pyrobot

pyrobot is a modern Python ontology toolkit built on py-horned-owl.

It is not a line-for-line port of ROBOT. The design goal is a composable Python API for ontology manipulation, dataframe workflows, and plugin-driven extensions.

Installation

From the GitHub repo:

pip install git+https://github.com/cmungall/pyrobot.git

For local development:

git clone https://github.com/cmungall/pyrobot.git
cd pyrobot
python3 -m venv .venv
. .venv/bin/activate
pip install -e ".[dev]"

Initial scope

• ontology loading, creation, cloning, and serialization
• core operations: merge, extract, filter, convert, annotate, diff, reason, template, query
• pandas adapters for components and annotations
• lazy component streams for batch-oriented downstream processing
• entry-point based plugin registry for future operation packs and reasoner backends
• click-based CLI via pyrobot

Reasoning boundary

The current py-horned-owl and horned-owl ecosystem is strong on parsing, modeling, and serialization. Native reasoning is not exposed as a first-class API in py-horned-owl today, so pyrobot keeps reasoning behind a plugin boundary rather than hard-wiring an incomplete abstraction into the core package.

The built-in reason operation therefore performs structural reasoning over asserted named-class hierarchies and materializes transitive SubClassOf axioms. Full DL reasoners are left as pluggable backends.

Quick start

from pyrobot import OntologyDocument
from pyrobot.ops import EntityAnnotationChange, ExtractSpec, diff, extract, merge

left = OntologyDocument.from_string(
    """
    Prefix(rdfs:=<http://www.w3.org/2000/01/rdf-schema#>)
    Prefix(ex:=<http://example.org/>)
    Ontology(<http://example.org/left>
        Declaration(Class(ex:A))
        Declaration(Class(ex:B))
        SubClassOf(ex:A ex:B)
    )
    """,
    serialization="ofn",
)

right = merge([left, left.clone()], output_iri="http://example.org/merged")
subset = extract(right, ExtractSpec(seeds={"ex:A"}, max_hops=0))
report = diff(left, subset)
print(report.identical)

CLI

The package installs a pyrobot console script:

pyrobot --help

Convert an ontology:

pyrobot convert -i edit.owl -o edit.ofn --to ofn

Extract a local neighborhood around one or more seed terms:

pyrobot extract -i edit.ofn -o subset.ofn --seed CL:0000540 --seed UBERON:0002107 --max-hops 1

Add annotations:

pyrobot annotate \
  -i edit.ofn \
  -o annotated.ofn \
  --entity-annotation 'CL:0000540|rdfs:comment|reviewed by pyrobot' \
  --ontology-annotation 'rdfs:comment|release candidate'

Run structural reasoning over the asserted hierarchy:

pyrobot reason -i edit.ofn -o reasoned.ofn --backend structural

Build ontology content from a ROBOT-style template TSV:

ID	Label	Parent	Comment
ID	LABEL	SC	A rdfs:comment
ex:Neuron	Neuron	ex:Cell	created from template

pyrobot template -t terms.tsv -o templated.ofn --ontology-iri http://example.org/templated

Run a SPARQL query:

pyrobot query \
  -i templated.ofn \
  -q 'PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#> SELECT ?s ?o WHERE { ?s rdfs:subClassOf ?o . }'

Run a basic graph-pattern query without writing full SPARQL:

pyrobot query \
  -i templated.ofn \
  --pattern '?child|rdfs:subClassOf|?parent' \
  --select ?child \
  --select ?parent

Diff two ontologies:

pyrobot diff left.ofn right.ofn --json-output

Template directives

The current built-in template engine supports a focused subset of ROBOT-style directives:

• ID: subject IRI or CURIE for the row
• LABEL: rdfs:label
• TYPE: class, object property, data property, annotation property, or named individual
• SC: subclass axiom
• EC: equivalent classes axiom
• A <property>: literal annotation assertion
• AI <property>: IRI-valued annotation assertion
• AL <property>: language-tagged annotation assertion, using text@lang
• AT <property> <datatype>: datatype-valued annotation assertion

Development

pytest
pyrobot --help