import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers import TapexTokenizer, BartForConditionalGeneration
import pandas as pd
import torch

import numpy as np
import time
import os
#import pkg_resources

'''
# Get a list of installed packages and their versions
installed_packages = {pkg.key: pkg.version for pkg in pkg_resources.working_set}

# Print the list of packages
for package, version in installed_packages.items():
    print(f"{package}=={version}")
'''

# Load the chatbot model
chatbot_model_name = "microsoft/DialoGPT-medium" 
tokenizer = AutoTokenizer.from_pretrained(chatbot_model_name)
model = AutoModelForCausalLM.from_pretrained(chatbot_model_name)

# Load the SQL Model
sql_model_name = "microsoft/tapex-large-finetuned-wtq"
sql_tokenizer = TapexTokenizer.from_pretrained(sql_model_name)
sql_model = BartForConditionalGeneration.from_pretrained(sql_model_name)

data = {
    "year": [1896, 1900, 1904, 2004, 2008, 2012],
    "city": ["athens", "paris", "st. louis", "athens", "beijing", "london"]
}
table = pd.DataFrame.from_dict(data)

sql_response = None

def predict(input, history=[]):

    global sql_response
    # Check if the user input is a question
    is_question = "?" in input

    if is_question: 
        sql_encoding = sql_tokenizer(table=table, query=input + sql_tokenizer.eos_token, return_tensors="pt")
        sql_outputs = sql_model.generate(**sql_encoding)
        sql_response = sql_tokenizer.batch_decode(sql_outputs, skip_special_tokens=True)

        # Append the SQL model's response to the history
        sql_response_ids = tokenizer.encode(sql_response + tokenizer.eos_token, return_tensors='pt')
        history.extend(sql_response_ids[0].tolist())  # Add SQL response token IDs to history

        
        '''
        bot_input_ids = torch.cat([torch.LongTensor(history), sql_encoding], dim=-1)
        history = sql_model.generate(bot_input_ids, max_length=1000, pad_token_id=sql_tokenizer.eos_token_id).tolist()
        response = sql_tokenizer.decode(history[0]).split("<|endoftext|>")
        response = [(response[i], response[i + 1]) for i in range(0, len(response) - 1, 2)]
        '''
    else:
        # tokenize the new input sentence
        new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
    
        # append the new user input tokens to the chat history
        bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
          
        # generate a response
        history = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id).tolist()
    
        # convert the tokens to text, and then split the responses into the right format
        response = tokenizer.decode(history[0]).split("<|endoftext|>")
        response = [(response[i], response[i + 1]) for i in range(0, len(response) - 1, 2)]  # convert to tuples of list
        
    return response, history


import gradio as gr

interface = gr.Interface(
    fn=predict,
    theme="default",
    css=".footer {display:none !important}",
    inputs=["text", "state"],
    outputs=["chatbot", "state"],
    title="ST Chatbot",
    description="Type your message in the box above, and the chatbot will respond.",
)

if __name__ == '__main__':
    interface.launch()