Introduction
Throttling, quota limits and timeouts are the AI problems users meet most often. This article explains where those constraints actually come from, and gives you a short list of things to check on your own AI account and in ARVO Studio Setup before raising anything with Bluefort.
Which constraints the provider’s responsible for
Throughput, throttling, quotas, latency and regional availability are controlled by Microsoft and the selected model provider, not by Bluefort. That distinction matters when an AI operation is slow, is refused or does not return, because the limit being applied belongs to the customer’s own deployment or account rather than to ARVO Studio.
Note: AI runs on the customer’s own Azure AI Foundry deployment or Anthropic account and is paid for by the customer. Bluefort does not resell, supply or absorb AI capacity, so quota and rate limits are managed in the customer’s own subscription or account and are resolved there.
What to check
Work through the following, in this order.
- Check the quota available on your own deployment or account. Quota is held and adjusted in the customer’s own Azure subscription or Anthropic account, so that is where a shortfall is resolved.
- Check the region the deployment runs in, because regional availability is set by the provider and a model is not available everywhere.
- Check the maximum output tokens value in Setup. Keep the default unless the customer’s own deployment enforces a lower limit, because a value that is too low can truncate output and cause validation to fail.
- Confirm that only one AI connection is enabled, either Azure AI Foundry or Anthropic direct, and not both at once.
Reported failures
AI work runs in the background, so start at Background AI Jobs. That page shows active jobs with their status, start time and the error text if something fails, which tells you whether the operation was refused, timed out or stopped for another reason. Completed jobs leave the active queue, and their results remain in Runs.