Does Document Intelligence support in Canada East?

Question

Does Document Intelligence support in Canada East?

高高高 0

Hi,

I’m running into an issue with an AI Foundry project I created in Canada East.

The service consistently returns a 500 Internal Server Error (with no response body), whereas the same setup works perfectly in Canada Central.

User's image

I checked https://azure.microsoft.com/en-us/pricing/details/document-intelligence/ ,

and it seems like Document Intelligence is not supported in Canada East.

Could you clarify if Document Intelligence is actually available in Canada East?

0 comments

1 answer

Your answer

Answer 1

Anshika Varshney 13,320 Microsoft External Staff Moderator

Hi 高高高,

Yes, Azure AI Document Intelligence is supported in the Canada East region, but with some important nuances.

The Document Intelligence resource itself can be created in Canada East, and core capabilities like prebuilt models and custom document models are available when the resource is deployed there. However, not every feature or model version is guaranteed to be available in all regions at the same time.

If you’re seeing errors or missing features, it’s usually due to one of these reasons:

The specific model or API version you’re calling isn’t enabled in Canada East yet.
Your SDK or REST call is targeting an API version that’s not supported in that region.
The resource was created earlier and hasn’t been updated to use the latest supported API version for that region.

A couple of things you can check:

Confirm your endpoint is explicitly pointing to Canada East (for example, https://<resource-name>.cognitiveservices.azure.com created in Canada East).
Verify you’re using a region‑supported API version from the official Document Intelligence regional availability documentation.
If you need a model or capability that isn’t available yet in Canada East, the usual workaround is to deploy a separate Document Intelligence resource in another supported region and route only those requests there.

If you’re still blocked after validating the above, Do let me know.

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thankyou!

高高高 0 Reputation points

2026-02-18T16:03:04.1533333+00:00

Hi Anshika,

Thanks for the details.

I have a quick follow-up question regarding regional availability. The Azure pricing site (https://azure.microsoft.com/en-us/pricing/details/document-intelligence/ ) indicates that pricing isn't available for Canada East. Is the Document Intelligence service currently unavailable in that region?

Also, I noticed an endpoint appears when I set up a project in Microsoft Foundry. Is that endpoint ready for immediate use, or are there limitations I should be aware of?
Anshika Varshney 13,320 Reputation points Microsoft External Staff Moderator

2026-02-24T11:04:33.05+00:00
Hi 高高高,

This error usually indicates a transient service‑side failure during batch execution, not an issue with your input data. The batch starts processing, some items complete successfully, and then an internal backend error interrupts the remaining requests. That’s why you see partial outputs written to the output location.

These internal errors typically occur due to temporary backend instability, capacity issues, or validation pipeline failures. Since rerunning the same batch often succeeds, it strongly suggests the issue isn’t caused by the batch file itself.

There’s no guaranteed way to fully prevent this, but a few mitigations help reduce impact:

Split large batches into smaller ones so fewer items are affected if a failure occurs.

Track completed vs. missing items so retries can be targeted instead of rerunning everything.

Add retry logic for unprocessed inputs.

You’re generally charged only for the inputs that were actually processed. Items that never ran due to the internal error shouldn’t be billed, but checking usage metrics or confirming with support using the trace/activity ID is the safest way to verify.

Agreed, it’s frustrating to sort through partial completions. Designing the workflow to handle partial outputs (for example, using unique IDs per input) makes reruns much easier when these transient failures happen.

If you continue seeing this frequently, raising a support ticket with the trace ID can help the backend team confirm whether there’s an ongoing service issue.

Thankyou!

Share via

Does Document Intelligence support in Canada East?

1 answer

Your answer