Hello Nikolay Yordanov,
Welcome to the Microsoft Q&A and thank you for posting your questions here.
I understand that you are getting error in adaptive custom translation (preview).
This is not a formatting error in your files; it is a technical limitation in the current public preview of the feature. As long as the workspace originates from the API, the platform will reject uploads and hide the workspace from the Azure AI Foundry portal.
Therefore, it is important to know that Adaptive Custom Translation in its preview phase only supports file uploads and portal visibility for workspaces or projects created directly inside the Azure AI Foundry UI. API‑created containers are not fully wired into the backend upload and portal experience yet, which explains both the upload failure and the missing workspace. This is expected behavior during preview and will be resolved by GA.
To fix the issue, the best practice is to first create your Adaptive Custom Translation project/workspace inside the Azure AI Foundry portal. You can follow the Foundry resource creation quickstart here - https://learn.microsoft.com/en-us/azure/ai-services/multi-service-resource. Once created through the UI, the workspace becomes fully supported for TMX/TSV file uploads, portal visibility, and dataset indexing.
After the UI‑based workspace is created, return to the Adaptive Custom Translation interface in the portal and upload your bilingual TMX/TSV dataset. Adaptive CT supports 5–10,000 aligned pairs with up to 512 characters per pair. This ensures your data passes validation and allows the system to build a dataset index suitable for few‑shot adaptation.
If you prefer to automate the process using the API, you can now safely call the document import endpoint using the workspaceId retrieved from the new UI‑created workspace. For example:
curl -X POST \
"https://<your-resource-name>.cognitiveservices.azure.com/translator/customtranslator/api/texttranslator/v1.0/documents/import?workspaceId=<workspaceId>" \
-H "Authorization: Bearer <token>" \
-F 'DocumentDetails=[{"DocumentName":"my-document","DocumentType":"Adaptive","FileDetails":[{"Name":"data.tmx","LanguageCode":"en","OverwriteIfExists":true}]}]' \
-F "FILES=@path/to/data.tmx"
Full API details - https://learn.microsoft.com/en-us/azure/ai-services/translator/custom-translator/azure-ai-foundry/concepts/adaptive-custom-translation
Next, create a dataset index referencing the successfully imported document IDs and use that index in your translation requests via the 2025‑05‑01‑preview Text Translation API. Adaptive CT dynamically selects the most relevant examples during inference, and using the index ensures that your terminology and style guide the translation output.
Finally, because this feature is still in preview, expect minor UI inconsistencies, naming drift (workspace vs. project), and occasional backend quirks. The reliable path is always: create in the UI > upload in the UI > automate with the API only after the workspace exists in the UI. Following this pattern guarantees consistent success until the feature reaches General Availability.
I hope this is helpful! Do not hesitate to let me know if you have any other questions or clarifications.
Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.
