Storage format

Comments are plain text inside the note. They remain readable without the plugin, and any script can read them. The format is a promise: notes written with an older version keep working, and every change to it is backward compatible and recorded in the changelog.

Markers in the text

A commented passage is wrapped in two Obsidian comments that share an id. A comment placed at the cursor is an empty anchor:

The author argues that %%[a1%%memory is a reconstruction%%a1]%% rather than a recording.
A comment placed here %%[a2%%%%a2]%% without a selection.

The annotations block

The comment texts live in a single block at the end of the note:

%% annotations
[a1] 2026-10-31 #idea First comment
    continued (indented continuation lines)
  [a1.1] 2026-11-04 Reply to comment a1
[a2] 2026-10-31 Comment without a category
%%

A root comment has the id aN; its replies are aN.k, indented by two spaces. Each entry starts with [id] YYYY-MM-DD. An optional #tag (the category, root comments only) can follow, then the text. Indented lines that follow continue the entry. An uncategorised text that starts with # is written as \#….

%% is hidden in reading view, so the note stays clean without the plugin.

For scripts

The regular expression for a commented passage is %%\[(a\d+)%%([\s\S]*?)%%\1\]%%. For a cursor comment, group 2 is empty.

A comment whose markers were deleted stays in the block. The panel flags it as orphaned.

To write comments from a script, prefer the API to editing the text by hand: it keeps tags and events in step.

Automatic tags

When a comment gets a category, its tag is added to the note’s tags property. The tag is removed when no comment of that category remains, unless the note already had that tag before. The plugin keeps track of the tags it added in its data.json.