Skip to content

Qontinuum 0.4.0 — "Execution Intelligence"

Qontinuum grows from a quantum DevOps toolchain into the Quantum Engineering Platform. This release teaches it to reason about executions — not just run them — and lands two architectural foundations underneath: a provider/SDK plugin system and a versioned execution history database.

Everything here is backwards compatible: no public signature, spec string, or result-JSON key changed. pip install -U qontinuum.

Execution intelligence

Running a circuit is an engineering decision. The same job can cost 100× more, run 10× slower, or return noise instead of signal on the wrong device. Three new commands turn the evidence Qontinuum already has — the cost catalog, published calibration, and your local run history — into ranked, explained decisions:

  • qont recommend — rank devices for your workload under six strategies (cost, fidelity, speed, value, reliability, balanced), with --budget filtering. Every candidate carries the reasoning, assumptions, missing evidence, and a confidence score behind it.
  • qont plan — a concrete pre-execution plan: selected device, estimated runtime/cost/fidelity, risks, and ordered fallbacks.
  • qont health — per-device reliability blending calibration quality with your local hardware track record, always attributed and confidence-weighted.

It is rule-based, provider-agnostic, and fully offline — no AI model, no hardcoded vendors. See Execution intelligence.

Provider, backend & SDK plugins

Providers, simulator backends, and SDKs are now plugins. ibm, braket, aer, qiskit, cirq, and pennylane ship built in; anyone can add a new provider or SDK by installing a package that declares an entry point — no fork, no core changes. qont plugin list / show / doctor inspect and diagnose them. See Extending Qontinuum.

Versioned execution history & richer dashboards

The local history is now a versioned execution database: old lines are migrated up on read (so any prior file stays valid), and hardware runs record a structured execution block — provider, cost, runtime, routing strategy, outcome — designed as the ingestion format a future hosted platform consumes without a redesign. qont dashboard gains hardware cost-over-time, provider usage, and provider-comparison views, still as a single self-contained HTML file. See Observability.

Upgrade notes

  • No breaking changes. qont history stats gains fields (old JSON keys kept).
  • append_history gained an optional execution= argument; existing callers are unaffected.
  • qontinuum.__version__ now correctly reports the release version.

What's next

Deeper deployment intelligence (queue-time modelling, actual-vs-estimated cost variance) and, later, an opt-in privacy-respecting cloud that turns anonymous execution metadata into community hardware analytics. Telemetry stays off by default.