to respond to user queries based on the Qwen model.",
    # css="footer {visibility: hidden}"
)

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