app.py

# Documentation Process
# clone https://github.com/TheMITTech/shakespeare

import subprocess
subprocess.run(["git", "clone", "https://github.com/TheMITTech/shakespeare.git"])

#
from glob import glob
files = glob("./shakespeare/**/*.html")

# Documents copy and move
import shutil
import os

os.mkdir('./data')
destination_folder = './data/'

for html_file in files:
  shutil.move(html_file, destination_folder + html_file.split("/")[-1])

# Documents Read with BueatifulSoup
from langchain.document_loaders import BSHTMLLoader, DirectoryLoader
bshtml_dir_loader = DirectoryLoader("./data", loader_cls=BSHTMLLoader)### YOUR CODE HERE
data = bshtml_dir_loader.load()

# Prepare tokenization process for the documents
from transformers import AutoTokenizer
bloomz_tokenizer = AutoTokenizer.from_pretrained("bigscience/bloomz-1b7")

# Now that we have our tokenizer - let's split our documents into bitesized pieces! Let's split our documents on the newline character!
from langchain.text_splitter import CharacterTextSplitter
text_splitter = CharacterTextSplitter.from_huggingface_tokenizer(bloomz_tokenizer, chunk_size=100, chunk_overlap=0, separator='\n')
documents = text_splitter.split_documents(data)

# Set an embedding for the documents
from langchain.embeddings import HuggingFaceEmbeddings
embeddings = HuggingFaceEmbeddings()

# Store them onto Chroma DB
from langchain.vectorstores import Chroma
persist_directory = "vector_db"
vectordb = Chroma.from_documents(documents, embedding = embeddings, persist_directory=persist_directory)

#
vectordb.persist()
vectordb = None

#
vectordb = Chroma(persist_directory=persist_directory, embedding_function=embeddings)

# First up, let's load our model!
from langchain import HuggingFacePipeline

llm = HuggingFacePipeline.from_model_id(
    model_id="bigscience/bloomz-1b7", 
    task="text-generation", 
    model_kwargs={"temperature" : 0, "max_length" : 500})

# Now let's set up our document vector store as a Retriever tool so we can leverage it in our chain!
doc_retriever =  vectordb.as_retriever()

# Combine them all
from langchain.chains import RetrievalQA
shakespeare_qa = RetrievalQA.from_chain_type(llm = llm, chain_type = "stuff", retriever=doc_retriever)

def make_answer(query):
    return shakespeare_qa.run(query)

if __name__ == "__main__":
    # Make a gradio interface
    import gradio as gr
    
    gr.Interface(
        make_answer,
        [gr.inputs.Textbox(lines = 2, label = "Question")],
        gr.outputs.Textbox(label = "Answer"),
        title = "GenerativeQA",
        description = "GenerativeQA is a question and answer model.",
        
    ).launch()