from gpt_index import GPTSimpleVectorIndex
from llama_index.indices.query.query_transform.base import HyDEQueryTransform
import gradio as gr
from gradio import Interface, Textbox
import sys
import os
from datetime import datetime, timedelta
import pytz
import huggingface_hub
from huggingface_hub import Repository, HfApi
from datetime import datetime
import csv

os.environ["OPENAI_API_KEY"] = os.environ['SECRET_CODE']

AUS_TIMEZONE = pytz.timezone('Australia/Sydney')

# Best practice is to use a persistent dataset 
DATASET_REPO_URL = "https://huggingface.co/datasets/peterpull/MediatorBot"
DATA_FILENAME = "data.txt"
INDEX_FILENAME = "index2.json"
DATA_FILE = os.path.join("data", DATA_FILENAME)
INDEX_FILE = os.path.join("data", INDEX_FILENAME)

#this will be called later to upload the chat history back to the dataset
api=HfApi()

# we need a HF access token - read I think suffices becuase we are cloning the distant repo to local space repo.
HF_TOKEN = os.environ.get("HF_TOKEN")
print("HF TOKEN is none?", HF_TOKEN is None)
print("HF hub ver", huggingface_hub.__version__)

#Clones the distant repo to the local repo
repo = Repository(
    local_dir='data', 
    clone_from=DATASET_REPO_URL, 
    use_auth_token=HF_TOKEN)

#PRINT file locations
print(f"Repo local_dir: {repo.local_dir}")
print(f"Repo files: {os.listdir(repo.local_dir)}")
print (f"Index file:{INDEX_FILENAME}")



def generate_text() -> str:
    with open(DATA_FILE) as file:
        text = ""
        for line in file:
            row_parts = line.strip().split(",")
            if len(row_parts) != 3:
                continue
            user, chatbot, time = row_parts
            text += f"Time: {time}\nUser: {user}\nChatbot: {chatbot}\n\n"
        return text if text else "No messages yet"

def store_message(chatinput: str, chatresponse: str):
    if chatinput and chatresponse:

        now = datetime.now() # current time in UTC
        aus_time = now.astimezone(AUS_TIMEZONE) # convert to Australia timezone
        timestamp = aus_time.strftime("%Y-%m-%d %H:%M:%S")
        user_input = f"User: {chatinput}"
        chatbot_response = f"Chatbot: {chatresponse}"
        separator = "-" * 30
        message = f"{timestamp}\n{user_input}\n{chatbot_response}\n{separator}\n"
  
        with open(DATA_FILE, "a") as file:
            file.write(message)
            print(f"Wrote to datafile: {message}")
              
    #need to find a way to push back to dataset repo
        HF_WRITE_TOKEN = os.environ.get("WRITE_TOKEN")
        api.upload_file(
            path_or_fileobj=DATA_FILE,
            path_in_repo='data.txt',
            repo_id="peterpull/MediatorBot",
            repo_type="dataset",
            commit_message="Add new chat history",
            use_auth_token=HF_WRITE_TOKEN)
            
    return generate_text()

def get_index(index_file_path):
    if os.path.exists(index_file_path):
        #print 500 characters of json header
        print_header_json_file(index_file_path)
        index_size = os.path.getsize(index_file_path)
        print(f"Size of {index_file_path}: {index_size} bytes") #let me know how big json file is.
          
        loaded_index = GPTSimpleVectorIndex.load_from_disk(index_file_path)
        return loaded_index
    else:
        print(f"Error: '{index_file_path}' does not exist.")
        sys.exit()

def print_header_json_file(filepath):
    with open(filepath, 'r') as f:
        file_contents = f.read()
        print ("JSON FILE HEADER:")
        print(file_contents[:500]) # print only the first 500 characters

index = get_index(INDEX_FILE)
# define the conversation_history list
conversation_history = []

# passes the prompt to the chatbot
def chatbot(input_text, history=conversation_history):
    hyde= HyDEQueryTransform(include_original=True)
    prompt = f"In character as John Haynes, please respond to: {input_text}. Only reply with contextual information or say you cannot find an answer. End with a reflective question."
    response = index.query(prompt, response_mode="default", verbose=True, query_transform=hyde)
    store_message(input_text,response)

   # append the current input and response to the conversation history
    history.append((input_text, response.response))
    
 # return the response and updated conversation history
    return [(input_text, response.response)], history

with open('about.txt', 'r') as file:
    about = file.read()

examples=[["What are three excellent questions to ask at intake?"],["How do you handle high conflict divorce cases?"],["Which metaphors do you steer parties away from in mediation? Which do you prefer?"]]
description="GPT3_Chatbot drawing on contextual mediation material, v0.6H"
title="The MediatorBot"

iface = Interface(
    fn=chatbot,
    inputs=[Textbox("Enter your question"), "state"],
    outputs=["chatbot", "state"],
    title=title,
    description=description,
    article=about,
    examples=examples)
                                         
iface.launch()