Chronos Scheduler changelog

Added the AI agent run task type (ai_agent_run) for schedules, chained events, platform events, and webhook-triggered events;

AI agent run tasks execute a published agent by its required agent_alias, with optional agent_version; pipeline_id is optional;

Added allow-listed scalar payload interpolation for prompt_template, agent_input, and idempotency_key;

Added a 64 KiB limit for trigger fire request bodies and loop protection for AI run platform events;

Fixed a bug where the SDK token refresh on 401 would reuse the same stale token instead of performing a fresh login handshake — the SDK client is now properly disconnected and recreated on reconnect;

Added a periodic Core Hub connection health check that calls GET /authentication/me every 5 minutes (configurable) and automatically triggers a force-reconnect when the token is rejected or Core Hub is unreachable;

Added CHRONOS_HEALTH_CHECK_ENABLED, CHRONOS_HEALTH_CHECK_INTERVAL_SECONDS, and CHRONOS_HEALTH_CHECK_TIMEOUT_SECONDS environment variables;

Added a warning log when the refreshed token is identical to the rejected token, making stale-connection issues easier to diagnose;

Added the Query Studio action type (query_studio) for schedules, chained events, platform events, and webhook-triggered events;

Query Studio tasks can run custom SQL or a saved query against a selected pipeline agent; execution goes through CoreHub Query Studio (120s timeout) and is never placed on the crontab command line;

Writable Query Studio execution requires an explicit query_read_only opt-in;

Fixed job creation failing with HTTP 500 when the UI sent query_read_only: null for non-Query-Studio jobs (null is now treated as read-only);

Added origin vs broadcast routing for platform-event trigger flows — origin loops the action back to the event source, broadcast fans out as before; table reorg, truncate, and DDL default to origin, cluster-wide events default to broadcast;

Pause-before-redo now treats Warning as settled (alongside Hold and Error), so redo proceeds when entities report errorState.severity=WARNING;

Redo operations (entity, pipeline, group) now pause CoreHub (stop / stop-group) and wait until every affected entity is Hold before issuing the redo; no more fixed sleep;

If the pause fails or Hold is not reached before timeout, the redo is aborted;

Fixed an issue where the trigger_flows table was created without the platform_event column on a fresh database, causing GET /api/triggers/ to return a 500 error at container bootstrap time;

Breaking change: this version is only compatible with Core Hub 2.2.10+ onwards;

Added chained events with sync/async execution modes and configurable webhook timeouts;

Added webhook triggers for external events;

Added platform-bound events via persistent webhook registration in CoreHub with retry and filter support;

All API endpoints are now secured with JWT authentication via CoreHub introspection;

Added role-based access control (SUPER_ADMIN, MANAGER, VIEWER, EXTERNAL_MODULE);

Optimized job loading with batch queries and cached timezone resolution;

GSSD-1195: Chronos now automatically refreshes its own auth token after core hub’s session expires;

Swapped Linux base image from Debian to Chainguard/Wolfi for improved security and smaller footprint;

Fixed an issue with connectivity loss towards Core Hub’s websockets;

GSSD-976: Fixed an issue with the scheduler not firing Snapshot events over Groups;

PythonSDK version bump;

Improved reconnect to Core Hub reliability;

Security patches;

PythonSDK version bump;

Improved reconnect to Core Hub reliability;

Quarterly security update;

GSSD-870: Fixed the scheduler event loop getting stuck if multiple concurrent schedules are fired;

Improved artifacts management for offline mode in Windows environments;

Added support for pipeline maintenance mode scheduling;

GSSD-697: Redo not firing snapshot;

Support for new schedule types (redo api);

Improved reconnection behavior when Core Hub ip address changes;

Fixed a regression caused by too strict user permissions in Linux containers;

GSSD-608: System time was winning over the user-selected timezone;

Docker image now uses a unprivileged user for security reasons;

GSSD-442: Fixed timezone management from ENV var "TZ";

Improved CORE_HUB_URL handling;

Moved tawards new /opt/gluesync/shared volume mount (while maintaining backward compatibility);

Updated Python SDK to the latest version 0.11;

Re-send license headers at every reconnection to the Core Hub;

When setting an hardcoded CORE_HUB_URL in the environment variables, Chronos will still attempt to retry connection to the Core Hub with exponential backoff;

GS SDK updated to the latest version 0.11;

File logging (now enabled by default, 1Gb Max size, 10 files);

Major refactor an env vars management to support Nanoserver images and Windows containers, especially legacy 2019;

Libs upgrades;

Support for nanoserver images;

Support for nanoserver images;

Fixed an issue that was causing Chronos to crash if the Core Hub APIs where responding 404;

Updated websockets library to use the latest version;

Auto set targeted objects in a scheduled jobs to STOP before executing a Snapshot task;

Version control;

New ENV variable: FIRE_ONCE;

Fixed an issue that could prevent the Scheduler icon from being displayed in the Gluesync UI;

GSSD-338: fixed an issue that could cause one-shot-snapshots targeting a group to fail;

Official support for Windows-based containers;

GSPRJ-1123: fixed a bug that was causing last_run, next_run and start_time to work on different timezones formats;

CORE_HUB_URL has now become GLUESYNC_HOST environment variable to be consistent with other modules;

Fixed an issue that could cause scheduled jobs with "snaphost only mode" to trying to run a snapshot + cdc instead;

CI/CD improvements;

Use Pip’s provided Gluesync python SDK;

Support for groups' scheduling;

CI/CD improvements;

Removed logs;

Fixed an issue that could cause the scheduler to crash if after several attempts to connect to the Core Hub it fails to schedule a task;

Chronos now provides evidence whether a job has been scheduled via a Cron-like expression or a human readable format;

Fixed an issue that could cause the scheduler to continue using an outdated Core Hub URL after CoreHub’s restart;

Fixed an issue that could cause the scheduler to continue using an outdated Core Hub API token after CoreHub’s restart;

Improved error handling, now properly forwarded to API and UI in case of a schedule failure;

Enhanced logging for better debugging experience;

Minor bugfixes and improvements.