Ai.sle6
Local RAG system for natural language retail product discovery
A fully local Retrieval-Augmented Generation (RAG) system for retail product discovery. The system indexes 120 products across 8 departments and supports natural language queries via similarity search with weighted attribute embeddings — all running locally without sending data to external APIs.
Query routing, similarity search, and LLM inference all run on-device using Llama 3.2 via Ollama, making it a privacy-conscious approach to conversational product search.
Tech Stack: Python, LanceDB, Sentence Transformers (all-MiniLM-L6-v2), Llama 3.2, Ollama, Flask