Getting Started

1. Prepare your data

Qruba expects prepared data:

• one row per candidate;

• a stable candidate identifier, usually candidate_index;

• numeric or boolean prepared values;

• clear scales, for example 0 to 1000;

• validation labels kept separate from decision inputs;

• no empty critical fields.

2. Build a workflow

Typical workflow:

Dataset Input
-> Predicate / Oracle or Decision Model or QIntent
-> Output
-> Traces / Verification

3. Choose an execution route

Qruba can work with different execution routes depending on deployment and license:

• logical or semantic validation;

• QuEST/statevector route;

• Aer-style simulation routes;

• IBM hardware routes when configured;

• private Docker routes for controlled environments.

4. Review results

After execution, review selected candidates, rankings, backend, execution path, reliability status, exported output, traces and evidence reports.

For hardware workflows, avoid reading a single accuracy metric as full physical validation. Qruba separates semantic decisions from hardware evidence and reliability.