Creating Workflows

Create a workflow from a built-in template when the same ordered calibration procedure must be saved, shared, run, or scheduled for a project.

Creating and changing workflows requires project edit permission. Start with the Fake backend when you are learning the editor or checking a new task sequence without laboratory hardware.

Start from a Template

1. Confirm the active project, then open Workflow. Saved workflows are listed in one folder per creator, with the subfolders set by each workflow's Folder property inside. Your own folder comes first and starts open; select another creator's folder header to open it.
2. Select New Flow.
3. Under TEMPLATE, confirm the backend; the configured default backend is preselected, and the list holds only the templates written for it, so a Fake backend never offers hardware templates. Then select the workflow closest to the procedure you need. Simple Calibration is the smallest general-purpose starting point on the qubex backend.
4. Under PROPERTIES, enter a unique Flow Name using letters, numbers, and underscores.
5. Confirm Entrypoint Function, Username, and Chip ID. The entrypoint must match the function marked as the flow in the source.
6. Edit the source, then select Save Flow.

Save Flow only creates a workflow. When the project already has a workflow with that name, including one created while you were editing, nothing is saved: the page keeps your source and properties and asks for another Flow Name. To change an existing workflow, open it and select Save, which replaces it.

Selecting another template replaces the current editor contents. Choose the template before making source changes.

Property	Meaning
Flow Name	Project-unique name of the saved workflow; it cannot be renamed later
Description	Purpose and expected use of the workflow
Entrypoint Function	Function QDash starts when the workflow runs
Username	Required default shown in the editor; the signed-in user owns each execution
Chip ID	Chip reserved while the workflow runs
Tags	Comma-separated labels added to execution history
Folder	Optional a/b path that files the workflow under subfolders in your space on the Workflow page; see Organize your workflows

Saving checks the Python syntax and entrypoint, then registers the workflow for execution. Correct any reported source, entrypoint, or registration error and save again.

Choose the Targets

Set either MUX targets or individual qubit targets in the template. Standard templates give MUX targets priority when both are present.

Source value	Use it for
mux_ids	All qubits belonging to one or more MUXes
exclude_qids	Qubits to remove from a MUX selection
qids	An explicit list of individual qubits

For an immediate run, select targets in Execute Flow as described in Run a Workflow. Changes in that dialog apply only to that run and do not change the saved source or schedules.

For scheduled runs, give the template a safe default target. In Simple Calibration, add this fallback immediately before the block that creates the targets:

if mux_ids is None and qids is None:
    mux_ids = [0]

Replace 0 with the MUX that the saved workflow should use. To target individual qubits instead, set qids = ["0", "1"]. If every caller supplies targets through an API, leave the template's missing-target check unchanged instead of adding a fallback.

Run Box Setup Before Spectroscopy

Use Box Setup on QuEL-1 or QuEL-3 hardware to align each selected MUX's readout capture before running resonator spectroscopy. The template requires mux_ids and runs CorrectCaptureDelay once per MUX.

1. Select Box Setup under TEMPLATE and set a flow name.

2. For an immediate run, select MUX IDs in Execute Flow. For scheduled runs, set a default MUX target in the source by replacing the template's missing-target check with:

   if mux_ids is None:
       mux_ids = [0]

   Replace 0 with the MUX you want to align.

3. Save the flow, then select Execute. Review the capture delay result before running a spectroscopy workflow.

Choose Tasks and Execution Order

The task list runs from top to bottom. Open Tasks in another tab to find the task names enabled for the active backend, then replace the default list in the template:

if tasks is None:
    tasks = [
        "Configure",
        "CheckT1",
        "CheckRamsey",
    ]

Keep configuration tasks and measurement tasks in their required order. A later task reads the calibration state produced or applied by earlier tasks.

The mode value controls how one-qubit work is arranged:

Mode	Behavior
scheduled	Groups work around hardware-sharing constraints; the default for Simple Calibration
synchronized	Runs the same task across the selected qubits before moving to the next task
serial	Runs the selected work sequentially
simultaneous_spectroscopy	Batches supported spectroscopy tasks across compatible targets

Keep the template's mode unless the procedure requires a specific ordering. The selected tasks must support the chosen mode and backend.

Schedule Two-Qubit Calibration

Use 2Q Calibration after the candidate qubits have the one-qubit calibration values required by the selected tasks. CR (cross-resonance) scheduling chooses directed coupling pairs and groups pairs that can run together without sharing qubits or conflicting hardware.

1. Open New Flow and choose 2Q Calibration under TEMPLATE.
2. Set a default qids list in the source for runs started from the application, as described in Choose the Targets. Include at least two connected qubits.
3. Set the direction and parallelism defaults in the source using the settings below.
4. Keep the template's task list or select the coupling tasks required by your procedure.
5. Save and test the workflow on a small target before expanding it.

Source setting	Meaning	Default
inverse	Reverse the selected CR direction; check that your procedure supports this direction	False
max_parallel_ops	Maximum coupling pairs in one parallel group; hardware conflicts can reduce this number	10

Groups run one after another. Within a group, coupling pairs can run in parallel, and each pair runs its task list from top to bottom. A pair can advance to its next task while another pair is still measuring. Pairs within one MUX are scheduled before pairs across MUXes. If no eligible pair remains after direction and hardware checks, scheduling fails.

In a combined one-qubit and two-qubit workflow, filter the one-qubit results before generating the CR schedule so the two-qubit work uses only the remaining candidates.

How this works: CR Gate Scheduler.

Set Parameters for One Task

Add task_parameters to the workflow's calibration service when one named task needs different run settings, calibration inputs, or output-saving behavior:

from qdash.workflow import sweep

task_parameters={
    "CheckRamsey": {
        "run": {
            "shots": 1024,
            "time_range": sweep.arange(0, 10_000, 100),
        },
        "input": {
            "control_frequency": 7.012,
        },
        "update_calibration_parameters": False,
    },
},

Values under run control the experiment. Use sweep.linspace, sweep.logspace, sweep.arange, or sweep.range for generated ranges so their construction remains visible in the workflow, or sweep.values(0, 4, 16, 64) for points you list yourself. A parameter that the task defines as one kind of range accepts any of these forms. Other run parameters use plain values. Values under input replace the current calibration input for every selected target in that task. Set update_calibration_parameters to False for an exploratory measurement whose outputs must remain in history without becoming the current calibration values.

QDash rejects a parameter placed in the wrong section, a parameter the task does not declare, or an input the task does not allow users to override. See Task Parameters before overriding calibration inputs or allowing outputs to update shared state.

Use Earlier Calibration Results in Later Steps

Later tasks use earlier outputs once those outputs pass validation and become current calibration values. This also applies across workflow steps: finishing a step does not reset the accepted values. Experiment settings such as shots and sweep ranges come from the workflow's settings and task defaults, rather than from previous measurement outputs.

For example, if a frequency measurement changes the accepted frequency from 5.000 GHz to 5.010 GHz, a later task that uses that calibration value resolves 5.010 GHz. If the measurement is configured to keep outputs only in history, the later task continues to use 5.000 GHz. An explicit input override takes priority over either value.

Review the effective inputs in Execution to confirm what the later task actually used. How this works: State Between Steps.

Save and Test the Workflow

1. Select Save Flow for a new workflow. For an existing workflow, select Edit, make the change, then select Save.
2. Confirm that Flow saved successfully! appears without a save error.
3. Select Execute and review Flow Name, Username, Chip ID, and Tags.
4. Confirm that no other run should use the chip. A workflow reserves the entire chip until all its steps finish.
5. Select Execute, then open Execution to follow task order, status, parameters, and results.

Use a small target and non-destructive task settings for the first run. Confirm the resulting task order and effective parameters before adding a schedule or expanding the targets.

How this works: Workflow Authoring Quickstart.

Organize Your Workflows

The Workflow page lists each workflow under its creator and, inside that space, under the folders you choose. Filing happens in the editor's Explorer panel, which shows one creator's space as folder/name.py; the selector at its top switches to another creator, and Open in editor on a creator's folder on the Workflow page opens the editor on that space. Moving a workflow changes only where it is listed: its name, source file, schedules, and running executions are untouched.

• Move a workflow: drag its row onto a folder, or onto the empty area below the tree for the top of the space; or right-click it and pick a folder from Move … to. Dragging a folder moves everything in it.
• Make a folder: select the folder button at the top of the Explorer, or right-click a folder and choose New folder, then type its name. The folder stays in the Explorer until you move a workflow into it or reload.
• Rename a folder: right-click it and choose Rename folder, or press F2 on it. Every workflow under it follows.
• Set the folder while saving: the Folder property takes a path such as daily/t1; / nests, and every level it names is created.

Workflows stay in their creator's space. Moving needs project edit permission, like saving.