FEED 1.8.0

FEED 1.8.0 releases the revised AI model support and moves the server to Node 24.

• Refreshed, shared model catalogue with current provider SDKs, thinking controls, language coverage, model-expiry warnings, and save-time entitlement verification.
• Clearer provider errors and administrator alerts, including actionable handling of unavailable models and temporary provider limits.
• Working Fable document classification and more accurate spending records for billed failures and parallel classification batches.
• Finalized in-app release notes without the confusing “Unreleased” heading.

Validation: backend 1,299 tests passed (2 skipped); frontend 578 tests passed. Both ARM64 production images build successfully. Development feature checks covered Fable classification, Spanish DOCX translation, Arabic shopping-list PDF export, long-document classification, and a 40-item bulk retry.

Production AI tests were intentionally omitted at the operator's request. Fresh API keys and production model configuration are handled by the operator. Deployment checks cover running image identity, health, migrations, and matching version labels.

Docker images: et2geiger/feed-backend:1.8.0 and et2geiger/feed-frontend:1.8.0 (linux/arm64); both are also published as latest.