src/retrival_generation.py

import os
import logging
import warnings
from dotenv import load_dotenv
from src.data_preprocessing import DataSplitting
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_groq import ChatGroq
from langchain_community.vectorstores import FAISS
from langchain.prompts import PromptTemplate
from langchain.schema import StrOutputParser
from langchain.schema.runnable import RunnableParallel, RunnablePassthrough
from langchain.retrievers import ContextualCompressionRetriever
from langchain.retrievers.document_compressors.cross_encoder_rerank import CrossEncoderReranker
from langchain_community.cross_encoders import HuggingFaceCrossEncoder

warnings.filterwarnings("ignore")

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(name)s - %(message)s",
    handlers=[logging.StreamHandler()]
)


logger = logging.getLogger("RetrievalGeneration")

load_dotenv()
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
if not GROQ_API_KEY:
    raise ValueError("GROQ_API_KEY not found in environment variables.")

llm = ChatGroq(model=os.getenv("GROQ_MODEL", "llama-3.1-8b-instant"))


class RetrievalGeneration:
    def __init__(self, vectorstore_path: str = "faiss_store"):
        self.vectorstore_path = vectorstore_path
        self.vectorstore = None
        self.rag_chain = None
        logger.info("RetrievalGeneration initialized with path: %s", vectorstore_path)

    def init_vectorstore(self, rebuild: bool = False):
        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
        logger.info("Embeddings model loaded.")

        if os.path.exists(self.vectorstore_path) and not rebuild:
            logger.info("Loading existing FAISS index from: %s", self.vectorstore_path)
            self.vectorstore = FAISS.load_local(
                self.vectorstore_path, embeddings, allow_dangerous_deserialization=True
            )
        else:
            logger.warning("Building new FAISS index...")
            chunks = DataSplitting(chunk_size=2000, chunk_overlap=800).chunking()
            logger.info("Data split into %d chunks", len(chunks))
            self.vectorstore = FAISS.from_documents(chunks, embeddings)
            self.vectorstore.save_local(self.vectorstore_path)
            logger.info("FAISS index saved at: %s", self.vectorstore_path)

        return self.vectorstore

    def build_rag_chain(self, k: int = 10,top_n: int = 5):
        if not self.vectorstore:
            raise ValueError("Vectorstore not initialized. Run init_vectorstore() first.")

        logger.info("Creating retriever from FAISS vectorstore (top_k=%d)...", k)
        cross_encoder_model = HuggingFaceCrossEncoder(model_name="cross-encoder/ms-marco-MiniLM-L-6-v2")

        compressor = CrossEncoderReranker(model=cross_encoder_model, top_n=top_n)  

        retriever = ContextualCompressionRetriever(base_compressor=compressor,base_retriever=self.vectorstore.as_retriever(search_kwargs={"k": k}))


        prompt = PromptTemplate(
        template="""
You are a professional and concise AI assistant that answers questions
about the career, education, skills, projects, certifications, and professional 
background of **Katta Sai Pranav Reddy**.

Your job is to:
- Use ONLY the provided context to answer.
- Be recruiter-friendly: structured, clear, and professional in tone.
- If the question is unrelated to Katta Sai Pranav Reddy’s professional profile, 
  politely decline by saying: 
  "I can only answer questions related to the professional background of Katta Sai Pranav Reddy."
- If the context does not provide enough information, say: 
  "The available information does not cover that detail."

Context:
{context}

Question:
{question}

Answer (clear, structured, recruiter-focused):
""",
        input_variables=["context", "question"]
)

        self.rag_chain = (
            RunnableParallel({
                "context": retriever,
                "question": RunnablePassthrough()
            })
            | prompt
            | llm
            | StrOutputParser()
        )

        logger.info("RAG chain with reranking successfully built.")
        return self.rag_chain