app.py

import os
import time

import langchain
# loaders
from langchain.document_loaders import PyPDFLoader, OnlinePDFLoader, Docx2txtLoader, UnstructuredWordDocumentLoader, UnstructuredPowerPointLoader
# splits
from langchain.text_splitter import RecursiveCharacterTextSplitter
# embeddings
from langchain.embeddings import HuggingFaceEmbeddings, OpenAIEmbeddings
# vector stores
from langchain.vectorstores import Chroma
# huggingface hub
from huggingface_hub import InferenceClient
from langchain import HuggingFaceHub
# models
from langchain.llms import OpenAI
# retrievers
from langchain.chains import RetrievalQA
import gradio as gr

HUGGINGFACEHUB_API_TOKEN = os.environ["HUGGINGFACEHUB_API_TOKEN"]

def build_context(openai_key,files,urls):
  if openai_key != "":
    embeddings = OpenAIEmbeddings(model_name="text-embedding-ada-002", openai_api_key=openai_key)
  else:
    embeddings = HuggingFaceEmbeddings(
                model_name="sentence-transformers/all-MiniLM-L6-v2", model_kwargs={'device': 'cpu'}
            )
  documents = []
  if files is not None:
    print("files not none")
    for idx, file in enumerate(files):
      if file.name.endswith('.pdf'):
        loader = PyPDFLoader(file.name)
        documents.extend(loader.load())
      elif file.name.endswith('.docx'):
        loader = Docx2txtLoader(file.name)
        documents.extend(loader.load())
      elif file.name.endswith('.ppt') or file.name.endswith('.pptx'):
        loader = UnstructuredPowerPointLoader(file.name)
        documents.extend(loader.load())
  if urls != "":
    print("urls not none")
    list_urls = urls.split(sep=",")
    for url in list_urls:
      loader = OnlinePDFLoader(url)
      documents.extend(loader.load())
  text_splitter = RecursiveCharacterTextSplitter(chunk_size=800,chunk_overlap=0,length_function=len,separators=["\n\n", "\n", " ", ""])
  chunked_documents = text_splitter.split_documents(documents)
  global vectordb
  vectordb = Chroma.from_documents(
    documents=chunked_documents,
    embedding=embeddings,
    )
  return "loaded"

def llm_response(openai_key, message, chat_history):
  if openai_key != "":
    llm = OpenAI(
        temperature=0, openai_api_key=openai_key, model_name="gpt-3.5-turbo", verbose=False
    )
  else:
    llm = HuggingFaceHub(repo_id='MBZUAI/LaMini-Flan-T5-248M',
                        huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN,
                         model_kwargs={"max_length":512,"do_sample":True,
                                       "temperature":0.2})
  qa_chain = RetrievalQA.from_chain_type(llm = llm,
                                       chain_type = "stuff",
                                       retriever = vectordb.as_retriever(search_kwargs = {"k": 10}), 
                                       return_source_documents = False,
                                       verbose = True)
  result = qa_chain(message)["result"]
  chat_history.append((message, result))
  time.sleep(2)
  return "", chat_history
  
def loading():
    return "Loading..."

def clear_chromadb():
  ids = vectordb.get()["ids"]
  for id in ids:
    vectordb._collection.delete(ids=id)

with gr.Blocks(theme=gr.themes.Soft()) as demo:
  with gr.Row():
    openai_key = gr.Textbox(label="Enter your OpenAI API Key if you want to use the gpt-3.5-turbo-16k model. If not, the open source LaMini-Flan-T5-248M is used")
  with gr.Row():
      pdf_docs = gr.Files(label="Load pdf files", file_types=['.pdf','.docx','.ppt','.pptx'], type="file")
      urls = gr.Textbox(label="Enter one of multiple online pdf urls (comma separated if multiple)")
  with gr.Row():
    load_docs = gr.Button("Load documents and urls", variant="primary", scale=1)
    loading_status = gr.Textbox(label="Loading status", placeholder="", interactive=False, scale=0)
  with gr.Row():
    with gr.Column(scale=1):
      msg = gr.Textbox(label="User message")
      chatbot = gr.Chatbot()
  with gr.Row():
      clearchat = gr.ClearButton([msg, chatbot], value="New chat",)
      cleardb = gr.Button(value="Reset context (for loading new documents)", variant="secondary")
  load_docs.click(loading, None, loading_status, queue=False)
  load_docs.click(build_context, inputs=[openai_key,pdf_docs, urls], outputs=[loading_status], queue=False)
  msg.submit(llm_response, [openai_key, msg, chatbot], [msg, chatbot])
  cleardb.click(clear_chromadb)
    
demo.queue(concurrency_count=3)
demo.launch()