Licensed
Quantum Toolkit (Sqetch)
Our toolkit on top of the open-source Aria runtime: circuit builders and optimizer passes, QML with autograd, error correction, Lean 4 and Rocq export, a finance lab, and a reproducible binary distribution.
For teams that want the Aria runtime plus a tested library of circuits, models and proofs they can build on.
The pieces
How it fits together.
Open source
Aria runtime
The language and the simulation engines, Apache-2.0. Everything in the toolkit runs on it.
→Licensed
EvaluationQuantum Toolkit (Sqetch)
Nine Rust crates on top of Aria: circuit algebra, optimizer passes, QML, error correction, photonics, data and backtesting, a QPU abstraction and Python bindings.
quantum-examples
Self-contained binaries and a guided tour of sixteen demos that each check themselves against a numeric value. Evaluation builds are time-limited and qubit-capped.
→Circuits
What the toolkit adds to Aria.
- Circuit algebra30+ gate types, symbolic parameters, annotations, composition operators and a circuit! macro for building circuits in Rust.
- Optimizer passesIdentity, adjacent and adjoint-pair cancellation, rotation and T merging, commutation reordering, gate decomposition and a Clifford+T pipeline. Every pass is checked against the statevector on random circuits.
- ExportersOpenQASM 2.0, OPTICQASM, JSON, omega-IR, a C FFI, Lean 4 and Rocq.
- Error correctionRepetition, Steane and surface codes with an MWPM decoder, plus measurement-based and blind delegated computation.
- BuildersBell, GHZ, QFT, QPE, Trotter steps for H₂, Ising and Heisenberg, QSVT, block encodings, QML encodings, and Shor including an EC-DLP estimate.
- Paper reproductions25 reproductions, 16 gate-model and 9 photonic, each printing a headline metric with a pass or fail against the paper's own mathematics.
QML
Machine learning that runs on a QPU.
- QuantumLayerA PyTorch-compatible layer with parameter-shift gradients, plus adjoint differentiation, CMA-ES and surrogate training.
- Train classical, deploy quantumRandom-feature surrogates of quantum models, an IQP Born machine trained analytically, and a photonic generative model whose deployment is boson sampling.
- Quantum Oracle SketchingAn emulated oracle built from samples, following the O(1/N²) sample-complexity law.
- GPU-awareCPU, Apple Metal and CUDA simulation with fused kernels and batched parameter sweeps.
Shipping
How you get it.
- Binary distributionReproducible builds for macOS (Metal and CPU) and Linux (x86-64, arm64, RISC-V), with a sha256 manifest you can re-verify.
- PythonPyO3 bindings and a stdlib-only socket client.
- LicensingThe toolkit is licensed per team. The Aria runtime underneath stays Apache-2.0.