Sparql Endpoint#

The Sparql Endpoint task exposes your RDF knowledge graph as a real SPARQL endpoint, and adds a natural-language field on top of it: ask a question in plain words and the assistant writes the SPARQL query for you, which you can then read, correct and run.

Set up the task#

1. In the assistant editor, on the General tab, click Advanced and select the Sparql Endpoint tile.
2. Add your RDF file on the Add data tab, if the assistant has none yet.
3. Click Use to open the assistant's page, whose first tab is the endpoint.

Write SPARQL yourself#

The lower half of the page is a SPARQL 1.1 editor over your graph, backed by qEndpoint. It opens with a SELECT * … LIMIT 10 query, which is a quick way to see what is actually in the graph.

• Query tabs — keep several queries side by side; + adds a tab, the × on a tab closes it.
• Run (▶) — execute the query in the active tab. Results appear underneath with the time they took.
• Prefixes — insert the usual prefix declarations, so you can write skos:Concept instead of the full IRI.
• Timeout — how many seconds a query may run before it is abandoned. Raise it for expensive queries.
• Share — produce a link that reopens the endpoint with this query already loaded.

Reading the results#

Results can be read in four ways:

• Table — one row per result, with sortable columns; IRIs are clickable.
• Response — the raw response the endpoint returned, which is what you want when debugging a query.
• Chart — plot the results, for counts and aggregations.
• Clickable — entities rendered so you can walk the graph by clicking from one to the next.

Above the results you also get a filter box, a page-size selector, Simple view and Ellipse switches that shorten long IRIs, and a download button for the result set.

Ask in natural language#

The Natural language query field at the top is the AI part of this task:

1. Type a question about the graph in plain words, for example "Which concepts are defined in the vocabulary?".
2. The assistant translates it into SPARQL and puts the query into the editor.
3. Read the query, adjust it if it missed something, and run it. What you see is exactly what will be executed — nothing happens behind your back.

Configure the task#

Everything is configured from the assistant editor, not from the endpoint page:

• Instructions (General tab) — the prompt that turns a question into SPARQL. This is the place to describe your schema's peculiarities.
• LLM Settings (General tab) — the model, its temperature and its speed. A low temperature is what you want here: SPARQL is not a place for creativity.
• Retrieving (Add data tab → Parsing Options) — how much of the graph's vocabulary is retrieved as context, plus the synonym groups that map your users' words onto the labels in the graph.