README.md

Advanced Conversational AI for E-Commerce: V1 vs. V2

This repository showcases the evolution of a RAG-based chatbot, demonstrating a clear progression from a self-hosted prototype (v1) to a professional, API-driven conversational AI (v2).

---

V2: The Professional Multilingual Assistant (API-Driven)

This version represents a modern, production-ready system built on a decoupled architecture with full support for English and Malay.

• Architecture (/v2_multilingual_api):
  • Backend: A secure FastAPI server implementing an advanced Retrieve-and-Re-rank pipeline. It uses multilingual models, Pinecone for vector storage, and the Google Gemini API for generation.
  • Frontend: A responsive Streamlit chat interface.
• Knowledge Base: /knowledge_base/v2_multilingual

---

V1: The Self-Hosted Prototype (Malay-Only)

This version is the original project, serving as a technical baseline.

• Architecture (/v1_malay_selfhosted): A single Streamlit application using a self-hosted mt5-small model and a local FAISS index.
• Knowledge Base: /knowledge_base/v1_malay