Stanza

Documentation: https://docs.conductorquantum.com/stanza

Stanza is a Python framework for building tune-up sequences for quantum devices. Configure devices with YAML, write routines as decorated functions, and execute them with automatic logging.

Quick Start

pip install cq-stanza

Or with Quantum Machines drivers:

pip install "cq-stanza[qm]"

Define Your Device and Routines

Configure your device topology and routine parameters in YAML. Parameters defined in the config are automatically passed to decorated routines:

# device.yaml
name: "My Quantum Device"

gates:
  G1: {type: BARRIER, control_channel: 1, v_lower_bound: -3.0, v_upper_bound: 3.0}
  G2: {type: PLUNGER, control_channel: 2, v_lower_bound: -3.0, v_upper_bound: 3.0}

contacts:
  SOURCE: {type: SOURCE, control_channel: 3, measure_channel: 1, v_lower_bound: -3.0, v_upper_bound: 3.0}
  DRAIN: {type: DRAIN, control_channel: 4, measure_channel: 2, v_lower_bound: -3.0, v_upper_bound: 3.0}

gpios:
  VSS: {type: INPUT, control_channel: 5, v_lower_bound: 0, v_upper_bound: 5}
  VDD: {type: INPUT, control_channel: 6, v_lower_bound: -5, v_upper_bound: 0}

routines:
  - name: sweep_barrier
    parameters:
      gate: G1
      v_start: -2.0
      v_stop: 0.0
      n_points: 100
      contact: DRAIN

instruments:
  - name: qdac2-control
    type: CONTROL
    driver: qdac2
    ip_addr: 127.0.0.1
    slew_rate: 1.0

Device Groups

Group definitions let you carve a device into logical regions for per-routine isolation. Add a groups section that lists the gates, contacts, and GPIOs that belong to each group. By default every routine receives the full, unfiltered device; if you set group on a routine entry, Stanza hands that routine a filtered view containing only the group’s pads.

• Reservoir-style gates are the only gates that can appear in multiple groups. Sharing them is useful when two regions need access to the same bus or measurement reservoir while everything else stays isolated.
• Contacts (and GPIOs) can be shared freely between groups. Shared contacts make it easy to measure a combined signal from multiple groups without redefining channel hardware.

If you need to bias two groups at once and measure through a shared contact, define the routine without a group parameter and filter the groups manually:

routines:
  # Receives the full device; routine chooses how to partition it
  - name: cross_group_calibration
    parameters:
      bias_voltage: 0.3
      control_group: control
      sensor_group: sensor
      measure_contact: OUT

  # Receives only the control group view
  - name: health_check
    group: control

@routine
def cross_group_calibration(ctx, bias_voltage, control_group, sensor_group, measure_contact):
    device = ctx.resources.device
    control_configs = device.filter_by_group(control_group)
    sensor_configs = device.filter_by_group(sensor_group)

    # Get control gates from each group
    control_gates = [
        name for name, config in control_configs.items()
        if config.control_channel is not None
    ]
    sensor_gates = [
        name for name, config in sensor_configs.items()
        if config.control_channel is not None
    ]

    # Bias each group independently using the main device
    device.jump({gate: bias_voltage for gate in control_gates})
    device.jump({gate: bias_voltage / 2 for gate in sensor_gates})

    # Measure through the shared contact
    reading = device.measure(measure_contact)
    return {"current": reading}

With this pattern you can mix and match group-specific routines (which set group) and cross-group routines (which omit it) in the same configuration.

Write a Routine

Routine parameters from YAML are passed as kwargs. You can override them at runtime:

import numpy as np
from stanza.routines import routine

@routine
def sweep_barrier(ctx, gate, v_start, v_stop, n_points, contact):
    """Sweep a barrier gate and measure current."""
    device = ctx.resources.device
    voltages = np.linspace(v_start, v_stop, n_points)
    v_data, i_data = device.sweep_1d(gate, voltages.tolist(), contact)
    return {"voltages": v_data, "currents": i_data}

Run It

First, initialize a session directory for your experiment data:

# Initialize a new timestamped session directory
stanza init

# Or with a custom name
stanza init --name my_experiment

# Check current session
stanza status

This creates a directory structure like:

your-project/
├── .stanza/
│   └── active_session.json                       # Tracks the active session
├── 20251020143022_untitled/                      # Timestamped session directory
│   ├── 20251020143022_untitled_notebook.ipynb    # Jupyter notebook for the session
│   └── .stanza/
│       └── config.json                           # Session metadata
└── device.yaml                                   # Your device config

The Jupyter notebook is pre-configured with Stanza imports and ready for running your routines. When you provide a custom name with --name, the notebook will be named {timestamp}_{name}.ipynb instead.

Then run your routines - all data will be logged to the active session directory:

from stanza.routines import RoutineRunner
from stanza.utils import load_device_config

# Load configuration
config = load_device_config("device.yaml")

# Create runner - automatically loads routine parameters from config
runner = RoutineRunner(configs=[config])

# Execute with config parameters
result = runner.run("sweep_barrier")

# Or override specific parameters at runtime
result = runner.run("sweep_barrier", gate="G2", n_points=50)
print(result["currents"])

Core Features

CLI Session Management: Initialize timestamped experiment directories with stanza init to organize your data.

Device Abstraction: Define quantum devices with gates, contacts, and instruments in YAML. Access them uniformly in code.

Decorator-Based Routines: Write tune-up sequences as simple Python functions with the @routine decorator.

Resource Management: Access devices, loggers, and other resources through a unified context object.

Result Tracking: Store and retrieve results from previous routines to build complex workflows.

Automatic Logging: Sessions and data are logged automatically with support for HDF5 and JSONL formats.

Type Safety: Built on Pydantic for configuration validation and type checking.

Live Plotting

Visualize your data in real-time as routines execute. Stanza supports two plotting backends:

Quick Start with CLI

Enable live plotting in your notebook or script:

# Plots in browser (default) - opens at http://localhost:5006
stanza live-plot enable

# Plots in notebook cells (requires: pip install jupyter_bokeh)
stanza live-plot enable --backend inline

Once enabled, the DataLogger automatically detects the configuration and streams data to plots as your routines run.

Python API

You can also enable live plotting programmatically:

Server Backend - Opens plots in a browser window (works in any environment):

from stanza.plotter import enable_live_plotting

# Enable live plotting with browser backend
backend = enable_live_plotting(runner.context.resources.logger, backend="server", port=5006)

# Run your routine - plots update live at http://localhost:5006
result = runner.run("sweep_barrier")

Inline Backend - Displays plots directly in notebook cells (requires jupyter_bokeh):

from stanza.plotter import enable_live_plotting

# Enable inline plotting
backend = enable_live_plotting(runner.context.resources.logger, backend="inline")

# Run your routine - plots appear and update in the notebook cell
result = runner.run("sweep_barrier")

Plots automatically update as data is logged, supporting both 1D line plots and 2D heatmaps for real-time visualization of sweeps and measurements.

Jupyter Integration

Stanza provides CLI commands to manage a Jupyter notebook server for interactive development and analysis.

Server Management

Start a Jupyter server in the background that survives terminal closure:

# Start server in current directory on default port 8888
stanza jupyter start

# Start in specific directory with custom port
stanza jupyter start /path/to/notebooks --port 8889

# Check server status (shows PID, URL, uptime)
stanza jupyter status

# Open Jupyter in browser with authentication token
stanza jupyter open

# Stop the server gracefully
stanza jupyter stop

Notebook Monitoring

Monitor and interact with running notebooks:

# List all active notebook sessions
stanza jupyter list

# View log files for active notebooks
stanza jupyter logs

# Tail a specific notebook's output (Ctrl+C to detach)
stanza jupyter logs my_notebook.ipynb

# Attach to a notebook with kernel control (Ctrl+C kills kernel, ESC exits)
stanza jupyter attach my_notebook.ipynb

The logs command is useful for monitoring long-running experiments, while attach provides active control for debugging and development. Both commands support showing initial context with the -n/--lines option (default: 10 lines).

Architecture

Stanza separates concerns into three layers:

1. Configuration (YAML) - Define device topology and routine parameters
2. Routines (Python functions) - Implement tune-up logic
3. Execution (Runner) - Orchestrate resources and logging

@routine
def my_routine(ctx, **params):
    # Access resources
    device = ctx.resources.device
    logger = ctx.resources.logger

    # Use previous results
    prev_result = ctx.results.get("previous_routine")

    # Your logic here
    return result

Device Operations

Stanza devices support common operations:

# Jump to voltages
device.jump({"G1": -1.5, "G2": -0.8})

# Set VSS and VDD for GPIO pins
device.jump({"VSS": 1.5, "VDD": -1.5})

# Zero specific pads or all pads
device.zero(["G1", "G2"])  # Zero specific pads
device.zero()              # Zero all pads

# Measure current
current = device.measure("DRAIN")

# Check current voltage
voltage = device.check("G1")

# Sweep operations
v_data, i_data = device.sweep_1d("G1", voltages, "DRAIN")
v_data, i_data = device.sweep_2d("G1", v1, "G2", v2, "DRAIN")
v_data, i_data = device.sweep_nd(["G1", "G2"], voltages, "DRAIN")

Breakout Box Operations

Stanza supports digital breakout boxes for routing control and measurement signals. Configure channels with breakout_channel and instruments with breakout_line:

# device.yaml with QSwitch breakout box
gates:
  G1: {type: BARRIER, control_channel: 1, breakout_channel: 1, v_lower_bound: -3.0, v_upper_bound: 3.0}
  G2: {type: PLUNGER, control_channel: 2, breakout_channel: 2, v_lower_bound: -3.0, v_upper_bound: 3.0}

contacts:
  DRAIN: {type: DRAIN, control_channel: 3, measure_channel: 1, breakout_channel: 3, v_lower_bound: -3.0, v_upper_bound: 3.0}

instruments:
  - name: qdac2
    type: GENERAL
    driver: qdac2
    ip_addr: 192.168.1.100
    port: 5025
    slew_rate: 100.0
    breakout_line: 9  # QSwitch relay that connects to this instrument

  - name: qswitch
    type: BREAKOUT_BOX
    driver: qswitch
    ip_addr: 192.168.1.102
    port: 5025

Control the breakout box in your routines:

@routine
def configure_breakout(ctx):
    device = ctx.resources.device

    # Ground all breakout lines for safety
    device.ground_breakout_lines()

    # Unground specific lines to enable control
    device.unground_breakout_lines()

    # Connect all lines to their configured instruments
    device.connect_breakout_lines()

    # Disconnect breakout lines when not in use
    device.disconnect_breakout_lines()

Built-in Routines

Stanza includes health check routines for device characterization:

from stanza.routines.builtins import (
    noise_floor_measurement,
    leakage_test,
    global_accumulation,
)

# Run health checks
runner.run("noise_floor_measurement", measure_electrode="DRAIN", num_points=10)
runner.run("leakage_test", leakage_threshold_resistance=50e6, num_points=10)
runner.run("global_accumulation", measure_electrode="DRAIN", step_size=0.01, bias_gate="SOURCE", bias_voltage=0.005)

These routines include automatic analysis and fitting for device diagnostics.

Examples

See the cookbooks directory for:

• Basic device configuration
• Writing custom routines
• Running built-in routines
• Jupyter notebook workflows

Development

git clone https://github.com/conductorquantum/stanza.git
cd stanza
pip install -e ".[dev]"
pytest

License

MIT License - see LICENSE for details.