#10665 [SOA] Skip content extraction for image attachments
The Sales Order Agent now skips content extraction for image attachments and uses their stored text content instead, improving processing reliability while maintaining relevance filtering for all attachments.
- 3files
- 2AL files
- 2objects
Pull request · merged into main (BC30) on 2026-08-27 by qasimikram · 3 files (+35 -0), 2 AL · fix · tier official · reviewed (checked by Opus)
What changed
- Image attachments bypass the extraction model (which rejects them) and use pre-stored text content
- New IsImageAttachmentContentType function identifies standard and non-standard image MIME types
- Non-image attachments continue through the existing extraction and system-ID marker check path
- Relevance evaluation still applies to all attachments regardless of extraction bypass
- Change is marked temporary pending future extraction model support for images
- Behaviour changes for users or extensions.
AL objects touched
- Codeunit 4421 “SOA Attachment MLLM”, SalesOrderAgent: modified
- Codeunit 4400 “SOA Setup”, SalesOrderAgent: modified
Also merged into
- #10738 into
releases/29.xon 2026-08-27
Context
- Work items: AB#647838
- Labels: AL: Apps (W1), Team: SCM
- Systems: development
Source: metadata of a merged pull request in microsoft/BCApps (bcapps): title, labels, changed paths and counts, with a summary in our words. No source text and no description beyond one short quote (D10, D61).