We focused on making the platform more stable and reliable. If you’ve experienced unexpected disconnections with Twitter/LinkedIn integrations or noticed scheduled tasks not running on time, this update addresses those issues.
OAuth Integrations Are Now More Reliable
We fixed three core issues affecting OAuth authentication stability:
No More Disconnections from Temporary Errors
Previously, if the Twitter API returned a 403 error (usually temporary rate limiting), we would immediately mark the entire integration as invalid. Now the system distinguishes between “real authentication failures” (401) and “temporary access issues” (403). This means your social media integrations won’t disconnect just because an API has a brief hiccup.
Concurrent Requests No Longer Conflict
When multiple tasks used the same OAuth token simultaneously, we had race conditions that caused token refresh failures. We’ve optimized the token refresh locking mechanism, so you can now safely use Twitter and LinkedIn integrations concurrently.
Automatic Detection and Notification
When there’s a real authentication issue (like a manually revoked token), the system automatically marks the integration as invalid and notifies you to re-authorize. No more silent failures.
Scheduled Tasks Finally Run on Time
Fixed a critical bug: interval-based scheduled tasks (like “every 2 days” or “every 12 hours”) weren’t triggering at all. The issue was in database field initialization and time unit parsing. Now all periodic tasks run on schedule.
If you previously set up interval tasks that never ran, they should work now.
Other Improvements
- Cleaner Artifact System: Removed the task_id field (since one artifact can be used by multiple tasks), added a refresh button, and support for video embedding
- Unified Email Configuration: Resolved inconsistent email URLs across different contexts
- Simplified Code Architecture: Cleaned up redundant components in the base agent, making the system easier to maintain