Fix: Permanent model override for decommissioned Groq models in AgentRunnerService (Phase 9)

Phase 7: Self-Optimizing Agents (Feedback analysis on rejection, lesson injection on retry)

Phase 7: Integrated Vectorized Long-Term Memory (Index on Approval, Retrieve on Run)

Harden report quality checks and fix Space version loading

Add guided mode, profiles admin, and output quality guardrails

Update Hugging Face Space

Deploy Aubm Docker Space