@vpavlyshyn if you mean semantic triples it’s what RDF is all about and in Logseq we use Properties for that:
Page
key:: value
- Block
key:: value
where key is the label of the relation and value can be a page or block reference.
Once you export the graph as RDF you can feed a LLM like Llama with it and query your graph like you would with ChatGPT: