little big brain
Product / Ask

Ask in plain words.
Get the answer and its sources.

One call answers a question from your graph. The answer cites its rows, and the server checks the citations.

[01] In the console

From a question to cited rows.

A router selects the route. Then a model runs SPARQL in a bounded loop and answers.

The Question answering page: 22 questions answered and routed, with calls, tokens and costThe Question answering page: 22 questions answered and routed, with calls, tokens and cost
One check opened: the answer from Claude Sonnet 5.5, the judge Claude Opus 5.5 keeps it, with the SPARQL query and its rowsOne check opened: the answer from Claude Sonnet 5.5, the judge Claude Opus 5.5 keeps it, with the SPARQL query and its rows
The query comes with it.

The response holds the SPARQL and the rows behind the answer.

Bounded work.

At most 4 rounds and 12 steps. A question takes about 6 to 12 seconds.

Six routes.

Lookup, aggregate, search, history and schema. A sixth route marks a question unanswerable.

The Question answering page: Jev routes and Claude Sonnet 5.5 answers; 22 questions, 76 calls, 411,148 tokens, $0.33The Question answering page: Jev routes and Claude Sonnet 5.5 answers; 22 questions, 76 calls, 411,148 tokens, $0.33
Question answering on the demo graph.
[02] Usage

Answers you can account for.
Calls, tokens and cost per day.

The Question answering page names the models behind the answers. It counts questions, calls, tokens and cost.

  • Jev routes. Claude Sonnet 5.5 answers.
  • Activity by day and by graph.
  • A sample of answers goes to the judge.
Answer questions
[03] What is included

What the response holds.

An answer for your users, and the parts behind it for your code.

POST /v1/query/ask

The answer, the query, its rows and the steps that led there.

mode: route

The route alone, in about 0.3 seconds.

askStream

Server-sent events while the answer forms.

names · describe · compare

Tools for your own agent loop, with no model calls inside.

lbb_query

Modes question, names, describe, commit_at and compare over MCP.

models

Tokens and cost per model, in the response.

[04] How it works

A route, a query loop, an answer.

Step 01

Route.

Jev, from TypeSafe, selects one of six routes.

Step 02

Query.

Claude Sonnet 5.5 writes SPARQL and reads the rows, for up to 4 rounds.

Step 03

Answer.

The answer cites rows. The server checks the citations against them.

[05] FAQ

Questions about Ask.

What does a question cost?

About 1 to 2 cents in model use. The response lists the tokens and the cost.

Is there a daily limit?

500 questions per environment per UTC day.

Can my agent use it?

Yes. Call lbb_query with mode question over MCP, or query.ask in the SDKs.

Which plans include Ask?

Cloud and Pro. Free includes search by meaning.

Next step

Try it on your data.