LabOne Q Experiment Maintainer

Scope

Use for development and debugging of:

• •experiments/*.py
• •analysis/*.py
• •qpu_types/fixed_transmon/*.py
• •validation notebooks under examples/selectiveRIP/*.ipynb

Hard Rules

• •Modify paired experiment/analysis modules together when behavior changes.
• •For new module requests, run a Comprehension Gate before coding (no implementation until user confirms the brief).
• •Enforce acquisition contract in create_experiment(...) with explicit errors.
• •Keep workflow branching that depends on values inside @workflow.task (avoid Reference logic in workflow body).
• •Treat workflow-body values as Reference; do not call dict methods on unresolved references.
• •Persist calibration values only through new_parameter_values and apply with update_qpu(...).
• •Save workflow artifacts only inside @workflow.task.
• •In notebooks, append validation cells; do not rewrite existing cells unless requested.
• •Keep workflow option values concrete and type-safe (bool, str, int, etc.); do not pass unresolved Reference into option setters.
• •Keep tune-up option defaults explicit (do_analysis=True, update=False) unless intentionally overridden.

Workflow

1. •If this is a new experiment/analysis request, complete Comprehension Gate and get user confirmation.
2. •Read both experiment and analysis files for the target flow.
3. •Implement behavior change in experiment sequence/options and analysis outputs consistently.
4. •Verify update path (old_parameter_values / new_parameter_values) for persisted parameters.
5. •Verify temporary_parameters handling and qpu.copy_qubits() usage when runtime overrides are needed.
6. •Add or update notebook validation cells (pass/fail style).
7. •Run quick syntax checks before finishing.

Validation

• •Workflow executes without OptionBuilder/Reference type errors.
• •Matrix/count outputs are finite and normalized on the correct axis by definition.
• •Reconstructed density matrix checks: Hermitian, trace≈1, min eigenvalue ≥ tolerance.
• •If update=True, expected parameter keys are present in new_parameter_values.
• •If workflow output is needed in notebooks, materialize task outputs and handle workflow_result.tasks as TaskView/sequence (not plain dict).

Do Not

• •Do not call workflow.save_artifact from notebook helper functions.
• •Do not silently coerce unsupported acquisition modes.
• •Do not change existing notebook cells when append-only validation is sufficient.
• •Do not use deprecated update_qubits in new code; use update_qpu.

References

For concrete error signatures and fixes, see references/common-errors.md. For doc-backed workflow contracts and update semantics, see references/applications-library-contracts.md. For User Manual guardrails (Reference/TaskView/acquisition/debug contracts), see references/laboneq-user-manual-guards.md. For request understanding and implementation standardization, see references/comprehension-gate.md.

Reference trigger policy:

• •Core layer first (laboneq.dsl, Experiment, acquire_loop_rt, AcquisitionType, compile/session/device mapping): read references/laboneq-user-manual-guards.md.
• •Applications layer first (laboneq_applications, experiment_workflow, .options(), do_analysis, update, update_qpu, temporary_parameters): read references/applications-library-contracts.md.
• •New module request (make new experiment/analysis, add workflow from scratch, major redesign): read references/comprehension-gate.md first, then continue with layer-specific references.
• •Boundary changes (experiment + analysis + update path edited together): read both references, then cross-check with references/common-errors.md.
• •Error-first fallback: Reference/OptionBuilder/TaskView/update_qpu issues -> applications contracts + common errors; compile/device/signal-map issues -> user-manual guards + common errors.