Does Graph packages for CMS 12 extract text from media files?

Johan Kronberg

Vote:

expand_less 0 expand_more

If so, what is required?

I'm testing with a content type that has pdf extension set. Seeing that _fulltext does get some meta attributes etc but nothing from the text content of the file.

{
  GenericMedia(
    limit:100
    where: {
      Name: { contains: "test.pdf" }
   }
 ) {
   items
   {
     _fulltext
   }
 }
}

These are the versions installed:

<PackageReference Include="Optimizely.ContentGraph.Cms" Version="4.3.0" />
<PackageReference Include="Optimizely.Commerce.GraphSearchProvider" Version="1.3.1" />
<PackageReference Include="Optimizely.Graph.Commerce" Version="1.3.1" />

#343410

Edited, Oct 05, 2026 9:27

Scott Reed

Vote:

expand_less 0 expand_more

I think the extracted content should be in the Content field. See https://docs.optimizely.com/graph/docs/text-extraction-from-media

Not the _fulltext but it should be included in CMS12, I believe for pdfs, docx, xls, xlsz, txt

#343412

Oct 05, 2026 9:55

Johan Kronberg

Vote:

expand_less 0 expand_more

Thanks! I must be missing something because I get no results using:

{
  GenericMedia(limit: 100, where: { Content: { exist: true } }) {
    items {
      _fulltext
      Content
      ContentType
    }
  }
}

#343415

Oct 05, 2026 10:09

Johan Kronberg

Vote:

expand_less 0 expand_more

I've now seen text extraction work in my local 13 PaaS solution... It goes into _fulltext same way as I confirmed later with SaaS.

There's ton of files in this setup that I was hoping text could be extracted from, but where it wasn't, that fooled me. A simpler PDF (exported from Word) works though.

I will try in 12 with this file shortly as well...

#343439

Edited, Oct 06, 2026 11:30

Mardy

workspace_premiumworkspace_premiumworkspace_premium

workspace_premium

Vote:

expand_less 0 expand_more

I tested this with both CMS 12 and CMS 13.

In CMS 12, extraction worked for the PDF/DOCX files I tested. The extracted text was available in Content, _fulltext and _json, and Content: { match: "..." } successfully found text from the documents.

One thing I noticed is that Content: { exist: true/false } does not seem to behave reliably for checking whether extracted text exists. I could get exist: false results even when Content was populated. I also found that a normal string field with an empty value ("") still matches exist: true.

In CMS 13, the schema is different: the Content field is no longer present, but the extracted text is available in _fulltext, and searching it works correctly.

So extraction itself seems to work in both versions, but the schema/representation differs, and on CMS 12 I wouldn't rely on Content.exist to determine whether extraction succeeded.

#343449

Edited, Oct 06, 2026 13:39