QLever and "dynamic graphs"

[kontost]

I’ve been exploring QLever and its high-performance SPARQL capabilities, and I wanted to clarify its intended use case compared to traditional triplestores (like e.g.GraphDB).

From my understanding, QLever operates on a static, pre-indexed dataset, meaning that once data is loaded and indexed, it cannot be modified via SPARQL UPDATE (e.g., inserting or deleting triples on the fly). Instead, the only way to change data is to modify the source files and reindex everything.

Given this, I wanted to ask:
Is my understanding correct that QLever currently cannot support evolving or dynamically updated RDF graphs? Is there any other way of doing this? For example, lets say consuming data from Rabbitmq queues, then transformed to RDF, and inserted to a hosted Qlever triplestore.
If not, do you recommend any best practices for working with frequently changing datasets in a QLever-based setup?
I’d love to hear thoughts from the team or anyone experienced with the above in regards to the QLever environment.

Thanks!

[hannahbast]

@kontost That was true for an earlier version of QLever. SPARQL 1.1 Update has always been on our agenda and will be fully implemented by 30.06.2025. The basic operations INSERT DATA, DELETE DATA, INSERT { ... } WHERE { ... }, and DELETE { ... } WHERE { ... } are implemented already now. More information and some preliminary performance tests here: https://github.com/ad-freiburg/qlever/wiki/QLever-support-for-SPARQL-1.1-Update

[kontost]

Just to confirm, currently I can dynamically modify data in QLever only through the supported SPARQL 1.1 UPDATE operations (INSERT DATA, DELETE DATA, etc.), but there is no way to directly push a .ttl file via an API to update the QLever repository without reindexing.
Is that correct?

[aindlq]

You can always transform ttl into INSERT DATA, + I saw that qlever team is working on Graph Store Protocol implementation, see ad-freiburg/qlever#1727

[hannahbast]

@kontost Like @aindlq said, SPARQL 1.1 Graph Store HTTP Protocol will be implemented soon and then you can simply add a whole TTL file via a POST request.

How large is your base dataset and how large are the additions relative to that?