app.py

import os
from chromadb.config import Settings

from langchain.chains import RetrievalQA
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.vectorstores import Chroma
import os
import requests
from fastapi import FastAPI, UploadFile, File
from typing import List, Optional
import urllib.parse
from langchain.llms import HuggingFacePipeline
from transformers import AutoTokenizer, AutoModelForCausalLM
import transformers
from torch import cuda, bfloat16
import gradio as gr
import gc
import torch


persist_directory = "db"
source_directory = 'source_documents'
embeddings_model_name = "all-MiniLM-L6-v2"
model = "tiiuae/falcon-7b-instruct"
chunk_size = 500
chunk_overlap = 50
target_source_chunks = 4
# Define the folder for storing database
persist_directory = 'db'


embeddings = HuggingFaceEmbeddings(model_name=embeddings_model_name)
llm = HuggingFacePipeline.from_model_id(model_id=model, task="text-generation", device=0, model_kwargs={"temperature":0.1,"trust_remote_code": True, "max_length":100000, "top_p":0.15, "top_k":0, "repetition_penalty":1.1, "num_return_sequences":1, "torch_dtype":bfloat16})


# Define the Chroma settings
CHROMA_SETTINGS = Settings(
        chroma_db_impl='duckdb+parquet',
        persist_directory=persist_directory,
        anonymized_telemetry=False
)

import os
import glob
from typing import List
import argparse

from langchain.document_loaders import (
    CSVLoader,
    EverNoteLoader,
    PDFMinerLoader,
    TextLoader,
    UnstructuredEmailLoader,
    UnstructuredEPubLoader,
    UnstructuredHTMLLoader,
    UnstructuredMarkdownLoader,
    UnstructuredODTLoader,
    UnstructuredPowerPointLoader,
    UnstructuredWordDocumentLoader,
)

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.vectorstores import Chroma
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.docstore.document import Document
# from constants import CHROMA_SETTINGS
# from PyPDF2 import PdfReader
import requests

# Map file extensions to document loaders and their arguments
LOADER_MAPPING = {
    ".csv": (CSVLoader, {}),
    # ".docx": (Docx2txtLoader, {}),
    ".doc": (UnstructuredWordDocumentLoader, {}),
    ".docx": (UnstructuredWordDocumentLoader, {}),
    ".enex": (EverNoteLoader, {}),
    # ".eml": (MyElmLoader, {}),
    ".epub": (UnstructuredEPubLoader, {}),
    ".html": (UnstructuredHTMLLoader, {}),
    ".md": (UnstructuredMarkdownLoader, {}),
    ".odt": (UnstructuredODTLoader, {}),
    ".pdf": (PDFMinerLoader, {}),
    ".ppt": (UnstructuredPowerPointLoader, {}),
    ".pptx": (UnstructuredPowerPointLoader, {}),
    ".txt": (TextLoader, {"encoding": "cp1252"}),
    # Add more mappings for other file extensions and loaders as needed
}


def load_single_document(file_path: str) -> List[Document]:
    ext = "." + file_path.rsplit(".", 1)[-1]
    if ext in LOADER_MAPPING:
        loader_class, loader_args = LOADER_MAPPING[ext]
        loader = loader_class(file_path, **loader_args)
        return loader.load()

    raise ValueError(f"Unsupported file extension '{ext}'")

def load_documents(source_dir: str, ignored_files: List[str] = []) -> List[Document]:
    """
    Loads all documents from the source documents directory, ignoring specified files
    """
    all_files = []
    for ext in LOADER_MAPPING:
        all_files.extend(
            glob.glob(os.path.join(source_dir, f"**/*{ext}"), recursive=True)
        )
    filtered_files = [file_path for file_path in all_files if file_path not in ignored_files]

    with Pool(processes=os.cpu_count()) as pool:
        results = []
        with tqdm(total=len(filtered_files), desc='Loading new documents', ncols=80) as pbar:
            for i, docs in enumerate(pool.imap_unordered(load_single_document, filtered_files)):
                results.extend(docs)
                pbar.update()

    return results

def process_documents(ignored_files: List[str] = []) -> List[Document]:
    """
    Load documents and split in chunks
    """
    print(f"Loading documents from {source_directory}")
    documents = load_documents(source_directory, ignored_files)
    if not documents:
        print("No new documents to load")
        exit(0)
    print(f"Loaded {len(documents)} new documents from {source_directory}")
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
    texts = text_splitter.split_documents(documents)
    print(f"Split into {len(texts)} chunks of text (max. {chunk_size} tokens each)")
    return texts

def does_vectorstore_exist(persist_directory: str) -> bool:
    """
    Checks if vectorstore exists
    """
    if os.path.exists(os.path.join(persist_directory, 'index')):
        if os.path.exists(os.path.join(persist_directory, 'chroma-collections.parquet')) and os.path.exists(os.path.join(persist_directory, 'chroma-embeddings.parquet')):
            list_index_files = glob.glob(os.path.join(persist_directory, 'index/*.bin'))
            list_index_files += glob.glob(os.path.join(persist_directory, 'index/*.pkl'))
            # At least 3 documents are needed in a working vectorstore
            if len(list_index_files) > 3:
                return True
    return False

def ingest():
    # Load environment variables
    embeddings_model_name = "all-MiniLM-L6-v2"
    persist_directory = "db"
    model = "tiiuae/falcon-7b-instruct"
    source_directory = "source_documents"
    os.makedirs(source_directory, exist_ok=True)
    # Load documents and split in chunks
    print(f"Loading documents from {source_directory}")
    chunk_size = 500
    chunk_overlap = 50
    documents = load_documents(source_directory)
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
    texts = text_splitter.split_documents(documents)
    print(f"Loaded {len(documents)} documents from {source_directory}")
    print(f"Split into {len(texts)} chunks of text (max. {chunk_size} characters each)")

    # Create embeddings
    # embeddings = HuggingFaceEmbeddings(model_name=embeddings_model_name)

    # Create and store locally vectorstore
    db = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory, client_settings=CHROMA_SETTINGS)
    db.persist()
    db = None

def embed_documents(files):

    saved_files = []
    source_directory = "source_documents"

    # print(files)
    # Save the files to the specified folder
    for file_ in files:
        print(type(file_))
        os.makedirs(source_directory, exist_ok= True)
        filename = "file.pdf"

        file_path = os.path.join(source_directory, filename)
        saved_files.append(file_path)

        print(type(file_))
        print(file_path)
        # file_content = file_.read()
        with open(file_path, "wb") as f:
            print("write")
            f.write(file_)
    ingest()

    # Delete the contents of the folder
    [os.remove(os.path.join(source_directory, filename)) or os.path.join(source_directory, filename) for file in files]

    return {"message": "Files embedded successfully"}

def retrieve_documents(query: str):
    target_source_chunks = 4
    mute_stream = ""
    embeddings_model_name = "all-MiniLM-L6-v2"

    db = Chroma(persist_directory=persist_directory, embedding_function=embeddings, client_settings=CHROMA_SETTINGS)
    retriever = db.as_retriever(search_kwargs={"k": target_source_chunks})
    # Prepare the LLM
    callbacks = [] if mute_stream else [StreamingStdOutCallbackHandler()]

    qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever, return_source_documents=False)

    # Get the answer from the chain
    res = qa(query)
    print(res)
    answer = res['result']
    torch.cuda.empty_cache()
    gc.collect()
    return answer

with gr.Blocks() as demo:
  with gr.Row():
    with gr.Column():
        file_input = gr.File(file_count="multiple", file_types=["text", ".json", ".csv", ".pdf"], type= 'binary')
        initiate_btn = gr.Button(value="Generate Embedding")

    with gr.Column():
        question = gr.Textbox(label="Question")
        question_btn = gr.Button(value="Question_btn")
        answer = gr.Textbox(label="answer")

  initiate_btn.click(embed_documents, inputs=file_input, api_name="embed-file")

  question_btn.click(retrieve_documents, inputs=question , outputs=answer, api_name="llm")

demo.queue().launch()