Local-first experiment tracking for quantum computing.

Capture circuits, backend state, and configuration automatically — runs are reproducible, comparable, and easy to share. Access your data via Python API, CLI, or Web UI.

General-purpose experiment trackers (MLflow, W&B, neptune.ai) work well for classical ML, but quantum workloads need structure they don't provide out of the box: what actually executed (program + compilation), where it ran (backend/device), and how it ran (runtime options, calibration state).

Same circuit, same backend, different day — different results. devqubit helps you track why.

• Automatic circuit capture — QPY, OpenQASM 3, SDK-native formats
• Multi-SDK support — Qiskit, Qiskit Runtime, Braket, Cirq, PennyLane, CUDA-Q
• Content-addressable storage — deduplicated artifacts with SHA-256 digests
• Reproducibility fingerprints — detect changes in program, device, or config
• Run comparison — TVD analysis, structural diff, calibration drift detection
• CI/CD verification — baselines with configurable noise-aware policies
• Portable bundles — export/import runs as self-contained ZIPs
• Web UI — browse runs, view artifacts, compare experiments

📚 https://devqubit.readthedocs.io

Python 3.11+ required.

pip install devqubit

Install with an SDK adapter:

pip install "devqubit[qiskit]"          # Qiskit + Aer
pip install "devqubit[qiskit-runtime]"  # IBM Quantum Runtime
pip install "devqubit[braket]"          # Amazon Braket
pip install "devqubit[cirq]"            # Google Cirq
pip install "devqubit[pennylane]"       # PennyLane
pip install "devqubit[cudaq]"           # NVIDIA CUDA-Q
pip install "devqubit[all]"             # all adapters

Optional extras:

pip install "devqubit[ui]"              # local web UI

from qiskit import QuantumCircuit
from qiskit_aer import AerSimulator
from devqubit import track

qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
qc.measure_all()

with track(project="bell-state", run_name="baseline-v1") as run:
    backend = run.wrap(AerSimulator())
    job = backend.run(qc, shots=1000)
    counts = job.result().get_counts()

    run.log_param("shots", 1000)
    run.log_metric("p00", counts.get("00", 0) / 1000)

print(f"Run saved: {run.run_id}")

The adapter automatically captures the circuit, backend config, job metadata, and results.

from devqubit.compare import diff

result = diff("baseline-v1", "experiment-v2", project="bell-state")

print(result.identical)                 # False
print(result.program.structural_match)  # True — same circuit structure
print(result.tvd)                       # 0.023

devqubit diff baseline-v1 experiment-v2 --project bell-state

from devqubit.compare import verify_baseline, VerifyPolicy

result = verify_baseline(
    "nightly-run",
    project="bell-state",
    policy=VerifyPolicy(tvd_max=0.05, noise_factor=1.2),
)

assert result.ok

# With JUnit output for CI pipelines
devqubit verify nightly-run --project bell-state --junit results.xml

devqubit list                            # list runs
devqubit show <run> --project <proj>     # run details
devqubit diff <a> <b> --project <proj>   # compare runs
devqubit baseline set <proj> <run>       # set baseline
devqubit verify <run> --project <proj>   # verify against baseline
devqubit pack <run> -o bundle.zip        # export bundle
devqubit ui                              # launch web UI

See the CLI reference for all commands.

pip install "devqubit[ui]"

devqubit ui
# → http://127.0.0.1:8080

Browse runs, view artifacts, compare experiments, and manage baselines.

devqubit is a monorepo with multiple packages:

Users install devqubit (plus optional extras). All public API lives in the devqubit namespace; engine internals are not part of the public API.

We welcome contributions — bug fixes, docs, new adapters, or feature ideas.

1. Read CONTRIBUTING.md for setup and guidelines
2. Check open issues or start a discussion
3. Fork, branch, and submit a PR

git clone https://github.com/devqubit-labs/devqubit.git
cd devqubit
uv sync --all-packages
uv run pre-commit install
uv run pytest

Early project — high impact contributions. Jump in!

• 💬 Discussions — questions, ideas, feedback
• 🐛 Issues — bug reports, feature requests
• 📚 Docs — guides and API reference

Apache 2.0 — see LICENSE.