Microquantum

A lightweight, NumPy-only quantum computing SDK — MIT licensed and dependency-light.

microquantum is an independent quantum computing SDK. It is not a port or wrapper around Qiskit, Cirq, or OpenQASM — every circuit, operator, simulator and algorithm is implemented from scratch with NumPy as the only hard dependency.

Positioning: an SDK — a complete kit for building quantum applications. The core engine and algorithms form a library (you call it); the DomainAdapter pipeline and backend/provider abstractions form an embedded framework (it calls your code); a standardized result contract ties it all together for downstream solvers.

What's Included

Installation

Requires Python 3.10, 3.11, 3.12, or 3.13 and NumPy ≥ 1.20 (installed automatically).

MicroQuantum is a pure-Python/NumPy quantum SDK designed for cross-platform use on Windows, Linux, macOS, and BSD systems where the required Python and NumPy environments are available. The wheel is py3-none-any — no platform-specific binaries. Windows, Linux and macOS are verified in CI; BSD is supported at the portable Python/package level. See the platform support page for the exact verification status.

pip install microquantum

Or with uv:

uv pip install microquantum

Or from source:

git clone https://github.com/ajit-ai/microquantum.git
cd microquantum
uv sync --group dev

Documentation for the General Availability release is published at https://ajit-ai.github.io/microquantum/ (auto-deployed from the main branch); the Sphinx sources live in docs/.

Optional GPU acceleration (NumPy stays the default; the CuPy backend is opt-in):

pip install "microquantum[gpu]"

Verify:

python -c "import microquantum; print(microquantum.__version__)"

Quick Start

1. Bell state in two minutes

from microquantum import Executor, QuantumCircuit, StatevectorBackend

qc = QuantumCircuit(2)
qc.h(0)        # Hadamard on qubit 0
qc.cx(0, 1)    # CNOT (control=0, target=1)

result = Executor(backend=StatevectorBackend()).run(qc, shots=1024)
print(result.counts)          # {'00': ~512, '11': ~512}
print(result.most_frequent()) # '00' or '11'

Executor is the single high-level entry point: pass any Backend (statevector, noisy density matrix, MPS, tensor network, or a hardware provider) — or a NoiseModel for built-in noisy simulation. Every Backend also exposes a direct high-level backend.run(circuit, shots=1024, seed=None) call.

2. The standardized result contract

from microquantum.analytics.result import Result

result = Result(
    problem="optimization",
    solution={"route": "A->C->B", "cost": 42.0},
    confidence=0.91,
    qubit_count=8,
    runtime_ms=48.3,
    baseline={"route": "A->B->C", "cost": 51.7},
)
print(result.to_json())
print(result.improved_over_baseline)  # True

3. Running on real hardware

from microquantum import Executor, HardwareBackend, IBMQuantumCredentials, IBMQuantumProvider

provider = IBMQuantumProvider(IBMQuantumCredentials(api_token="..."))
backend = HardwareBackend(provider)

result = Executor(backend=backend).run(qc, shots=1024)

Architecture Principles

1. Independent implementation — all quantum operations are implemented from scratch using NumPy. No dependence on Qiskit, Cirq, OpenQASM, Strawberry Fields, or any other quantum SDK.

2. Layer separation — each layer only depends on layers below it: Analytics never touches raw matrices; domain adapters never bypass the core engine.

   Engine  →  Algorithms  →  Backends  →  Providers  →  NumPy
   

3. Big-endian qubit ordering — qubit 0 is the most significant bit; tensor axis 0 = qubit 0. This matches the mathematical convention.

4. Standardized results — every successful run funnels into a typed result (BackendResult, ExecutorResult, *Result, Result) that can be serialized with to_dict() / to_json().

5. Test-driven — a full test suite ships with the SDK and runs in CI.

Project Structure

microquantum/
├── src/microquantum/
│   ├── core/          # Quantum engine, circuits, gates, states, transpiler
│   ├── ir/            # Intermediate representation & compiler
│   ├── backends/      # Simulators, noise, tensor networks, Executor
│   ├── runtime/       # ExecutionPlan, ExecutionRuntime, strategies
│   ├── problems/      # Sampling / optimization / Hamiltonian / search
│   ├── algorithms/    # 23+ quantum algorithms
│   ├── experiments/   # Execution records, sweeps, experiments (MQ-07)
│   ├── analysis/      # Sampling / expectation / state analysis (MQ-07)
│   ├── optimization/  # QUBO / Ising toolchain
│   ├── providers/     # IBM Quantum & IonQ hardware clients (REST)
│   ├── adapters/      # Domain adapters (QuantumProblem/QuantumResult)
│   ├── analytics/     # CSV loading, result contract, analytics base
│   ├── qml/           # Quantum machine learning
│   ├── qec/           # Error correction codes
│   ├── benchmarks/    # Quantum benchmarking suite
│   ├── chemistry/     # Molecular Hamiltonians and ansätze
│   ├── mitigation/    # Error mitigation (ZNE, PEC, MEM)
│   ├── optimizers/    # Classical optimizers
│   └── stdlib/        # System standard library: bits, numbers, states
├── examples/          # Runnable demo scripts
├── docs/              # Sphinx documentation
└── pyproject.toml     # Package configuration

Development

# Install dev tooling (pytest, coverage, mypy, ruff, sphinx)
uv sync --group dev

# Run the test suite (with coverage report)
uv run pytest tests/

# Type check (strict, 0 errors expected)
uv run mypy src/microquantum/

# Lint / format
uv run ruff check src tests examples

# Build the docs
uv run sphinx-build docs docs/_build/html

See CONTRIBUTING.md for the contribution workflow and CHANGELOG.md for release history.

Roadmap

• v1.0.0 (current) — General Availability: the final planned MicroQuantum roadmap phase. Complete stdlib (microquantum.stdlib), locked package ecosystem, configurable runtime & tooling (RuntimeConfig, stage-tagged errors, runtime_info, the microquantum CLI), production/ stable packaging and the GA release notes — see docs/releases/ga.rst.
• v0.4.x — Developer Preview series: warning-free documentation with an auto-generated API reference, GitHub Pages deployment, a consolidated CI/packaging pipeline, and PyPI + TestPyPI releases via Trusted Publishing (python-publish.yml, testpypi-publish.yml).
• v0.3.0 — public open-source release: unified Backend.run(), serializable results (to_dict()), SPDX/legacy metadata cleanup, coverage + ruff gates, SDK-only docs.

The roadmap is complete; there are no Phase 121+ roadmap phases. Future enhancements are post-GA release work.

License

MIT — see LICENSE. Built as an independent, dependency-light quantum computing SDK: use it, fork it, build on it.