Cognee 0.5.1 - Issue mapping custom OWL ontology to Graph nodes during Cognify

[LStromann]

=Hi Cognee Community,

I can't get Cognee to map data to my custom OWL. Logs show it loads successfully, but nodes are classified as generic IndexSchema instead of my classes (e.g., ProjectTask). Metadata shows "ontology_valid": false.

Setup: FalkorDB, Qdrant, RDF/XML (.owl).

1. Ontology Snippet: Defined User, Message, ProjectTask with ObjectProperties like hasSent. (Code attached).

2. Data (data.json):

JSON

[{"timestamp": "2026-01-20T11:00:00Z", "user_id": "user_id_103", "role": "Backend", "content": "Done with auth endpoint.", "task_mention": "Auth API"}]
3. Python Snippet:

Python

await cognee.cognify(config = {
"ontology_config": {"ontology_resolver": RDFLibOntologyResolver(ontology_file = path), "infer_relationships": True},
"graph_config": {"force_ontology_mapping": True}
})
Logs: Lookup built: 5 classes, 0 individuals [OntologyAdapter].

Questions:

Does cognee.add() require a specific format to trigger OWL class detection?

Why is ontology_valid: false if classes are in the lookup?

Are there mandatory rdfs:label conventions for LLM recognition?

Thanks!

This discussion was automatically pulled from Discord.

[LStromann]

Hi <@605959891273908243> thanks for sharing all the details here!!

To answer your questions:

1. there's no special format required for cognee.add() issue isn't with how the data is ingested it happens later during the entity-matching step in cognify()

2. what happens is: the llm first extracts entities from your text and then cognee tries to match those entities against your ontology. If it can't find a close enough match the node still gets created but it's marked as ontology_valid = false

3. labels aren't strictly mandatory but having clear, simple class and entity names definitely helps.

let us know if this helps!!

[LStromann]

How could I create an ontology for that case? How would you reformat it? I've tried everything. I don't know if I'm setting up the ontology wrong or what. I've even lost hope.

[LStromann]

DM'd you