app.py

### Import Section ###
"""
IMPORTS HERE
"""
import os
import uuid
from dotenv import load_dotenv
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyMuPDFLoader
from qdrant_client import QdrantClient
from qdrant_client.http.models import Distance, VectorParams
from langchain_openai.embeddings import OpenAIEmbeddings
from langchain.storage import LocalFileStore
from langchain_qdrant import QdrantVectorStore
from langchain.embeddings import CacheBackedEmbeddings
from langchain_core.prompts import ChatPromptTemplate
from chainlit.types import AskFileResponse
from langchain_core.globals import set_llm_cache
from langchain_openai import ChatOpenAI
from langchain_core.caches import InMemoryCache
from operator import itemgetter
from langchain_core.runnables.passthrough import RunnablePassthrough
import chainlit as cl
from langchain_core.runnables.config import RunnableConfig
from langchain_community.llms import HuggingFaceEndpoint
from langchain_huggingface.embeddings import HuggingFaceEndpointEmbeddings
from langchain_core.prompts import PromptTemplate

import numpy as np
from numpy.linalg import norm

load_dotenv()

### Global Section ###
"""
GLOBAL CODE HERE
"""

RAG_PROMPT_TEMPLATE = """\
<|start_header_id|>system<|end_header_id|>
You are a helpful assistant. You answer user questions based on provided context. If you can't answer the question with the provided context, say you don't know.<|eot_id|>

<|start_header_id|>user<|end_header_id|>
User Query:
{query}

Context:
{context}<|eot_id|>

<|start_header_id|>assistant<|end_header_id|>
"""
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
hf_llm = HuggingFaceEndpoint(
    endpoint_url=f"{os.environ["YOUR_LLM_ENDPOINT_URL"]}",
    max_new_tokens=512,
    top_k=10,
    top_p=0.95,
    typical_p=0.95,
    temperature=0.01,
    repetition_penalty=1.03,
    huggingfacehub_api_token=os.environ["HF_TOKEN"]
)
hf_embeddings = HuggingFaceEndpointEmbeddings(
    model=os.environ["YOUR_EMBED_MODEL_URL"],
    task="feature-extraction",
    huggingfacehub_api_token=os.environ["HF_TOKEN"],
)

rag_prompt = PromptTemplate.from_template(RAG_PROMPT_TEMPLATE)
rag_chain = rag_prompt | hf_llm

def cosine_similarity(phrase_1, phrase_2):
  vec_1 = hf_embeddings.embed_documents([phrase_1])[0]
  vec2_2 = hf_embeddings.embed_documents([phrase_2])[0]
  return np.dot(vec_1, vec2_2) / (norm(vec_1) * norm(vec2_2))

def process_file(file: AskFileResponse):
    import tempfile

    with tempfile.NamedTemporaryFile(mode="w", delete=False) as tempfile:
        with open(tempfile.name, "wb") as f:
            f.write(file.content)

    Loader = PyMuPDFLoader

    loader = Loader(tempfile.name)
    documents = loader.load()
    docs = text_splitter.split_documents(documents)
    for i, doc in enumerate(docs):
        doc.metadata["source"] = f"source_{i}"
    return docs


### On Chat Start (Session Start) Section ###
@cl.on_chat_start
async def on_chat_start():
    """ SESSION SPECIFIC CODE HERE """
    files = None

    while files == None:
        # Async method: This allows the function to pause execution while waiting for the user to upload a file,
        # without blocking the entire application. It improves responsiveness and scalability.
        files = await cl.AskFileMessage(
            content="Please upload a PDF file to begin!",
            accept=["application/pdf"],
            max_size_mb=20,
            timeout=180,
            max_files=1
        ).send()

    file = files[0]
    msg = cl.Message(
        content=f"Processing `{file.name}`...",
    )
    await msg.send()
    docs = process_file(file)

    # Typical QDrant Client Set-up
    collection_name = f"pdf_to_parse_{uuid.uuid4()}"
    client = QdrantClient(":memory:")
    client.create_collection(
        collection_name=collection_name,
        vectors_config=VectorParams(size=1536, distance=Distance.COSINE),
    )

    # Adding cache!
    store = LocalFileStore("./cache/")
    cached_embedder = CacheBackedEmbeddings.from_bytes_store(
        hf_embeddings, store, namespace=hf_embeddings.model
    )

    # Typical QDrant Vector Store Set-up
    vectorstore = QdrantVectorStore(
        client=client,
        collection_name=collection_name,
        embedding=cached_embedder)
    
    for i in range(0, len(docs), 32):
        if i == 0:
            vectorstore = docs.from_documents(docs[i:i+32], hf_embeddings)
            continue
        vectorstore.add_documents(docs[i:i+32])

    retriever = vectorstore.as_retriever(search_type="mmr", search_kwargs={"k": 3})

    retrieval_augmented_qa_chain = (
        {"context": itemgetter("question") | retriever, "question": itemgetter("question")}
        | RunnablePassthrough.assign(context=itemgetter("context"))
        | rag_prompt | hf_llm
    )

    # Let the user know that the system is ready
    msg.content = f"Processing `{file.name}` done. You can now ask questions!"
    await msg.update()

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

### Rename Chains ###
@cl.author_rename
def rename(orig_author: str):
    """ RENAME CODE HERE """
    rename_dict = {"ChatOpenAI": "the Generator...", "VectorStoreRetriever": "the Retriever..."}
    return rename_dict.get(orig_author, orig_author)

### On Message Section ###
@cl.on_message
async def main(message: cl.Message):
    """
    MESSAGE CODE HERE
    """
    runnable = cl.user_session.get("chain")

    msg = cl.Message(content="")

    # Async method: Using astream allows for asynchronous streaming of the response,
    # improving responsiveness and user experience by showing partial results as they become available.
    async for chunk in runnable.astream(
        {"question": message.content},
        config=RunnableConfig(callbacks=[cl.LangchainCallbackHandler()]),
    ):
        await msg.stream_token(chunk.content)

    await msg.send()