Beyond RAG: A Relational Context Engine That Cuts Token Burn — Unblocked

Why watch

Peter Werry demonstrates how Unblocked's relational context engine outperforms vector search for coding agents by answering structural questions like PR history that RAG cannot handle. The talk shows ~50% token cost reduction in Claude sessions and walks through building an open-source document query engine with schema discovery, identity resolution, query validation, and full-text search. Targeted at AI engineers building context-aware coding agents who need to retrieve work history beyond just code.