Smart Assistant - Context-Aware RAG System
A domain-specific Retrieval-Augmented Generation (RAG) engine designed to automate customer support for E-commerce businesses. This system allows users to upload policy documents (PDFs) to create an instant AI assistant that answers questions accurately based strictly on the provided documents.
Key Features
Context-Awareness (Conversational Memory): The assistant understands follow-up questions and references to previous context (e.g., remembering which product or tier was discussed).
Zero-Hallucination Policy: The system is configured with temperature=0.0 and strict prompts to ensure it only answers using the provided context, preventing the invention of false information.
Fast Semantic Search: Uses advanced HuggingFace embeddings (bge-small-en-v1.5) for rapid and accurate retrieval.
User-Friendly Interface: Built with Streamlit for a smooth, interactive chat experience.
Tech Stack
LLM: Google Gemini 2.5 Flash
Embeddings: HuggingFace (BAAI/bge-small-en-v1.5)
Vector Database: ChromaDB
Orchestration: LangChain
Frontend: Streamlit