Bertos Refrigerators Company RAG

This is a production-grade RAG build for Bertos Refrigerators, turning their roughly 120-page commercial catalog into something a customer or sales rep can actually query in plain language instead of scanning a PDF for one spec.

The pipeline chunks the catalog by model and spec section, embeds it into a vector store, and answers pricing, dimension, and capacity questions by retrieving the specific passage the question is actually about, then citing it, rather than letting a model guess at a number from general training data. Getting this right was mostly chunking strategy and retrieval tuning, since a catalog's value is in exact figures that can't be paraphrased loosely.

The result is a Q&A layer that answers a specific spec or price from the real catalog page instead of an approximate guess, cutting the time a customer or rep spends hunting through a 120-page document for one number.