rag_node.py

from langchain_text_splitters import RecursiveCharacterTextSplitter
from pypdf import PdfReader
from vectorstore import insert_document, search_query, create_collection
from llm_node import refine_answer


def load_pdf(pdf_path: str):
    """Load PDF, split into chunks, and store embeddings in Qdrant."""
    reader = PdfReader(pdf_path)
    text = ""

    for page in reader.pages:
        extracted = page.extract_text()
        if extracted:
            text += extracted

    splitter = RecursiveCharacterTextSplitter(
        chunk_size=500,
        chunk_overlap=50
    )
    chunks = splitter.split_text(text)

    create_collection()

    for i, chunk in enumerate(chunks):
        insert_document(chunk, i)


def rag_answer(query: str) -> str:
    """
    Universal RAG Answering:
    - Handles factual questions
    - Handles step-by-step SOP questions
    - Handles summary / explanation questions
    - Works for B2B and non-B2B documents
    """
    docs = search_query(query)

    if not docs:
        return "No relevant information found in the document."

    context = "\n".join(docs)

    prompt = f"""
You are a professional AI assistant answering questions strictly using the given context.

Rules:
- Use ONLY the provided context.
- Do NOT repeat the full document.
- Be precise and useful.
- If the question asks for steps or a process, return a numbered list.
- If the question asks for a summary, provide a concise paragraph.
- If the question is factual, answer directly.
- Do NOT hallucinate.

Context:
{context}

Question:
{query}

Answer:
"""

    return refine_answer(prompt)