Is Treasure AI Down?

We have confirmed that the rollback is complete and the incident is resolved. --- [2026-11-09T23:08:24.004-08:00] (monitoring) A release deployed on November 9 caused some Bulk Load and Result Export jobs to fail. We have identified the root cause and completed the rollback. We are monitoring for any further failures. Affected Bulk Load jobs may contain javax.ws.rs.ClientErrorException: HTTP 409 Conflict in their logs. Affected Result Export jobs may report an unusually high number of invalid records.

The issue causing Amazon S3 Parquet imports to load date and time columns as text (string) instead of numbers (long) has been resolved. A fix was deployed on October 22, 2026 at 07:40 UTC, and new import jobs load these columns as numbers (long), as before October 13, 2026. Data imported between October 13, 2026 02:00 UTC and October 22, 2026 07:40 UTC may contain affected date and time columns. We will contact affected customers directly. We apologize for the inconvenience and thank you for your patience. --- [2026-10-22T00:44:03.436-07:00] (monitoring) We have deployed a fix for Amazon S3 Parquet imports loading date and time columns as text. New import jobs load these columns as numbers (long), as before October 13, 2026. We are monitoring to ensure stability. We will provide an update in 30 minutes. --- [2026-10-21T23:22:44.546-07:00] (identified) We have identified the cause of Amazon S3 Parquet imports loading date and time columns as text (string) instead of numbers (long) since October 13, 2026 02:00 UTC, and are deploying a fix. Until the fix is deployed, affected import jobs continue to complete with these columns loaded as text. Other columns and other data connectors are not affected. We are verifying data integrity for data imported during this period. We will provide an update in 30 minutes. --- [2026-10-21T21:06:37.687-07:00] (identified) We have identified the cause of Amazon S3 Parquet imports loading date and time columns as text (string) instead of numbers (long) since October 13, 2026 02:00 UTC, and are implementing a fix. Until the fix is deployed, affected import jobs continue to complete successfully with these columns loaded as text. Other columns and other data connectors are not affected. We are verifying data integrity for data imported during this period. We will provide an update in 1 hour. --- [2026-10-21T20:24:39.148-07:00] (identified) Since October 13, 2026 02:00 UTC, some import jobs that read Parquet files from Amazon S3 complete successfully but load date and time columns as text (string) instead of numbers (long). If the destination table already defines these columns as long, the imported values are empty (NULL). Only columns stored with a date or time type in the Parquet file are affected. Other columns and other data connectors are not affected. We have identified the cause, a change introduced in the October 13 release, and are preparing a fix. We are assessing the full scope of impact. We will provide an update in 30 minutes.

The recording transfer failures in Treasure AI Voice have been resolved. All systems are operating normally. Between approximately 22:55 UTC on October 6 and [fix time] UTC on October 7, new recordings could not be transferred from the mobile app. We deployed a server-side fix, and no app update is required. Recordings that failed to transfer during this period can be transferred again from the device. Thank you for your patience. --- [2026-10-06T21:54:16.872-07:00] (monitoring) We have deployed a server-side fix for the recording transfer failures in Treasure AI Voice. New transfers from the mobile app are succeeding again. No app update is required. Recordings that failed to transfer during the incident can be transferred again from the device. We are monitoring to ensure stability. --- [2026-10-06T21:27:41.324-07:00] (identified) We are continuing to work on a fix for this issue. --- [2026-10-06T21:21:03.468-07:00] (identified) Since approximately 22:55 UTC on October 6, users of the Treasure AI Voice mobile app may be unable to transfer new recordings from their device. The transfer fails with an "Unauthorized" error. This affects both Wi-Fi and Bluetooth transfers. We have identified the cause and are deploying a fix. Recordings that were already transferred, and their transcripts and summaries, are not affected. We will provide an update in 30 minutes.

On Monday, November 16 from 01:00 to 03:00 UTC, we will perform routine maintenance on the databases that power the Treasure AI Platform. During this maintenance, we will apply patches and updates to databases used by the Treasure AI Platform. These updates are necessary to ensure we have the latest security and performance improvements available. There may be a brief period of unavailability while we restart services, typically a few minutes. See below for information about the impact of this change to Treasure AI services. We will communicate with you on an ongoing basis via our status page (https://status.treasure.ai/). If you notice any problems outside this maintenance window, we encourage you to contact our support team for assistance (https://docs.treasure.ai/support/contacting-treasure-data-support). # Impact During the maintenance window and the recovery period that follows, the impact follows one rule: - Asynchronous processing keeps working, just slower. Ingestion, jobs, and message deliveries are queued and automatically retried, and complete on their own after maintenance. - Synchronous requests return errors (mostly 5xx). Retry them after we mark maintenance complete on our status page. Customers make experience the following: - Jobs execution delay All jobs (Presto Query Engine, Hadoop / Hive Query Engine, Data Connector Integrations, Bulk Import, Bulk Export, Result Export, and Partial Delete jobs submitted from Console, API, Workflow or triggered by our system according to the configured schedule) will be delayed and automatically retried until they succeed after the downtime period. We expect job processing to ramp up over 30 minutes after the maintenance is complete before returning to full throughput. If customers experience errors running Data Connector Integration Export jobs, please contact our Support team at support@treasure.ai to recover the jobs. To avoid data duplication in your systems from Treasure AI exports, Integration Export jobs won’t be automatically retried. - Console execution delay Data Workbench and Audience Studio will show that jobs are running longer due to job execution delays. The Table preview (sample data and records counter) in Data Workbench may show old information while maintenance is underway.  After maintenance is complete, we expect the processing of Data Workbench and Audience Studio jobs to recover gradually over 30 minutes. SQL or Workflow requests to create tables will be delayed and automatically retried after the maintenance period. - REST API The REST API endpoints to create tables, job submission, and job cancellation will be unavailable and will respond with a 500 error code. Components that use the REST API may also see performance degradation such as delays, authorisation failures, partial functionality, unavailability, and errors during the maintenance window. These components include “Web Interface”, “Mobile/Javascript REST API”, “Workflow”, “Workflow REST API”, “CDP API”, “Data Access API (beta)”, “ADL”, “Engage Studio REST APIs”, “Engage Studio Message Delivery”, “AI Agent Foundry, Foundry Workspace”, “Streaming Ingress / Egress connectors” Requests, workflows, and deliveries will be retried where safe to do so, and scheduled jobs will be processed after the maintenance period. - Data Connector Integrations connection settings During the maintenance, creating or updating Data Connector Integrations connections may fail. Please retry after the maintenance is complete. - AI Signals  During the maintenance window, requests to the ML Batch API used by AI Signals may temporarily fail, and machine learning training or prediction runs may not be created or executed. Treasure Workflow tasks that invoke the ML Batch API will retry automatically, so no customer action is expected for workflow-based runs. Customers that call the ML Batch API directly may receive failed requests during the maintenance window and should follow their normal retry handling after maintenance is complete. - AI Studio During the maintenance window, Treasure AI Studio users may experience errors when accessing chats, managing projects, or updating settings. What needs your attention - Campaigns scheduled during the window - check them and re-run any that failed. - Long-running workflows - workflows running longer than a few minutes may be terminated. Check and re-run them. - Integration Export jobs are not retried automatically, to avoid duplicating data in your systems. If one fails, contact support@treasure.ai to recover it. # Communication Beyond this notice, we will provide updates on our status page (https://status.treasure.ai/) approximately 1 hour before the beginning of the maintenance window, at the start and completion of the operation, and once the verification is complete. At that time, all systems will have returned to normal operations, and the Scheduled Maintenance will be closed. If you have any questions or concerns about this upgrade, please contact our Support team at support@treasure.ai.

On Tuesday, November 17 from 01:00 to 03:00 UTC, we will perform routine maintenance on the databases that power the Treasure AI Platform. During this maintenance, we will apply patches and updates to databases used by the Treasure AI Platform. These updates are necessary to ensure we have the latest security and performance improvements available. There may be a brief period of unavailability while we restart services, typically a few minutes. See below for information about the impact of this change to Treasure AI services. We will communicate with you via our status page (https://status.treasure.ai/ ) on the maintenance activities. If you notice any problems outside this maintenance window, we encourage you to contact our support team for assistance (https://docs.treasure.ai/support/contacting-treasure-data-support). # Impact During the maintenance window and the recovery period that follows, the impact follows one rule: - Asynchronous processing keeps working, just slower. Ingestion, jobs, and message deliveries are queued and automatically retried, and complete on their own after maintenance. - Synchronous requests return errors (mostly 5xx). Retry them after we mark maintenance complete on our status page. Customers make experience the following: - Jobs execution delay All jobs (Presto Query Engine, Hadoop / Hive Query Engine, Data Connector Integrations, Bulk Import, Bulk Export, Result Export, and Partial Delete jobs submitted from Console, API, Workflow or triggered by our system according to the configured schedule) will be delayed and automatically retried until they succeed after the downtime period. We expect job processing to ramp up over 30 minutes after the maintenance is complete before returning to full throughput. If customers experience errors running Data Connector Integration Export jobs, please contact our Support team at support@treasure.ai to recover the jobs. To avoid data duplication in your systems from Treasure AI exports, Integration Export jobs won’t be automatically retried. - Console execution delay Data Workbench and Audience Studio will show that jobs are running longer due to job execution delays. The Table preview (sample data and records counter) in Data Workbench may show old information while maintenance is underway. After maintenance is complete, we expect the processing of Data Workbench and Audience Studio jobs to recover gradually over 30 minutes. SQL or Workflow requests to create tables will be delayed and automatically retried after the maintenance period. - REST API The REST API endpoints to create tables, job submission, and job cancellation will be unavailable and will respond with a 500 error code. Components that use the REST API may also see performance degradation such as delays, authorisation failures, partial functionality, unavailability, and errors during the maintenance window. These components include “Web Interface”, “Mobile/Javascript REST API”, “Workflow”, “Workflow REST API”, “CDP API”, “Data Access API (beta)”, “ADL”, “Engage Studio REST APIs”, “Engage Studio Message Delivery”, “AI Agent Foundry, Foundry Workspace”, “Streaming Ingress / Egress connectors”. Requests, workflows, and deliveries will be retried where safe to do so, and scheduled jobs will be processed after the maintenance period. - Data Connector Integrations connection settings During the maintenance, creating or updating Data Connector Integrations connections may fail. Please retry after the maintenance is complete. - AI Signals During the maintenance window, requests to the ML Batch API used by AI Signals may temporarily fail, and machine learning training or prediction runs may not be created or executed. Treasure Workflow tasks that invoke the ML Batch API will retry automatically, so no customer action is expected for workflow-based runs. Customers that call the ML Batch API directly may receive failed requests during the maintenance window and should follow their normal retry handling after maintenance is complete. - AI Studio During the maintenance window, Treasure AI Studio users may experience errors when accessing chats, managing projects, or updating settings. What needs your attention - Campaigns scheduled during the window - check them and re-run any that failed. - Long-running workflows - workflows running longer than a few minutes may be terminated. Check and re-run them. - Integration Export jobs are not retried automatically, to avoid duplicating data in your systems. If one fails, contact support@treasure.ai to recover it. # Communication Beyond this notice, we will provide updates on our status page (https://status.treasure.ai/) approximately 1 hour before the beginning of the maintenance window, at the start and completion of the operation, and once the verification is complete. At that time, all systems will have returned to normal operations, and the Scheduled Maintenance will be closed. If you have any questions or concerns about this upgrade, please contact our Support team at support@treasure.ai.