Retrieval is not the same as evidence. We started tightening the boundary between candidate chunks and the context that should actually be used for synthesis.
This keeps the RAG path inspectable. Retrieval can return many candidates, but evidence selection needs its own reasoning and failure modes.
We tightened how retrieval diagnostics, evidence boundaries, and answer synthesis traces connect in the Agentic RAG loop.
We clarified the boundary between conversation memory, retrieved evidence, and answer synthesis in the Agentic RAG runtime.
Ingestion guardrails were added to make document handling and chunk quality more explicit before retrieval uses the content.