The _summary Convention

When a job completes, the Marqov dashboard displays its result. You can control what appears as summary cards at the top of the job detail page by including a _summary key in the result dict.

How It Works

The dashboard component (WorkflowExecutionDashboard) checks for a _summary field in the job result. If present, it renders each key-value pair as a card:

Keys become card titles (displayed as small muted text)
Values become card content (displayed as large bold text)

Cards are laid out in a responsive grid: 2 columns on mobile, 4 columns on desktop.

Returning _summary

The _summary dict must be a top-level key in the result returned by your main() function. Values must be strings.


async def main(client, params):
    dispatch = my_workflow(**params)
    result = await dispatch.run(client)
 
    return {
        "result": result,
        "_summary": {
            "Energy": f"{result['energy']:.4f} Ha",
            "Error": f"{result['error']:.4f} Ha",
            "Iterations": str(result["iterations"]),
            "Backend": params.get("backend", "local"),
        },
    }

This renders four cards on the dashboard:


+------------------+------------------+------------------+------------------+
|     Energy       |      Error       |   Iterations     |    Backend       |
|  -1.8512 Ha      |   0.0023 Ha      |      15          |     sv1          |
+------------------+------------------+------------------+------------------+

Requirements

Use the main(client, params) entry point strategy. The auto-detected single-workflow strategy does not give you a place to attach _summary to the result.
All values in the _summary dict must be strings. The dashboard renders them as-is.
The _summary key is passed through alongside result and _workflow_metadata in the enriched result format.

Example: Without vs With _summary

Without _summary

The job result is stored but the dashboard only shows the raw JSON output and the workflow execution graph. Users have to dig into the result to find key metrics.


async def main(client, params):
    dispatch = bell_experiment(params)
    result = await dispatch.run(client)
    return {"result": result}

With _summary

Key metrics are highlighted as cards at the top of the job page, making results scannable at a glance.


async def main(client, params):
    dispatch = bell_experiment(params)
    result = await dispatch.run(client)
 
    counts = result.get("counts", {})
    total = sum(counts.values())
 
    return {
        "result": result,
        "_summary": {
            "Shots": str(total),
            "State |00>": f"{counts.get('00', 0)} ({counts.get('00', 0) / total * 100:.1f}%)",
            "State |11>": f"{counts.get('11', 0)} ({counts.get('11', 0) / total * 100:.1f}%)",
            "Fidelity": f"{(counts.get('00', 0) + counts.get('11', 0)) / total * 100:.1f}%",
        },
    }

VQE Example


async def main(client, params):
    best_energy = float("inf")
    best_theta = 0.0
 
    for i in range(params.get("max_iterations", 20)):
        theta = optimize(best_theta)
        dispatch = vqe_step(theta, params)
        result = await dispatch.run(client)
 
        if result["energy"] < best_energy:
            best_energy = result["energy"]
            best_theta = theta
 
    return {
        "result": {
            "energy": best_energy,
            "theta": best_theta,
        },
        "_summary": {
            "Ground State Energy": f"{best_energy:.6f} Ha",
            "Optimal Theta": f"{best_theta:.4f} rad",
            "Iterations": str(params.get("max_iterations", 20)),
            "Chemical Accuracy": "Yes" if abs(best_energy - (-1.8512)) < 0.0016 else "No",
        },
    }

Tips

Keep the number of summary cards to 4-6 for a clean layout.
Format numbers with appropriate precision (.4f for energies, .1f for percentages).
Include units in the value string (e.g., “Ha”, “rad”, “ms”).
Use the summary for the most important metrics; put detailed data in result.