MassGen v0.1.52 adds quality gates, novelty injection, and provenance for multi-agent answers
MassGen v0.1.52 introduces a final answer modal, quality gates, novelty injection, and answer provenance to tighten multi-agent AI workflows. The [release](https://github.com/massgen/MassGen/releases/tag/v0.1.52) debuts a dedicated Final Answer Modal with tabs for rendered answers, post-eval signals, file lists, and a review workspace with diffs, plus an action bar to Approve/Reject/Cancel with rework controls. A new Substantive Gate halts coordination when only incremental changes are planned, classifying outputs as transformative/structural/incremental and detecting decision_space_exhausted to stop low-value loops. Novelty Injection applies configurable “creative pressure” (none/gentle/moderate/aggressive) when agents converge, escalating after restarts, while answer provenance tags (e.g., agent1.2) and label mapping make lineage auditable in the UI. The build also changes first-answer behavior so initial responses don’t auto-restart on quality-check failure, reducing churn in early iterations.