Retrieval-Augmented Generation

Overview

Retrieval-Augmented Generation (RAG) is a technique that combines a language model's generative capabilities with external document retrieval, allowing the model to answer questions grounded in a specific knowledge base rather than relying solely on its training data. Foxconn's FoxBrain team deployed a hybrid RAG architecture for internal office workloads, combining two complementary retrieval strategies:

• Document-RAG — extracts facts from documents and generates an initial answer via vector search and full-text retrieval.
• Graph-RAG — traverses a knowledge graph to infer deeper relational insights unavailable from direct document lookup.

The FoxBrain RAG system's knowledge base spans three Smart Platforms — Smart EV, Smart Manufacturing, and Smart Cities — backed by three data store types: Document Store, Vector Store, and Knowledge Graph. This hybrid approach allows the system to answer both direct factual questions (handled well by Document-RAG) and complex multi-hop relational queries (requiring Graph-RAG traversal).