import subprocess
def install(name):
    subprocess.call(['pip', 'install', name])

install('PyPDF2')
import PyPDF2
import numpy as np
install('tensorflow')
import tensorflow as tf
install('tensorflow-hub')
import tensorflow_hub as hub
import time
import gradio as gr
import re

module_url = "https://tfhub.dev/google/universal-sentence-encoder/4"
model = hub.load(module_url)
print ("module %s loaded" % module_url)

def Model(f, q):
    reader = PyPDF2.PdfReader(f)
    text = ''
    for i in range(len(reader.pages)):
        text = text + reader.pages[i].extract_text()
    text = text.replace('\n',' ')
    data = re.split(r"[.]\s+(?=[A-Z1-9])", text)
    embs = model(data)

    doc = [{"txt":d[0], "emb":np.array(d[1]).tolist()} for d in zip(data, embs)]

    embedding = []
    for i in range(len(doc)):
        embedding.append(doc[i]['emb'])

    para = []
    for i in range(len(doc)):
        para.append(doc[i]['txt'])

    qry = q
    q = np.array(model([qry])).tolist()
    y = np.argsort(np.inner(q, embedding))[0]

    ans = ''
    for i in y[-3:][::-1]:
        ans = ans + para[i] + '\n'
    return ans

with gr.Blocks(css=".gradio-container-3-18-0 {background-color: lightskyblue} .contain.svelte-1iguv9h.svelte-1iguv9h.svelte-1iguv9h {background-color: lightskyblue} .lg.svelte-1ma3u5b {background-color: darkorange}") as demo:
    gr.HTML('<h1 style="text-align:center; background-color: lightskyblue;">Answer Market Requirements Document (MRD)</h1>')
    f = gr.File(label='Upload requirements and specifications of company (.pdf file) here!')
    q = gr.Textbox(placeholder="Enter Query", label='Enter Query here!')
    submit_btn = gr.Button(value="Answer")
    output = gr.Textbox(label='RESULTS!')
    submit_btn.click(fn=Model, inputs=[f,q], outputs=output, scroll_to_output=True, api_name="answer-mrd")

demo.launch(auth=("admin","12e4"))