app.py

import os
from typing import List

from langchain.embeddings.openai import OpenAIEmbeddings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain.chains import (
    ConversationalRetrievalChain,
)
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.chat_models import ChatOpenAI
from langchain.prompts.chat import (
    ChatPromptTemplate,
    SystemMessagePromptTemplate,
    HumanMessagePromptTemplate,
)
from langchain.docstore.document import Document
from langchain.memory import ChatMessageHistory, ConversationBufferMemory
from dotenv import load_dotenv

import chainlit as cl

load_dotenv()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)

system_template = """Answer the question based only on the following context. If you cannot answer the question with the context, please respond with 'I don't know':

Context:
{context}

Question:
{question}
----------------
{summaries}"""
messages = [
    SystemMessagePromptTemplate.from_template(system_template),
    HumanMessagePromptTemplate.from_template("{question}"),
]
prompt = ChatPromptTemplate.from_messages(messages)
chain_type_kwargs = {"prompt": prompt}


@cl.on_chat_start
async def on_chat_start():
    msg = cl.Message(
        content=f"Processing Nvidia.pdf..."
    )
    await msg.send()

    # Load PDF directly from local data directory
    pypdf_loader = PyPDFLoader("data/nvidia.pdf")
    texts = pypdf_loader.load_and_split()
    texts = [text.page_content for text in texts]

    # Create metadata for each chunk
    metadatas = [{"source": f"{i}-pl"} for i in range(len(texts))]

    # Create a FAISS vectorstore
    embeddings = OpenAIEmbeddings()
    docsearch = await cl.make_async(FAISS.from_texts)(
        texts, embeddings, metadatas=metadatas
    )

    message_history = ChatMessageHistory()

    memory = ConversationBufferMemory(
        memory_key="chat_history",
        output_key="answer",
        chat_memory=message_history,
        return_messages=True,
    )

    # Create a chain that uses the FAISS vectorstore
    chain = ConversationalRetrievalChain.from_llm(
        ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0, streaming=True),
        chain_type="stuff",
        retriever=docsearch.as_retriever(),
        memory=memory,
        return_source_documents=True,
    )

    # Let the user know that the system is ready
    msg.content = f"Processing Nvidia.pdf done. You can now ask questions!"
    await msg.update()

    cl.user_session.set("chain", chain)


@cl.on_message
async def main(message):
    chain = cl.user_session.get("chain")  # type: ConversationalRetrievalChain
    cb = cl.AsyncLangchainCallbackHandler()

    res = await chain.acall(message.content, callbacks=[cb])
    answer = res["answer"]
    source_documents = res["source_documents"]  # type: List[Document]

    text_elements = []  # type: List[cl.Text]

    if source_documents:
        for source_idx, source_doc in enumerate(source_documents):
            source_name = f"source_{source_idx}"
            # Create the text element referenced in the message
            text_elements.append(
                cl.Text(content=source_doc.page_content, name=source_name)
            )
        source_names = [text_el.name for text_el in text_elements]

        if source_names:
            answer += f"\nSources: {', '.join(source_names)}"
        else:
            answer += "\nNo sources found"

    await cl.Message(content=answer, elements=text_elements).send()