import requests
import gradio as gr
import preprocessor as tweet_cleaner
# from transformers import pipeline

# pretrained_name = "w11wo/indonesian-roberta-base-sentiment-classifier"

# sentiment = pipeline(
#     "sentiment-analysis",
#     model=pretrained_name,
#     tokenizer=pretrained_name,
#     max_length=512,
#     truncation=True,
# )

API_URL = "https://api-inference.huggingface.co/models/w11wo/indonesian-roberta-base-sentiment-classifier"
headers = {"Authorization": "Bearer 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"}

def format_sentiment(predictions):
  formatted_output = dict()

  for p in predictions:
    if p['label'] == 'positive':
      formatted_output['Positif'] = p['score']
    elif p['label'] == 'negative':
      formatted_output['Negatif'] = p['score']
    else:
      formatted_output['Netral'] = p['score']
  
  return formatted_output

def query(payload):
	response = requests.post(API_URL, headers=headers, json=payload)
	return response.json()

def clean_tweet(tweet):
    return tweet_cleaner.clean(tweet)

def get_sentiment(input_text):
  res = query({"inputs": clean_tweet(input_text)})
  formatted_output = format_sentiment(res[0])
  return formatted_output

examples = list()
examples.append("Semoga saja pelayanan BPJS ke depannya semakin baik. #BPJSKesehatan #TerimaKasihBPJS #BPJSMelayani https://t.co/iDETFSXFJR")
examples.append("min ini mau bayar ko ga bisa yaa m banking sama shopee nya kenapa. Help min udah tenggat nih")
examples.append("Kenaikan harga bpjs yg makin mahal bikin rakyat jadi tambah sengsara pak!")

iface = gr.Interface(
            fn = get_sentiment, 
            inputs = 'text', 
            outputs = ['label'], 
            title = 'Analisis Sentimen Twitter', 
            description="Dapatkan sentimen positif, negatif, atau netral untuk tweet yang dimasukkan.",
            examples=examples
        )
                     
iface.launch(inline = False)