Dyber SDK

Write gate-model quantum circuits in Python, OpenQASM 3, Qiskit or Cirq, and run them on an exact local statevector simulator or on any server that speaks the Dyber job protocol. The resource report and the optional noise come from an illustrative H-cat resource model (see "What the simulator is (and is not)"). No H-cat hardware exists, and the hcat_qpu backend is not available.

30-second quickstart

pip install dyber

from dyber import Dyber, Circuit
c = Circuit(2, "bell")
c.h(0); c.cx(0, 1); c.measure_all()
job = Dyber().backend("local_simulator").run(c, shots=1000)
print(job.result().counts())        # {'00': ~500, '11': ~500}

Five lines, a Bell state. Qubit 0 is the leftmost bit in result strings.

Three access modes

Same circuits, same job protocol, pick the surface that fits your workflow.

1. Python SDK

from dyber import Dyber, Circuit

dy = Dyber()
print(dy.backends())                 # local_simulator (online), hcat_qpu (not available)

c = Circuit(2, "bell")
c.h(0); c.cx(0, 1); c.measure_all()

res = dy.backend("local_simulator").run(c, shots=1000, loss_per_cycle=0.005).result()
print(res.counts())                  # measurement counts
print(res.resources)                 # model values (illustrative), not a hardware operating point
print(res.noise)                     # illustrative noise summary

2. Command line

The dyber command ships with the package (or use python -m dyber.cli):

dyber run bell.qasm --shots 1000     # run an OpenQASM file on the simulator
dyber run bell.qasm --json           # machine-readable output for scripts
dyber backends                       # list backends and their status
dyber examples                       # bundled example circuits to get started
dyber serve                          # start the reference control plane (REST API)
dyber version

3. REST API

Start the reference control plane and submit jobs over HTTP. The protocol is documented in PROTOCOL.md.

dyber serve      # listens on http://127.0.0.1:8787

curl -s -X POST http://127.0.0.1:8787/jobs \
  -H "Content-Type: application/json" \
  -d '{"qasm": "OPENQASM 3.0; qubit[2] q; h q[0]; cx q[0], q[1];", "shots": 1000}'

Then GET /jobs/{id}/result returns counts plus the model resource report and noise summary. Jobs can also be submitted as circuit JSON ({"program": {...}}, see PROTOCOL.md). The hcat_qpu backend is not available.

Bring your own language

Dyber deliberately has no proprietary circuit language. Circuits come in and out through open formats and the tools you already use:

• OpenQASM 3 import and export: dyber.qasm.loads(text) and dyber.qasm.dumps(circuit). OpenQASM 2 programs are accepted on import.
• Qiskit adapter: dyber.interop.from_qiskit(qc) and to_qiskit(circuit) convert to and from a Qiskit QuantumCircuit.
• Cirq adapter: dyber.interop.from_cirq(circuit) and to_cirq(circuit) convert to and from a Cirq Circuit.

If your circuit already exists somewhere, it runs here. The adapters are duck-typed: neither Qiskit nor Cirq is a dependency of this package.

Examples

What the simulator is (and is not)

The local_simulator backend is an exact statevector simulator for up to 14 qubits. Noise is off unless you pass loss_per_cycle; with it, the simulator applies illustrative noise taken from the SDK's resource report. The resource report (code distance, cats per logical qubit, hardware modes) is always returned; without loss_per_cycle it uses 0.005.

Model value: abstract circuit model with assumed loss heralds. Not a hardware operating point. The model's thresholds (0.79 percent per cycle by default, 0.898 percent uncorrected) and distances describe the model only. At or above the model threshold the report returns the above-threshold regime with a NISQ estimate. No H-cat hardware exists or has been measured, and the hcat_qpu backend is not available.

The native summary in each result is an illustrative count of H-cat operations. In the H-cat encoding (J. Lee et al., PRX Quantum 5, 030322 (2024)), both logical states carry even cats. The logical Z value is read from the single photon, the logical X value from the photon's H/V outcome times the cat photon count mod 4, and cat photon-number parity only flags photon loss.

How it is layered

Development

pip install -e .[dev]        # from sdk/
python -m pytest tests -q

The native lowering here is illustrative: a count of H-cat operations, not a compiler. License: Apache-2.0.

Changelog

0.2.1

• Labeling and documentation release, with the output changes listed below. Resource-report thresholds, distances and cat counts are labeled as model values (abstract circuit model with assumed loss heralds, not a hardware operating point); the simulator's optional noise is described as illustrative; the H-cat readout is described as in the published encoding; the hcat_qpu backend is described as not available.
• The resource report and the noise summary carry a basis field with that label. The above-threshold note now ends its label with "not a hardware operating point".
• Breaking for code that reads native['parity_measurements']: use native['readouts']. The native summary key parity_measurements is renamed readouts (the op measure_parity is now hybrid_readout).
• The GET / descriptor's backend field is now local_simulator.
• hcat_qpu no longer probes the network: dyber backends and Dyber().backends() always report it as not available (no H-cat hardware exists), and running a job on it raises ConnectionError. An unreachable remote URL now gets a message that does not mention hcat_qpu.
• No change to circuits, OpenQASM, interop, simulator sampling, the numbers the model returns, or the job request format (protocol 1.0).

0.2.0

• OpenQASM 3 import and export, Qiskit and Cirq adapters, the dyber CLI, a REST API that accepts OpenQASM, examples and a test suite.