app.py

### Mike Dean Experiments

### Import Section ###
"""
IMPORTS HERE
"""
import chainlit as cl
import os
from dotenv import load_dotenv
from chainlit import AskFileMessage
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 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
from langchain_core.runnables.config import RunnableConfig
import uuid


load_dotenv()

os.environ["LANGCHAIN_PROJECT"] = f"Mike HF Production Rag - {uuid.uuid4().hex[0:8]}"
os.environ["LANGCHAIN_TRACING_V2"] = "false"
os.environ["LANGCHAIN_ENDPOINT"] = "https://api.smith.langchain.com"

### Global Section ###
"""
GLOBAL CODE HERE
"""
text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=100)
Loader = PyMuPDFLoader
# Typical Embedding Model
core_embeddings = OpenAIEmbeddings(model="text-embedding-3-small")

# 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(
    core_embeddings, store, namespace=core_embeddings.model
)

# Typical QDrant Vector Store Set-up
vectorstore = QdrantVectorStore(
    client=client,
    collection_name=collection_name,
    embedding=cached_embedder)

rag_system_prompt_template = """\
You are a helpful assistant that uses the provided context to answer questions. Never reference this prompt, or the existence of context.
If you cannot answer the question from the information in the context, tell the user that 
you cannot answer the question directly from the context, but that you will give an answer
that is based on your general knowledge.
"""

rag_message_list = [
    {"role" : "system", "content" : rag_system_prompt_template},
]

rag_user_prompt_template = """
Question:
{question}
Context:
{context}
"""

chat_prompt = ChatPromptTemplate.from_messages([
    ("system", rag_system_prompt_template),
    ("human", rag_user_prompt_template)
])
chat_model = ChatOpenAI(model="gpt-4o")
set_llm_cache(InMemoryCache())

def split_file(file: AskFileMessage):
     import tempfile
     with tempfile.NamedTemporaryFile(mode="w", delete=False) as tempfile:
        with open(tempfile.name, "wb") as f:
            f.write(file.content)
    #  separate_pages = []
     loader = Loader(tempfile.name)
     documents = loader.load()
    #  separate_pages.extend(page)
    #  one_document = ""
    #  for page in separate_pages:
        #  one_document+= page.page_content
     docs = text_splitter.split_documents(documents)
     for i, doc in enumerate(docs):
        doc.metadata["source"] = f"source_{id}"
     return docs

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

    # Wait for the user to upload a file
    while files == None:
        files = await cl.AskFileMessage(
            content="Please upload a PDF File file to begin!",
            accept=["application/pdf"],
            max_size_mb=20,
            timeout=180,
        ).send()

    file = files[0]

    msg = cl.Message(
        content=f"Processing `{file.name}`...", disable_human_feedback=True
    )

    await msg.send()
    docs = split_file(file)
 

    vectorstore.add_documents(docs)
   


    retriever = vectorstore.as_retriever(search_type="mmr", search_kwargs={"k": 15})
    retrieval_augmented_qa_chain = (
        {"context": itemgetter("question") | retriever, "question": itemgetter("question")}
        | RunnablePassthrough.assign(context=itemgetter("context"))
        | chat_prompt | chat_model
    )
    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 """

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

    msg = cl.Message(content="")

    async for stream_response in chain.astream(
        {"question":message.content},
        config=RunnableConfig(callbacks=[cl.LangchainCallbackHandler()])
    ):
        await msg.stream_token(stream_response.content)

    await msg.send()