Order of content fields in Content Graph _fulltext property
When a text field is indexed by Content Graph in the _fulltext property, the values in the fields are appearing out of order from how they are configured in the CMS, where nested content is appearing before outer container content.
For example, our page contains an accordion block. The accordion block has a content area which allows topics blocks. The topics blocks contain a rich text field.
When querying graph the _fulltext property shows the values of the indexed content in this order:
Topics block content
Topics block name
Accordion block name
Container block name
We would expect to see the content in reverse order where the outer block names appear above the inner content:
Container block name
Accordion Block name
Topics block name
Topics block content
The customer is querying graph content to feed data into their own LLM, and the proper order of block or page contents is essential for the LLM since it affects how the model comprehends the context. For example, these blocks are being used as FAQ content, where the "question" is the name of the block and the "answer" is the content of the rich text field. Maintaining the order of fields here is critical to preserve the context of the content.
Cheryl Kellas
shared this idea
We’re currently gathering additional feedback on this idea. If you have any more context, examples, or specific use cases, please share them - it really helps us understand how this would best support your workflow. Thanks for contributing to the conversation!
-
Saad-Al- Muttakee
commented
Hello,
I am part of the team that was behind the feature request. I want to add more information regarding the ticket.If there are multiple content areas in a page, their text content appears in reverse inside the _fulltext property. For example, we have an article page with two content areas, "TopContent" and "BottomContent". The texts from the bottom content area appear right before the texts from the "TopContent".
I also want to share some of my thoughts regarding the reported issue. It seems to me that the "name" property of the block appearing at the end of the "_fulltext" is the root cause of the problem. This is true for all the blocks I have tried to work with. When we are referencing one block inside another, this issue compounds into a complex structure.