Release wave: QUBODrivers 0.6 / QUBOTools 0.13 compat + benchmark conformance

[bernalde]

Part of the benchmarking-readiness epic: JuliaQUBO/QUBO.jl#40 (WS2 — coordinated driver release wave). Blocked by the QUBODrivers metadata/seed standardization and conformance-suite issues on that epic (WS1) — adopt both in a single PR here so the package is only touched once.

Problem

QUBO.jl v0.6 requires QUBODrivers = "0.6" / QUBOTools = "0.13". DWave.jl's current compat is:

QUBODrivers = "0.3.3, 0.4, 0.5"
QUBOTools   = "0.10, 0.11, 0.12"

The intersection with QUBO.jl v0.6 (and with the QUBOLib-served QOBLib benchmarking stack) is empty — this package cannot be installed alongside the rest of the ecosystem. A scheduled ecosystem-resolution canary is being added to QUBO.jl that will stay red until this lands.

Package-specific work

• Largest driver in the wave: Neal, Greedy, Tabu, Random samplers plus the cloud QPU path. Conformance CI should run the offline samplers; the cloud path is exercised only when DWAVE_API_TOKEN is present (current test-skip pattern is fine — keep it).
• Expose RandomSeed for Neal (the underlying simulated-annealing sampler accepts a seed); cloud QPU declares supports_seed = false.
• D-Wave timing is rich (qpu_access_time, qpu_sampling_time, etc.): map QPU access time to metadata["time"]["effective"] and preserve the full timing dict under a namespaced breakdown key — TTS analysis on QPU results depends on this being faithful, not approximated by Python-side wall time (current behavior).
• PythonCall = "0.9.26" pin: verify it does not conflict with PySA.jl (0.9.23) and CIMOptimizer (0.9) in one environment — if it does, relax to "0.9" with a CondaPkg-level pin instead.

Common work (all drivers in the wave)

1. Bump compat: QUBODrivers = "0.6" (drop pre-0.5 entries unless there's a reason to keep them) and QUBOTools = "0.13" where it is a direct dependency. Fix any breakage from the upgrade.
2. Adopt the QUBODrivers metadata schema: metadata["time"]["effective"] (backend solve time) always set; seeds recorded under metadata["seeds"]; backend name/version under metadata["backend"].
3. Declare capability traits honestly (supports_seed, honors_final_reads, enforces_time_limit).
4. Adopt the standard RandomSeed attribute where the backend supports seeding (alias any existing seed attribute to it; keep the old one as deprecated).
5. CI runs the upgraded QUBODrivers.test(Optimizer) with benchmark_conformance = true.
6. Tag a release and verify it registers (General registry compat must match).

Checkpoints (anti-drift)

• Before starting: wait for the reference driver run posted on the epic (first wave driver, suggested MQLib.jl) and use its conformance output as the template — deviations should be deliberate, not accidental.
• Mid-PR: post the conformance-suite output here (all sets, including which are trait-skipped and why).
• Before merge: resolve a temp environment with QUBO@0.6 + this branch via Pkg.develop — the actual coexistence check, not just compat-file edits.

Acceptance criteria

• Registered release with QUBODrivers 0.6 / QUBOTools 0.13 compat, conformance suite green in CI, package resolves alongside QUBO.jl v0.6.

[bernalde]

Reference driver checkpoint is ready enough to start this package's wave PR.

MQLib status:

• Add QUBODrivers 0.6 benchmark conformance MQLib.jl#27 merged into main.
• Reference checkpoint posted on the epic: Epic: Benchmark-ready JuliaQUBO ecosystem for the QOBLib campaign QUBO.jl#40.
• MQLib release-prep PR Prepare v0.7.0 release MQLib.jl#28 is open, CI-green, and bumps the release to v0.7.0; registration is still pending.
• Final reference stack uses QUBODrivers = "0.6.1" and QUBOTools = "0.13".
• Reference conformance result: supports_seed=true, honors_final_reads=true, enforces_time_limit=true, Benchmark Conformance 14/14, no trait-skipped benchmark sets.

Implementation guidance for DWave.jl:

• Target QUBODrivers = "0.6.1" and QUBOTools = "0.13".
• Run benchmark conformance in CI against offline samplers; keep the cloud QPU path gated on DWAVE_API_TOKEN.
• Expose RandomSeed for Neal where supported; declare cloud QPU seed support honestly, likely supports_seed = false.
• Map D-Wave QPU access/solver timing to metadata["time"]["effective"] and preserve the richer D-Wave timing fields under a namespaced breakdown.
• Verify PythonCall/CondaPkg resolution alongside PySA and CIMOptimizer before release.

Use MQLib.jl#27/#28 as the reference shape: compat bump, metadata schema, trait declarations, unconditional QUBODrivers.test(...; benchmark_conformance = true), conformance transcript, QUBO v0.6 coexistence check, then release/registration.

[bernalde]

@JuliaRegistrator register branch=main

[JuliaRegistrator]

Registration pull request created: JuliaRegistries/General/157970

Recommendation: add release notes

Release notes help your users know what has changed in the new version. For breaking releases in particular, it's recommended to document what breaking changes were made (or point to where changes can be found, such as a changelog).

To add release notes, just add markdown formatted text underneath the comment after the text "Release notes:" and it will be added to the registry PR, and if TagBot is installed it will also be added to the release that TagBot creates. i.e.

@JuliaRegistrator register

Release notes:

## Breaking changes

- blah

To add release notes to this registration, just re-invoke JuliaRegistrator and the PR will be updated.

Tagging

After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.

This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via:

git tag -a v0.7.0 -m "<description of version>" 9053f14336df27d4f2c4f4917b0676059b7827a6
git push origin v0.7.0

[bernalde]

@JuliaRegistrator register branch=main

Release notes:

Breaking changes

• Narrow package compatibility to the current JuliaQUBO stack: Julia 1.10+, QUBODrivers 0.6.1, and QUBOTools 0.13.

Changes

• Add QUBODrivers 0.6 benchmark-conformance support for the D-Wave QPU wrapper and offline samplers.
• Declare sampler capability traits for seed handling, final read counts, and time-limit behavior.
• Emit QUBODrivers-compatible benchmark metadata, including D-Wave backend, read count, seed, timing, and chip information.
• Update the D-Wave Ocean SDK dependency stack to the audited 9.3.0-era Python packages.

[JuliaRegistrator]

Registration pull request updated: JuliaRegistries/General/157970

Tagging

After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.

This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via:

git tag -a v0.7.0 -m "<description of version>" 9053f14336df27d4f2c4f4917b0676059b7827a6
git push origin v0.7.0

[bernalde]

DWave v0.7.0 is released.

• General registration: New version: DWave v0.7.0 JuliaRegistries/General#157970
• GitHub release: https://github.com/JuliaQUBO/DWave.jl/releases/tag/v0.7.0