Docs/Settings/Ontology

Ontology

The Ontology tab in Settings adds terms of your own to this thoughtbase's knowledge graph, beside the ones Minerva ships with. A vocabulary is a Turtle (.ttl) file: a lab's schema, say, or a published vocabulary you want to query with. Once it's imported, its prefixes work in every query and its classes and properties appear when you write SPARQL.

What you can do

Import a vocabulary

Use Import Turtle… and pick a .ttl file. Before anything is saved, Minerva shows what the file declares: its prefixes, its classes and properties, and how many statements it adds. Choose Import to add it, or Cancel. Importing a file with the same name as one you already have replaces it.

See what's loaded

Each vocabulary is listed with its prefixes and how many classes and properties it declares. A file that can't be used is listed as Not loaded, with the reasons beneath it.

Remove a vocabulary

Remove deletes the file from the thoughtbase and takes its terms out of the graph. Queries that use its prefix stop working until you import it again.

Using its terms

Say the file starts with @prefix lab: <https://lab.example/ns#> . From then on you can write lab:Sample in the Query panel, in a ```sparql block or in a view, with no PREFIX line. Typing lab: in the Query panel offers the vocabulary's classes and properties, whether or not any note uses them yet. And when you ask the AI to query your graph, it sees the vocabulary alongside Minerva's own.

The vocabulary is stored in the thoughtbase, in .minerva/ontology/, so it travels with the thoughtbase: your queries keep working when it's shared or opened on another machine.

When a file can't be imported

A vocabulary can add terms, but it can't change Minerva's. Minerva checks the whole file, and won't import it if it:

describes one of Minerva's own terms, such as thought:Claim. Building on one is fine: declaring your class a subclass of thought:Claim works;
gives a standard prefix like skos: or dc: a different meaning;
contradicts itself or another vocabulary you've imported. For example, it gives one property two ranges, or one term two labels in the same language;
isn't valid Turtle, or is larger than 1 MB.

The review lists each problem by name, so you can fix the file and import it again. The same checks run when a shared thoughtbase arrives with vocabularies in it, or a file is edited outside Minerva: a file that fails is listed here and isn't loaded, and the rest of the graph loads as usual.

A vocabulary describes terms; it doesn't give orders

When the AI reads a vocabulary, Minerva marks it as material from the thoughtbase, not as instructions. A comment inside a file someone shared with you can describe a term, but it can't tell the AI what to do.