import google.generativeai as genai
import datetime
from urllib.request import Request, urlopen
from pypdf import PdfReader
from io import StringIO
import io
import pandas as pd
import os
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers import pipeline
from openai import OpenAI

openai_key = "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"
client = OpenAI(api_key = openai_key)

#tokenizer = AutoTokenizer.from_pretrained("mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis")
#model = AutoModelForSequenceClassification.from_pretrained("mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis")
pipe = pipeline("text-classification", model="mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis")

def filter(input_json):

  sym = pd.read_excel('symbol.xlsx',header = None)
  sym_list  = sym.iloc[:,0].to_list()

  desc = pd.read_excel('Descriptor.xlsx',header = None)
  desc_list  = desc.iloc[:,0].to_list()
    
  if input_json['FileURL']==None or input_json['FileURL'].lower()=='null':
    return [0,"File_URL"]
  if input_json['symbol']== 'null' or input_json['symbol'] not in sym_list:
    return [0,"symbol"]
  if input_json['TypeofAnnouncement'] not in ['General_Announcements','Outcome','General']:
    return [0,"Annoucement"]
  if input_json['Descriptor'] not in desc_list:
    return [0,"Desc"]
      
  url = 'https://www.bseindia.com/xml-data/corpfiling/AttachLive/'+ input_json['FileURL'].split('Pname=')[-1]
  req = Request(url, headers={'User-Agent': 'Mozilla/5.0'})
  html = urlopen(req)
  cont = html.read()
  reader = PdfReader(io.BytesIO(cont))
  content = ''
  for i in range(len(reader.pages)):
    content+= reader.pages[i].extract_text()
  document = content

  

  return [1, document]

def summary(input_json):

    key = os.getenv("key")
    genai.configure(api_key=key)
    model = genai.GenerativeModel('gemini-pro')
    
    output = {}
    filtering_results = filter(input_json)
    if filtering_results[0] == 0:
        return 0
    #return filtering_results[1]
    
    long_text = filtering_results[1]
    mobile = model.generate_content("Summarize this Financial letter in 60 words to be used as a news article. {}".format(long_text))
    output['mobile'] = mobile.text
    web = model.generate_content("Summarize this Financial letter in 128 words to be used as a news article. {}".format(long_text))
    output['web'] = web.text
    tag = model.generate_content("1 word Financial SEO tag for this news article {}".format(mobile.text))
    output['tag'] = tag.text 
    headline = model.generate_content("Headline for this News Article {}".format(web.text))
    output['headline'] = headline.text
    utc_now = datetime.datetime.utcnow()
    ist_now = utc_now.astimezone(datetime.timezone(datetime.timedelta(hours=5, minutes=30)))
    output['Time'] = ist_now.strftime("%I:%M %p")
    month_name = ist_now.strftime("%B")
    output['Date'] = f"{ist_now.day} {month_name}, {ist_now.year}"
    #senti = pipe(mobile.text)
    #output['sentiment'] = senti[0]['label']
    senti = model.generate_content("Answer in one word the sentiment of this News out of Positive, Negative or Neutral {}".format(web.text))
    output['sentiment'] = senti.text
    # response = client.images.generate(
    # model="dall-e-3",
    # prompt=headline.text,
    # size="1024x1024",
    # quality="standard",
    # n=1
    # ) 
    # output["image_url"] = response.data[0].url
    
    return output