app.py

import gradio as gr
import boto3
from botocore import UNSIGNED
from botocore.client import Config


from huggingface_hub import AsyncInferenceClient

from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.llms import HuggingFaceHub
from langchain.embeddings import HuggingFaceHubEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import RetrievalQA
from langchain.prompts import ChatPromptTemplate
from langchain.document_loaders import WebBaseLoader

from typing import Iterator


text_splitter = RecursiveCharacterTextSplitter(chunk_size=350, chunk_overlap=10)
model_id = HuggingFaceHub(repo_id="HuggingFaceH4/zephyr-7b-beta", model_kwargs={"temperature":0.1, "max_new_tokens":300})
embeddings = HuggingFaceHubEmbeddings()


s3 = boto3.client('s3', config=Config(signature_version=UNSIGNED))
s3.download_file('rad-rag-demos', 'vectorstores/chroma.sqlite3', './chroma_db/chroma.sqlite3')
     
db = Chroma(persist_directory="./chroma_db", embedding_function=embeddings)
db.get()


retriever = db.as_retriever()

global qa 
qa = RetrievalQA.from_chain_type(llm=model_id, chain_type="stuff", retriever=retriever, return_source_documents=True)

def generate(
    message: str,
    chat_history: list[tuple[str, str]],
    system_prompt: str,
    max_new_tokens: int = 1024,
    temperature: float = 0.6,
    top_p: float = 0.9,
    top_k: int = 50,
    repetition_penalty: float = 1.2,
) -> Iterator[str]:
    conversation = []
    if system_prompt:
        conversation.append({"role": "system", "content": system_prompt})
    for user, assistant in chat_history:
        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
    conversation.append({"role": "user", "content": message})

    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
    input_ids = input_ids.to(model.device)

    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
    generate_kwargs = dict(
        {"input_ids": input_ids},
        streamer=streamer,
        max_new_tokens=max_new_tokens,
        do_sample=True,
        top_p=top_p,
        top_k=top_k,
        temperature=temperature,
        num_beams=1,
        repetition_penalty=repetition_penalty,
    )
    t = Thread(target=model.generate, kwargs=generate_kwargs)
    t.start()

    outputs = []
    for text in streamer:
        outputs.append(text)
        yield "".join(outputs)


def add_text(history, text):
    history = history + [(text, None)]
    return history, ""

def bot(history):
    response = infer(history[-1][0])
    history[-1][1] = response['result']
    return history

def infer(question):
    
    query = question
    result = qa({"query": query})
    return result

css="""
#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
"""

title = """
<div style="text-align: center;max-width: 700px;">
    <h1>Chat with PDF</h1>
    <p style="text-align: center;">Upload a .PDF from your computer, click the "Load PDF to LangChain" button, <br />
    when everything is ready, you can start asking questions about the pdf ;)</p>
</div>
"""


with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.HTML(title)      
        chatbot = gr.Chatbot([], elem_id="chatbot")
        with gr.Row():
            question = gr.Textbox(label="Question", placeholder="Type your question and hit Enter ")
    question.submit(add_text, [chatbot, question], [chatbot, question]).then(
        bot, chatbot, chatbot
    )

if __name__ == "__main__":

    demo.launch()