zet

Zettel Model

The Zettel Model describes a zettel as contained within its own directory (a much more specific version of a KEG Knowledge Node). We’ll use jq query notation to give an overview. Requires understanding ZettelMark as well.

|Identifier |Description | |:————–:|:————–: | |.title |Title heading text (<=50 cols) | |.body |All text of zettel without title | |.video |Video link (if found) | |.tags[] |Tags from the tag line | |.data |Everything in data.yaml | |.figure |Figure related info (empty if no figure) | |.figure.height|Height of figure (if has one) | |.figure.width |Width of figure (if has one) | |.refs[].text |Text of reference (may match url or zet)| |.refs[].url |Fully qualified external Web URL | |.refs[].zet |Internal zettel ID (if has one) | |.links[]` |Other zettels that link to this one |

The refs[].url is special in that if the reference is nothing but text it will be returned as a link to the configured external search engine provider (DuckDuckGo by default).

Creation of .audio was decided against since .video can be used for audio only. Besides, .audio is best fulfilled by simply having a natural language processor read back the text content of the zettel.

A .summary (first paragraph) was considered and dismissed since the size of most zettels is too small to warrant it, and we want to encourage smaller zettels that are the most focused they can be.

A .figure name was dismissed because at most a single figure per zettel and it is always named figure-<width>-<height>.{jpg,png,gif}.