#import streamlit as st
import gradio as gr
from transformers import pipeline
from huggingface_hub import InferenceClient
#import gc

#st.header("Sentiment-demo-app")
#st.subheader("Please be patient and wait up to a minute until the demo app is loaded.")
#st.caption("This is a very simple demo application for a zero-shot classification pipeline to classify positive, neutral, or negative sentiment for a short text. Enter your text in the box below and press CTRl+ENTER to run the model.")

title = "Sentiment-demo-app"
description = """This is a very simple demo application for a sentiment classification pipeline to classify positive, neutral, or negative sentiment for a short text. Enter your text in the box below and press CTRl+ENTER to run the model.
Please be patient until the demo app is loaded. """

sentiment = pipeline("text-classification", model='tabularisai/multilingual-sentiment-analysis') #"zero-shot-classification" model='facebook/bart-large-mnli')

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

def get_sentiment(text):
  output = sentiment(text)
  return f'The sentence was classified as "{output[0]["label"]}" with {output[0]["score"]*100}% confidence'

demo = gr.Interface(
    fn=get_sentiment,
    inputs="text",
    outputs="text",
    title=title,
    description=description
)


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


#texts = st.text_area('Enter text here!')
#candidate_labels = ['Positive', 'Neutral', 'Negative']
#result = pipe(texts)

#if text:
#  out = pipe(text, result)
#  st.json(out)
#  del out
#  gc.collect()