from statistics import mode
import urllib.request
import gradio as gr
import subprocess
import unidecode
import requests
import string
import json


# Download files using wget
keys_url = "https://github.com/Nkluge-correa/Aira/raw/master/Aira-1/data/generated_data/keys_en.json"
answers_url = "https://github.com/Nkluge-correa/Aira/raw/master/Aira-1/data/original_data/answers_en.txt"

subprocess.run(["wget", keys_url, "-O", "keys_en.json"])
subprocess.run(["wget", answers_url, "-O", "answers_en.txt"])

# Load data from files
with open('answers_en.txt', encoding='utf-8') as fp:
    answers = [line.strip() for line in fp]

with open('keys_en.json', 'r') as fp:
    vocabulary = json.load(fp)


with open('keys_en.json') as json_file:
    dictionary = json.load(json_file)

def generate_ngrams(text, WordsToCombine):
    """
    Generates n-grams of length WordsToCombine from the input text.

    Args:
        text: A string representing the input text
        WordsToCombine: An integer representing the
            size of the n-grams to be generated

    Returns:
        A list of n-grams generated from the input text, where each
        n-gram is a list of WordsToCombine words
    """
    words = text.split()
    output = []
    for i in range(len(words) - WordsToCombine+1):
        output.append(words[i:i+WordsToCombine])
    return output


def make_keys(text, WordsToCombine):
    """
    Given a text and a number of words to combine, returns
    a list of keys that correspond to all possible combinations
    of n-grams (sequences of n consecutive words) in the text.

    Args:
        - text (str): The input text.
        - WordsToCombine (int): The number of words to combine.

    Returns:
        - sentences (list of str): A list of all the keys, which are
        the n-grams in the text.
    """
    gram = generate_ngrams(text, WordsToCombine)
    sentences = []
    for i in range(0, len(gram)):
        sentence = ' '.join(gram[i])
        sentences.append(sentence)
    return sentences


def chat(message, history):
    """
    A function that generates a response to a user input message
    based on a pre-built dictionary of responses.

    Args:
        message (str): A string representing the user's input message.
        history (list): A list of tuples containing previous
        messages and responses.

    Returns:
        tuple: A tuple containing two lists of tuples. The first list is
        the original history with the user's input message and the bot's
        response appended as a tuple. The second list is an updated history
        with the same tuples.
    """
    history = history or []
    text = message.lower()
    sentences = []
    values = []
    new_text = text.translate(str.maketrans('', '', string.punctuation))
    new_text = unidecode.unidecode(new_text)

    if len(new_text.split()) == 1:
        if new_text in dictionary.keys():
            l = [dictionary[new_text]] * 100
            values.append(l)
        new_text = new_text + ' ' + new_text

    else:
        if new_text in dictionary.keys():
            l = [dictionary[new_text]] * 100
            values.append(l)

    for i in range(1, len(new_text.split()) + 1):
        sentence = make_keys(new_text, i)
        sentences.append(sentence)

    for i in range(len(sentences)):
        attention = sentences[i]
        for i in range(len(attention)):
            if attention[i] in dictionary.keys():
                l = [dictionary[attention[i]]] * i
                values.append(l)

    if len([item for sublist in values for item in sublist]) == 0:
        bot_input_ids = "I'm sorry, either I didn't understand the question, or it is not part of my domain of expertise... :( Try asking it in another way or using other words. Maybe then I can help you!"
        history.append((message, bot_input_ids))
        return history, history

    else:
        values = [item for sublist in values for item in sublist]
        prediction = mode(values)
        bot_input_ids = answers[int(prediction)-1]
        history.append((message, bot_input_ids))
        return history, history

title = "Basic Chatbot - By Teeny-Tiny Castle 🏰"

head = (
  "<center>"
  "<img src='https://d2vrvpw63099lz.cloudfront.net/do-i-need-a-chatbot/header-chat-box.png' width=400>"
  "This is an example of a rules-based closed domain chatbot. It knows a couple of answers to questions related to AI."
  "<br>"
  "</center>"
)

ref = (
"<center>"
"To see its full version (ML style) of this bot, go to <a href='https://playground.airespucrs.org/aira'>this link</a>."
"</center>")

# create a chat interface
chatbot = gr.Chatbot()

demo = gr.Interface(
    chat,
    ["text", "state"],
    [chatbot, "state"],
    allow_flagging="never",
    title=title,
    description=head,
    article=ref
)

demo.launch()