v1.1.0
Memory decision layer (salience, supersede, contradiction, offline judge), hybrid specialist-prompt format (ADR-0037), research-engagement gate, output-style rebalance, OpenCode/local-models/docling work. Security: hono 4.12.25 (high CVE). See CHANGELOG [1.1.0] and the Upgrade notes.