Customer impact

Between August 6, 2026 at 10:00 am UTC and 1:00 pm UTC, some organizations in the US region were unable to complete document ingestion in UiPath Context Grounding. A small number of search requests in the same region were also slow or timed out.

Action required: please re-submit the affected ingestion requests. Ingestion retries a failing request automatically for a limited number of attempts. Once those attempts are exhausted the request is marked failed and is not retried again, so affected documents will not appear in your index until the request is submitted again. Failed requests are listed in the ingestion history for each index.

No action is required for search. Those requests were affected only while the issue was ongoing, and subsequent searches completed normally.

Root cause

A sudden increase in concurrent document ingestion triggered a high number of simultaneous document validation steps, which created a capacity bottleneck on the underlying infrastructure resource beyond its scaling capacity.

Once that resource was saturated, ingestion operations began exceeding their time limits and failing. Automatic retries of the failed operations added further load, which sustained the condition. Search requests served by the same resource were delayed behind the same contention.

Detection

Automated alerts were flagged as the condition developed, and an automated infrastructure resource capacity alert triggered at 10:37 am UTC brought it to the team's attention.

Response

• 10:03 am UTC — Automated low severity alerts started coming in.
• 10:37 am UTC — Automated alert for resource capacity issue paged the team.
• 12:23 pm UTC — As a mitigation step the impacted resource's capacity was increased.
• 12:57 pm UTC — Ingestion and search operations stopped failing and response times returned to normal.

Follow-up

• The fix is deployed. The validation step has been reimplemented to enforce the same limits at a small fraction of the previous cost, so this level of concurrent ingestion now sits well within available capacity. It was released to the affected US region on August 7, ahead of schedule, and reaches all remaining regions by early September.
• We are improving how quickly we detect issues like this. We are adding monitoring that tracks whether document ingestion is completing successfully for customers, so problems are identified and acted on directly rather than inferred from underlying system alerts. This will be in place across all regions by the end of August.