qubestack-pennylane

QuREKA backend plugin for PennyLane.

Run PennyLane circuits on IonQ hardware, the QPerfect Mimiq emulator, IQM superconducting QPUs, or Rigetti superconducting QPUs through the QuREKA platform. The plugin ships four PennyLane devices that delegate execution to the same QuREKA gateway used by qubestack-cudaq; the gateway then forwards jobs to the actual provider.

PennyLane circuit
       │
       ├──►  qureka.ionq    ──►  POST /api/quantum-jobs  ──►  IonQ Cloud v0.4
       │                                                     (qpu.forte-1, ...)
       │
       ├──►  qureka.mimiq   ──►  POST /api/quantum-jobs  ──►  QPerfect Mimiq
       │                          QuREKA gateway              (sdt.qubesim*)
       │                          (qubestack-job-engine)
       │
       ├──►  qureka.iqm     ──►  POST /api/quantum-jobs  ──►  AWS Braket
       │                                                     (IQM Garnet / Emerald)
       │
       └──►  qureka.rigetti ──►  POST /api/quantum-jobs  ──►  AWS Braket
                                                             (Rigetti Ankaa-3, ...)

Supported Backends

IonQ (qureka.ionq)

Mimiq (qureka.mimiq)

IQM (qureka.iqm)

Rigetti (qureka.rigetti)

These match the device maps in qubestack-cudaq and the routing in qubestack-job-engine. Server-side routing happens by deviceCode prefix: ionq.* → IonQ Cloud v0.4 directly, sdt.qubesim* → Mimiq, and iqm.* / rigetti.* → AWS Braket.

Installation

pip install qubestack-pennylane

Requires Python 3.12 (matches qubestack-cudaq's runtime baseline). The plugin pins pennylane>=0.44,<0.45; tested against the current PyPI stable PennyLane 0.44.1.

Quick Start

import pennylane as qml

# IonQ hardware
dev = qml.device(
    "qureka.ionq",
    wires=2,
    backend="ionq.forte-1",     # see Supported Backends
    api_key="<QUREKA_API_KEY>",  # or set QUREKA_API_KEY env var
    shots=1000,
)

# Mimiq emulator (drop-in alternative)
dev = qml.device(
    "qureka.mimiq",
    wires=2,
    backend="sdt.qubesim-mimiq",
    api_key="<QUREKA_API_KEY>",
    shots=1000,
)

# IQM superconducting QPU (via AWS Braket)
dev = qml.device(
    "qureka.iqm",
    wires=2,
    backend="iqm.garnet",   # or "iqm.emerald"
    api_key="<QUREKA_API_KEY>",
    shots=1000,
)

# Rigetti superconducting QPU (via AWS Braket)
dev = qml.device(
    "qureka.rigetti",
    wires=2,
    backend="rigetti.ankaa-3",
    api_key="<QUREKA_API_KEY>",
    shots=1000,
)

@qml.qnode(dev)
def bell():
    qml.Hadamard(wires=0)
    qml.CNOT(wires=[0, 1])
    return qml.counts(wires=[0, 1])

print(bell())
# {'00': ~500, '11': ~500}

Configuration

Supported Operations

The common qis gateset that the QuREKA gateway can forward to all four providers (IonQ / Mimiq / IQM / Rigetti):

Hadamard, PauliX/Y/Z, S, T, SX, Identity, CNOT, CZ, SWAP, RX, RY, RZ, PhaseShift, IsingXX, IsingYY, IsingZZ, plus the adjoints of S, T, SX, RX, RY, RZ, IsingXX/YY/ZZ.

Higher-level operations (e.g. QFT, Toeplitz) must be decomposed first via qml.transforms.decompose or qml.transforms.unitary_to_rot before submission.

Development

git clone <repo>
cd qubestack-pennylane

# Python 3.12 is required.
python3.12 -m venv .venv
source .venv/bin/activate

pip install --upgrade pip setuptools wheel
pip install -e . pytest responses
pytest -v

Architecture

• docs/ionq-integration.md — end-to-end IonQ flow and side-by-side comparison with qubestack-cudaq.
• docs/mimiq-integration.md — Mimiq emulator integration: routing, JWT-based gateway interaction, big-endian protobuf result wrapper.
• docs/iqm-integration.md — IQM QPU integration via the gateway's AWS Braket route (Garnet / Emerald), whitelist validation, big-endian Braket measurement bitstrings.
• docs/rigetti-integration.md — Rigetti QPU integration via the gateway's AWS Braket route (Ankaa-3), whitelist validation, big-endian Braket measurement bitstrings.
• docs/testing-guide.md — how to run the full test suite and what each one verifies.
• docs/implementation-summary.md — high-level summary of what was built.

License

Apache License 2.0