Using content

[gminnick-presidio]

Sorry if this is not the correct place to post this. I do not use Github often and was not sure if there was a forum or discussion area for repos. When using the github with our AI frontend (which uses Claude) I find many times information gets missed. My AI also tells me that just pointing to the llms.txt leads to a huge pull of information as it tries to work its way to the index.md for each publication and then to find what I need either within the index, or within each .md file within the publication. Is there a better way to index the information, such as a tags line on each .md file? Or better yet, an index for the most common terms, functions or processes? I am currently testing something like this. I downloaded the whole repo as a zip and then extracted all files. I had my AI create a PowerShell script to parse every .md file in the it-operations-management folder and pull the headers. And from that make a best guess at what in the header would be a key term or process. And from that build a tag-based/keyword index. Having the documentation in a format for AI ingestion is great but I am not sure if I am not using it correctly or if it still needs some tweaking.

[ekimbernow]

Not sure if I can give you much assistance.

Each topic has metadata that includes relevant keywords--most topics have them but not all do.

Some topics will be classified with a "product" metadata field that is the name of the associated product, but not all topics are classified to a specific product.

The titles are usually fairly indicative of what a given topic applies to and the breadcrumb metadata field gives you title hierarchy from highest to lowest, which can be useful.

Otherwise, each publication has an index.md file in it's top-level directory that lists all the topics, so that's the best way to navigate since the order of the directories may not match the publication order.

The directory structure mirrors the servicenow.com/docs URL structure, which in turn reflects the publication's top-level URL component (usually reflective of the publication name) and, if present, the product classification, and then the topic's filename.

For example, in the Proactive Service Experience Workflows publication, all the topics are classified with a product, so the paths have two levels of directory:

https://github.com/ServiceNow/ServiceNowDocs/blob/australia/markdown/proactive-service-exp-workflows/product-support-for-technology/affected-customer-list-columns.md

But in ITOM, some topics have product classifications and some do not, so you see both subdirectories and markdown files under the top-level it-operations-management directory.