app.py

# You can find this code for Chainlit python streaming here (https://docs.chainlit.io/concepts/streaming/python)

# OpenAI Chat completion
import os
from openai import AsyncOpenAI  # importing openai for API usage
import chainlit as cl  # importing chainlit for our app
from chainlit.prompt import Prompt, PromptMessage  # importing prompt tools
from chainlit.playground.providers import ChatOpenAI  # importing ChatOpenAI tools
from dotenv import load_dotenv
from langchain.document_loaders import CSVLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_openai import OpenAIEmbeddings
from langchain.embeddings import CacheBackedEmbeddings
from langchain.storage import LocalFileStore
from langchain_community.vectorstores import FAISS
from datasets import load_dataset



load_dotenv()

# ChatOpenAI Templates
system_template = """You are a helpful assistant who always speaks in a pleasant tone!
"""

user_template = """{input}
Think through your response step by step.
"""

def setup():
    dataset = load_dataset("ShubhamChoksi/IMDB_Movies")
    print(dataset['train'][0])
    print("data from huggingface dataset\n")
    
    dataset_dict = dataset
    dataset_dict["train"] # TODO - what method do we have to use to store imdb.csv from ShubhamChoksi/IMDB_Movies?

    dataset_dict["train"].to_csv("imdb.csv")
    loader = CSVLoader(file_path='imdb.csv')
    data = loader.load()
    len(data)
    print(data[0])
    print("loaded data from csv\n")
    
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size = 1000,
        chunk_overlap = 100,
    )

    chunked_documents = text_splitter.split_documents(data)


    len(chunked_documents) # ensure we have actually split the data into chunks
    print(chunked_documents[0])
    
    openai_api_key =  os.getenv("OPENAI_API_KEY")

    embedding_model = OpenAIEmbeddings(openai_api_key=openai_api_key)
    
    
    store = LocalFileStore("./cache/")
    embedder = CacheBackedEmbeddings.from_bytes_store(
        embedding_model, store, namespace=embedding_model.model
    )
    
    vector_store = FAISS.from_documents(chunked_documents, embedder)

    vector_store.save_local("./vector_store")
    
    vector_store = FAISS.load_local("./vector_store", embedder, allow_dangerous_deserialization=True)

    retriever = vector_store.as_retriever()
    
    query = "What are some good westerns movies?"
    embedded_query = embedding_model.embed_query(query)
    similar_documents = vector_store.similarity_search_by_vector(embedded_query)
    for page in similar_documents:
        # TODO: Print the similar documents that the similarity search returns?
        print(page)
        print("00-----0000")
        print(page)
        print("-------------")

@cl.on_chat_start  # marks a function that will be executed at the start of a user session
async def start_chat():
    settings = {
        "model": "gpt-3.5-turbo",
        "temperature": 0,
        "max_tokens": 500,
        "top_p": 1,
        "frequency_penalty": 0,
        "presence_penalty": 0,
    }

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


@cl.on_message  # marks a function that should be run each time the chatbot receives a message from a user
async def main(message: cl.Message):
    settings = cl.user_session.get("settings")

    client = AsyncOpenAI()

    print(message.content)

    prompt = Prompt(
        provider=ChatOpenAI.id,
        messages=[
            PromptMessage(
                role="system",
                template=system_template,
                formatted=system_template,
            ),
            PromptMessage(
                role="user",
                template=user_template,
                formatted=user_template.format(input=message.content),
            ),
        ],
        inputs={"input": message.content},
        settings=settings,
    )

    print([m.to_openai() for m in prompt.messages])

    msg = cl.Message(content="")

    # Call OpenAI
    async for stream_resp in await client.chat.completions.create(
        messages=[m.to_openai() for m in prompt.messages], stream=True, **settings
    ):
        token = stream_resp.choices[0].delta.content
        if not token:
            token = ""
        await msg.stream_token(token)

    # Update the prompt object with the completion
    prompt.completion = msg.content
    msg.prompt = prompt

    # Send and close the message stream
    await msg.send()