🔍 Add Retriever Logic using FAISS Vector Store

[Messycodess]

🚀 Summary

This PR adds the retriever component responsible for searching relevant documents using FAISS vector store, based on user role and query embeddings.

🛠️ Features

• Load FAISS index per role
• Search top-k documents using query embeddings
• Return role-specific documents with similarity scores
• Preloaded vector indices for roles: HR, Marketing, Finance, Engineering, General

📁 New Files

• app/services/vector_store.py
• app/services/retriever_service.py
• vector_data/*.pkl and *.faiss

✅ Next Step

Start llm-integration branch for connecting retrieved docs to LLM response generation.