app.py

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, pipeline
import gradio as gr
import chromadb
from langchain.document_loaders import PyPDFDirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import ConversationalRetrievalChain
from langchain.memory import ConversationBufferMemory
from langchain_huggingface import HuggingFacePipeline

# Download the model from HuggingFace
model_name = "anakin87/zephyr-7b-alpha-sharded"
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    quantization_config=bnb_config
)

tokenizer = AutoTokenizer.from_pretrained(model_name)
tokenizer.bos_token_id = 1  # Set beginning of sentence token id

# Specify embedding model
embedding_model_name = "sentence-transformers/all-mpnet-base-v2"
model_kwargs = {"device": "cpu"}  # Using CPU since GPU is not available
embeddings = HuggingFaceEmbeddings(model_name=embedding_model_name, model_kwargs=model_kwargs)

# Load the documents (replace this with your document loading logic)
documents = ["Sample document text 1", "Sample document text 2"]

# Split the documents into small chunks
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
all_splits = text_splitter.split_documents(documents)

# Embed document chunks
vectordb = Chroma.from_documents(documents=all_splits, embedding=embeddings, persist_directory="chroma_db")

# Specify the retriever
retriever = vectordb.as_retriever()

# Build HuggingFace pipeline for using zephyr-7b-alpha
pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    use_cache=True,
    device_map="auto",
    max_length=2048,
    do_sample=True,
    top_k=5,
    num_return_sequences=1,
    eos_token_id=tokenizer.eos_token_id,
    pad_token_id=tokenizer.eos_token_id,
)

# Specify the llm
llm = HuggingFacePipeline(pipeline=pipeline)

# Define the create_conversation function
def create_conversation(query: str, chat_history: list) -> tuple:
    try:
        memory = ConversationBufferMemory(
            memory_key='chat_history',
            return_messages=False
        )
        qa_chain = ConversationalRetrievalChain.from_llm(
            llm=llm,
            retriever=retriever,
            memory=memory,
            get_chat_history=lambda h: h,
        )

        result = qa_chain({'question': query, 'chat_history': chat_history})
        chat_history.append((query, result['answer']))
        return '', chat_history

    except Exception as e:
        chat_history.append((query, e))
        return '', chat_history

# Define the Gradio UI
with gr.Blocks() as demo:
    chatbot = gr.Chatbot(label='My Chatbot')
    msg = gr.Textbox()
    clear = gr.ClearButton([msg, chatbot])

    msg.submit(create_conversation, [msg, chatbot], [msg, chatbot])

# Launch the Gradio demo
demo.launch()