import io
import os
from openai import OpenAI
from langchain.tools import StructuredTool, Tool
from io import BytesIO
import requests
import json
from io import BytesIO

import chainlit as cl


def get_image_name():
    """
    We need to keep track of images we generate, so we can reference them later
    and display them correctly to our users.
    """
    image_count = cl.user_session.get("image_count")
    if image_count is None:
        image_count = 0
    else:
        image_count += 1

    cl.user_session.set("image_count", image_count)

    return f"image-{image_count}"


def _generate_image(prompt: str):
    """
    This function is used to generate an image from a text prompt using
    DALL-E 3.

    We use the OpenAI API to generate the image, and then store it in our
    user session so we can reference it later.
    """
    client = OpenAI()

    response = client.images.generate(
        model="dall-e-3",
        prompt=prompt,
        size="1024x1024",
        quality="standard",
        n=1,
    )

    image_payload = requests.get(response.data[0].url, stream=True)

    image_bytes = BytesIO(image_payload.content)

    print(type(image_bytes))

    name = get_image_name()
    cl.user_session.set(name, image_bytes.getvalue())
    cl.user_session.set("generated_image", name)
    return name

def _youtube_rag(prompt: str):
    openai.api_key = os.environ["OPENAI_API_KEY"]
    flattened_texts = []
    
    #check if db exists
    if os.path.exists(persist_directory):
       #don't process transcripts
        if debug:
            print("Database exists, skipping transcript processing...")
    else: 
        print("Database does not exist")
    
    if debug:
        print("Initializing database...")
    docsearch = initialize_chroma_db(flattened_texts)

    docs = docsearch.get_relevant_documents(prompt)
    chat_model = ChatOpenAI(model_name="gpt-4-1106-preview")
    chain = load_qa_chain(llm=chat_model, chain_type="stuff")
    answer = chain.run(input_documents=docs, question=prompt)
    return answer

def generate_image(prompt: str):
    image_name = _generate_image(prompt)
    return f"Here is {image_name}."


# this is our tool - which is what allows our agent to generate images in the first place!
# the `description` field is of utmost imporance as it is what the LLM "brain" uses to determine
# which tool to use for a given input.
generate_image_format = '{{"prompt": "prompt"}}'
generate_image_tool = Tool.from_function(
    func=generate_image,
    name="GenerateImage",
    description=f"Useful to create an image from a text prompt. Input should be a single string strictly in the following JSON format: {generate_image_format}",
    return_direct=True,
)


def youtube_rag(prompt: str):
    answer = _youtube_rag(prompt)
    return f" {answer}."

import os
import openai
from langchain.chat_models import ChatOpenAI
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains.question_answering import load_qa_chain
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings.sentence_transformer import SentenceTransformerEmbeddings

debug = False

persist_directory = 'db'
embedding_function = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")

# Function to initialize or load the Chroma database
def initialize_chroma_db(texts):
    if os.path.exists(persist_directory):
        # Load existing database
        if debug:
            print("Loading existing database...")
        db = Chroma(persist_directory="./db", embedding_function=embedding_function)
    else:
        # Create and initialize new database
        #embeddings = OpenAIEmbeddings()
        if debug:
            print("Creating new database...")
        db = Chroma.from_texts(texts, embedding_function, persist_directory=persist_directory)
    return db.as_retriever()



# thisis the youtube rag tool - which is what allows our agent to rag the youtube vector db
# the `description` field is of utmost importance as it is what the LLM "brain" uses to determine
# which tool to use for a given input.
youtube_rag_format = '{{"prompt": "prompt"}}'
generate_image_tool = Tool.from_function(
    func=youtube_rag,
    name="Youtube_Rag",
    description=f"Useful to query the vector database containing youtube transcripts about Aaron Lebauer. Input should be a single string strictly in the following JSON format: {youtube_rag_format}",
    return_direct=True,
)