Smarter Automation Rejections and Clearer Blockers

This release makes automation decisions more reliable and easier to understand. It also preserves more article content during extraction, reducing cases where items need reprocessing.

Automation that rejects the right content

Pending articles with explicit publisher paywalls are now automatically rejected, including items that were already waiting for a decision. This helps keep queues clean and prevents time spent reviewing content that cannot be published.

When running in full automation, the system can now reject unrelated and non-article content without requiring publication-ready summaries or topic matches. This reduces unnecessary work and speeds up end-to-end processing for clearly ineligible items.

Articles affected by earlier extraction limits and classification validation errors can now be recovered with a corrected processing pass. This helps previously stuck items move forward without manual intervention.

Clearer blockers and fewer confusing admin actions

Publication blockers are now more precise, including paywalls, missing topic evidence, description-only text, date exclusions, and unavailable approved sources. This makes it easier to see exactly why something is blocked and what would need to change.

The admin interface now reports successful rejection decisions more clearly, and it hides repeat-analysis actions when the available topic evidence has not changed. You get a cleaner workflow with fewer dead-end actions.

Extraction now preserves longer article bodies and code examples, follows verified publisher redirects, and keeps richer feed summaries for analysis. This improves content quality and reduces avoidable rework caused by incomplete extraction.

Agent approval is also simpler, with the internal OAuth callback address hidden. Chrome and Edge extension packages now include agent client icons for a more polished experience.