An AI tool in Foundry for analyzing documents and media to classify content, extract entities, and generate structured understanding
Hello @Tim Atkins
For API version 2025-11-01, a built analyzer retains the labeled training data and uses it during analysis. Therefore, after the analyzer build has completed successfully, normal analysis requests shouldn’t need to read the original Blob container again. The Analyze API receives the analyzer ID and the new document input; it doesn’t require the training-container URL.
Based on that documented lifecycle, you can use a short-lived SAS for the build, provided it remains valid until the asynchronous build operation reaches Succeeded. Expiration afterward shouldn’t prevent that already-built analyzer from using its retained labeled data.
A new valid SAS would be required whenever the service must access the source data again, for example, when updating the labeled samples and rebuilding the analyzer. Microsoft’s customization guidance requires rebuilding after labeled samples are added or changed.
Microsoft doesn’t document a mode in which a successfully built 2025-11-01 analyzer silently stops using its retained examples because the original SAS expired.
References:
- Data, privacy, and security for Azure Content Understanding
- Improve document analyzers with labeled samples
- Content Analyzers - Analyze REST API
- Content Analyzers - Get REST API
Help make this community better for everyone: If this answer helped or resolved your issue, please accept it or upvote it. If not, share more details in a comment so we can continue the discussion and find the right solution. Thank you.