src/rag_pipeline.py

from langchain.chains import RetrievalQA
from langchain_google_genai import GoogleGenerativeAI
import re
from src.config import GOOGLE_API_KEY
from src.database import load_vector_database

llm = GoogleGenerativeAI(
        model="gemini-2.0-flash",
        api_key=GOOGLE_API_KEY
    )

vector_store = load_vector_database()
qa = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=vector_store.as_retriever()
)

def preprocess_text(text):
    text = text.lower()
    text = re.sub(r'[^a-zA-Z0-9\s]', '', text)
    return text

def get_answer(question, use_rag=False):
    query = preprocess_text(question)

    if use_rag:
        response = qa.invoke(query)
    else:
        response = llm.invoke(query)

    return response