Problem
Provider model retirements break unattended deployments on a schedule nobody in the deployment chose: Anthropic retires models with ≥60 days' notice (requests then hard-fail), OpenAI ≥6 months for GA models. Today a retired or renamed model id means silently failing detection jobs plus NULL-cost accounting. No reference OSS product surfaces this well — for a product whose value is unattended background LLM jobs, it is part of the compatibility promise.
What
Dependencies
#1369 (alert channel). Builds on #1367 (which owns the classification + alert); this issue owns the admin-visible banner, badge staleness, and docs.
Part of #1354.
Problem
Provider model retirements break unattended deployments on a schedule nobody in the deployment chose: Anthropic retires models with ≥60 days' notice (requests then hard-fail), OpenAI ≥6 months for GA models. Today a retired or renamed model id means silently failing detection jobs plus NULL-cost accounting. No reference OSS product surfaces this well — for a product whose value is unattended background LLM jobs, it is part of the compatibility promise.
What
Dependencies
#1369 (alert channel). Builds on #1367 (which owns the classification + alert); this issue owns the admin-visible banner, badge staleness, and docs.
Part of #1354.