We added structure-aware query handling to the Agentic RAG path. The goal was to stop treating every user request as a flat search string.
This gave the pipeline room to separate simple lookups from multi-part questions that need decomposition, context tracking, or stricter evidence selection. It did not replace retrieval; it made retrieval decisions more explicit before the system fetched context.
A public-safe engineering status note for Cortagent Agentic RAG: routing, retrieval, evidence selection, memory, caching, safety, and evaluation remain the active surface.
We clarified the boundary between conversation memory, retrieved evidence, and answer synthesis in the Agentic RAG runtime.
We shaped Agentic RAG evaluation cases around follow-ups, weak evidence, source targeting, and complex multi-part questions.