src/services/read_pdf.py

import os
from pyvi.ViTokenizer import tokenize
from langchain_community.document_loaders import PyPDFLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
import pymongo
from generate_embedding import generate_embedding

os.environ["OPENAI_API_KEY"] = "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"

# Connect DB
client = pymongo.MongoClient(
    "mongodb+srv://rag:p9vojYc9fafYwxE9@rag.xswi7nq.mongodb.net/?retryWrites=true&w=majority&appName=RAG"
)

db = client.rag
collection = db.pdf


def insertData(chunk):
    return collection.insert_many(chunk)


def deleteByUserId(user_id: str):
    return collection.delete_many({"user_id": user_id})


def readFromPDF():
    # load PDF
    loader = PyPDFLoader("data/cds.pdf")
    pages = loader.load_and_split()
    pages = list(filter(lambda page: page.metadata['page'] >= 10, pages))

    text_splitter = RecursiveCharacterTextSplitter(chunk_size=768, chunk_overlap=200)
    chunks = text_splitter.split_documents(pages)
    items = []
    for index, chunk in enumerate(chunks):
        print(index)
        items.append({"page_content": chunk.page_content, "index": index})
    return items


def indexData(user_id: str):
    items = readFromPDF()
    contents = []
    for item in items:
        tokenized_page_content = tokenize(item["page_content"])
        content = {
            "page_content": item["page_content"],
            "page_content_embedding": generate_embedding(tokenized_page_content),
            "user_id": user_id,
            "index": item["index"],
        }
        contents.append(content)
    deleteByUserId(user_id)
    insertData(contents)


indexData("cds.pdf")

# prompt = hub.pull("rlm/rag-prompt")
# llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)

# def format_docs(docs):
#     return "\n\n".join(doc.page_content for doc in docs)


# rag_chain = (
#     {"context": retriever | format_docs, "question": RunnablePassthrough()}
#     | prompt
#     | llm
#     | StrOutputParser()
# )