# Skip embedding generation for selected documents

**URL:** https://forum.opensearch.org/t/skip-embedding-generation-for-selected-documents/27375
**Category:** OpenSearch
**Tags:** discuss, index-management
**Created:** [October 27, 2025, 4:36pm UTC](https://forum.opensearch.org/t/skip-embedding-generation-for-selected-documents/27375 "2025-10-27T16:36:48Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![sousu](https://avatars.discourse-cdn.com/v4/letter/s/e95f7d/32.png) [@sousu](https://forum.opensearch.org/u/sousu)
#### Post date: [October 27, 2025, 4:36pm UTC](https://forum.opensearch.org/t/skip-embedding-generation-for-selected-documents/27375/1 "2025-10-27T16:36:48Z")

</div>

**Versions** : 3.1

Is there a way to skip embedding generation on a semantic field and process only the raw type? I mean like an option that could bypass the semantic step?

Currently I am updating a product that would require that semantic information is generated only for some documents. I could redefine an old field to a semantic field with text raw type an can use the raw information without making any changes to any other system component. But for this use case I only need that some of the documents generate semantic information, not all of them, and I understand that generating for all when not needed would have a considerable impact. It would be really good that this is possible as it would avoid the necessity to:

- create a new index with the same configurations aside from a field that is semantic;
- programmatically create an alias on demand (as both indexes would be templates) that make searching both indexes at the same time possible;
- make changes to other system components to adequate to those change.

I searched on the documentation to enable this behavior but could not find anything. If this is not yet possible please inform so I can write a feature request.