The Library

Foundation Models have powerful generalised knowledge but will perform less well at business-specific tasks, unless they have the right context. Libraries help you bring this context directly into Go, so you can improve the performance of the foundation models for your specific needs.

Libraries house universal files that can be used as inputs/references in any Go project - this makes it easier for companies to teach AIs to better solve specific tasks.

How does this work?

You can upload files to your global Library in the top left corner of Go, as below:

You can then provide one or more of these files as reference files as an input, so that GPT4 Omni in this case can refine its outputs to be more specific to your work. In this case, we've used a 'Company Glossary' that will show GPT4 the company-specific language that our example user would like to see in the output.

FAQs

• What happens if I delete, edit, or replace a library file that is an input to a property?
  • All dependent fields become stale (rather than automatically re-calculating)
  • If you need to recompute them, you can trigger this manually with the UI or API
• What are the cost/token implications of the library?
  • As it stands, like any other Go property, if a file in the library is used as an input, this consumes tokens
  • This means using multiple large reference files in each property could get expensive, fast
  • If this becomes an issue, flag this to your CSM and we'll look into this and try to find a solution
• What file types does the library support?
  • Upload = PDF, CSV, Image
  • Create = free text (see Create document option below)