How would you design a Retrieval-Augmented Generation pipeline to minimize hallucination when querying dense technical manuals for Trunk Tools?
