The most basic interaction is for citing, since being able to refer to a document is core to document relationships. Visual-Meta was designed for is to allow a user to copy text from a PDF and paste it as a full citation since the clipboard payload can have the copied text plus the full BibTeX payload from the Visual-Meta.
Basic view controls such as being able to fold the document into a Table of Contents.
Alternative views, such as concept maps of the text and graph views.
Paratexts, including letting the reader software understand where the References, Glossary and Endnotes are and how they are formatted
More custom/niche interactions are server based, where documents can surface their contents, from Dublin Core metadata to what values tables and charts contain, for instant access externally to internal data.