A pure-NumPy quantum computing framework. Built for learning, prototyping, and reference.
LQE is built to run on a phone — no Qiskit, no Cirq, no heavy toolchains required. Every module is pure NumPy with optional Numba JIT acceleration. No Qiskit, no Cirq, no heavy toolchains. Just import numpy and go.
Prefer a no-install browser experience? Open LQE Lite for the phone-friendly JupyterLite lessons. This repository remains the full local engine with the complete module set and advanced features.
from legitimate_quantum_engine import StatevectorSim
sim = StatevectorSim(2)
sim.apply("H", 0)
sim.apply("CNOT", 0, 1)
print(sim.state)
# [0.707+0.j 0.000+0.j 0.000+0.j 0.707+0.j]from legitimate_quantum_engine import auto_solve, tfi_hamiltonian
H = tfi_hamiltonian(4)
result = auto_solve(H, n_qubits=4)
print(f"Ground state energy: {result['energy']:.4f}")
print(f"Solver used: {result['solver_used']}")python shell.py
> new 2
> h 0
> cnot 0 1
> state
> probs
> quitgit clone https://github.com/ethancjohnson0806-source/Legitimate-Quantum-Engine.git
cd legitimate-quantum-engine
python -m legitimate_quantum_engine.tests.test_allgit clone https://github.com/ethancjohnson0806-source/Legitimate-Quantum-Engine.git
cd legitimate-quantum-engine
pip install -e .pkg install python numpy
python -c "from legitimate_quantum_engine import StatevectorSim; s=StatevectorSim(2); s.apply('H',0); print(s.state)"| Subpackage | Modules | What It Does |
|---|---|---|
core/ |
statevector, sparse_hamiltonian, stabilizer, noise_models | Simulation backends: exact, sparse, Clifford, noisy |
solvers/ |
vqe_ground, vqe_excited, qaoa, adapt_vqe, qite, qng, orchestrator | Ground-state, excited-state, and auto-selecting solvers |
tensor_networks/ |
mps_simulator, dmrg, tdvp, stochastic_mps, adaptive_trotter | MPS, DMRG, and time-evolution methods |
quantum_chemistry/ |
fermionic_encoding, uccsd_ansatz, chemistry_driver | Molecular Hamiltonians and UCCSD-VQE |
quantum_error_correction/ |
surface_code, mwpm_decoder, bp_decoder | Surface codes and decoders |
circuit_tools/ |
qasm_export, dynamic_circuits, circuit_knitting, entanglement_budget, duality, zx_optimizer, solovay_kitaev | Circuit manipulation and optimization |
characterization/ |
classical_shadows, shadow_tomography, shadow_amputation, quantum_volume, rb_suite, differential_sim | Tomography, benchmarking, cross-validation |
open_systems/ |
lindblad | Lindbladian dynamics for noisy evolution |
monte_carlo/ |
qmc | Variational Monte Carlo solver |
benchmarks/ |
vqe_benchmarks, synthesis_benchmarks, resource_estimator | Performance and resource estimation |
utils/ |
symmetry_solver, quantum_kernel, extended_gates, error_mitigation, gpu_accelerator, jit_backend | Utilities and optional acceleration |
The engine includes 53 core modules across 11 subpackages, plus experimental chemistry and utility modules.
python -m legitimate_quantum_engine.tests.test_allExpected output: 28/28 tests passing in about 6–8 seconds on the audit environment. These tests cover core simulation, solvers, tensor networks, error correction, chemistry smoke checks, and regression checks; they do not prove every module correct for every input.
LQE is honest about what it can and cannot do. No false promises.
| Module | Limitation |
|---|---|
statevector.py |
n <= 16 qubits (2^16 amplitudes) |
mps_simulator.py |
Fixed maximum bond dimension |
dmrg.py |
Two-site algorithm only |
tdvp.py |
Single-site, first-order, no sweep |
stabilizer.py |
Clifford gates only (Gottesman-Knill) |
surface_code.py |
Distance d <= 7 exact simulation |
vqe_ground.py |
SPSA optimizer, ~15% tolerance on hard landscapes |
chemistry_driver.py |
STO-3G basis only; pre-computed integrals for H2, LiH, BeH2 |
uccsd_ansatz.py |
H2 UCCSD energy path is tested; the general circuit-versus-fermionic-sector equivalence remains under investigation |
qmc.py |
Product-state ansatz only |
lindblad.py |
n <= 6 qubits |
zx_optimizer.py |
Basic spider fusion only |
solovay_kitaev.py |
Single-qubit gates only |
- Pure NumPy. Every core module uses only
numpy. Optional dependencies (Numba) are wrapped in try/except and gracefully fall back. - Phone-runnable. If it cannot run on a $40 Android phone in Termux, it does not belong in core.
- Test-first. The unified suite provides 28 regression and integration checks. Passing the suite is evidence for those tested cases, not a proof of universal correctness.
- Honest limitations. We document what does not work, how far it scales, and when to use a heavier framework.
Pull requests welcome. All changes should pass python -m legitimate_quantum_engine.tests.test_all with 28/28 tests and should document any known limitations.
MIT License -- see LICENSE for details.
Built with persistence, pure NumPy, and the belief that quantum computing should be accessible to everyone.