build(deps): bump AISimulate to 0.12.0

[saturley-hall]

Summary

• bump the Python aisimulate dependency from 0.12.0.dev2 to 0.12.0
• bump the Rust aisimulate-core dependency from 0.12.0-dev.2 to 0.12.0
• refresh all three Cargo lockfiles with the published crate checksum
• keep Dynamo's dependency consistency assertion aligned with the final release

This change is also intended to be cherry-picked to release/1.5.0.

Validation

• pytest -q on the five focused AISimulate dependency/manifest consistency tests: 5 passed
• cargo check --locked -p dynamo-mocker --no-default-features
• cargo metadata --locked --no-deps --format-version 1 for the root, Python binding, and KVBM manifests
• git diff --check
• verified the PyPI resolver stub points to https://pypi.nvidia.com/ and that final macOS, manylinux x86_64, and manylinux aarch64 0.12.0 wheels are present

Summary by CodeRabbit

• Chores

  • Updated the project’s AISimulate components from development build 0.12.0-dev.2 to the stable 0.12.0 release.
  • Aligned supported Python environments, container setup, benchmarks, and native integrations with the stable release.
  • Updated consistency checks to validate the stable version across the project.

• Tests

  • Adjusted dependency validation to reflect the released 0.12.0 version.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 3df82844-15ab-48e6-b0ad-102a00f36a96

📥 Commits

Reviewing files that changed from the base of the PR and between dc0f538 and c93dfd7.

⛔ Files ignored due to path filters (3)

• Cargo.lock is excluded by !**/*.lock
• lib/bindings/kvbm/Cargo.lock is excluded by !**/*.lock
• lib/bindings/python/Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (6)

• Cargo.toml
• benchmarks/pyproject.toml
• container/deps/requirements.aisimulate.txt
• lib/bindings/python/Cargo.toml
• pyproject.toml
• tests/dependencies/test_aiconfigurator_consistency.py

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

Walkthrough

The pull request updates Rust and Python aisimulate dependencies from development version 0.12.0.dev2 to stable version 0.12.0. The dependency consistency test now expects the stable Rust version.

Changes

Stable aisimulate dependency update

Layer / File(s)	Summary
Update dependency pins and validation Cargo.toml, lib/bindings/python/Cargo.toml, pyproject.toml, benchmarks/pyproject.toml, container/deps/requirements.aisimulate.txt, tests/dependencies/test_aiconfigurator_consistency.py	Rust, Python, container, and test dependency declarations now use the stable 0.12.0 release. The optional Python binding dependency retains its python feature.

Priority: ⬇️ Low

Estimated code review effort: 1 (Trivial) | ~5 minutes

Merge Risk: ⚪ Minimal · up to c93df

The dependency update is internally consistent and presents no identified merge-blocking risk.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description provides a clear summary and validation details, but it does not follow the required template. It omits the required Related Issues section and does not state whether the pull request …	Add the required Related Issues section. Either link the relevant issue with entries such as "Closes #XXXX" or check the confirmation that no related issue exists. Add the template sections for Overview, Details, and Where should the review…
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files. (5 skipped: 5 …	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: updating AISimulate dependencies to version 0.12.0.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Description check

Explanation

The description provides a clear summary and validation details, but it does not follow the required template. It omits the required Related Issues section and does not state whether the pull request is linked to an issue.

Resolution

Add the required Related Issues section. Either link the relevant issue with entries such as "Closes #XXXX" or check the confirmation that no related issue exists. Add the template sections for Overview, Details, and Where should the reviewer start, or map the existing Summary and Validation content to those sections.

Full details: Docstring Coverage

Explanation

Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files. (5 skipped: 5 unsupported.)

• Fix all pre-merge checks with AI

---

_{Comment @coderabbitai help to get the list of available commands.}

[dmitry-tokarev-nv]

Approved

I found no defects. Every place that names the AISimulate version is updated, and the new version behaves the same as the old one for the way this repository calls it.

What I checked, and the measurements

Relationship to #14541. Not a duplicate. #14541 moved 0.1.0.dev2 to 0.12.0.dev1 and carried the source changes for the queryable version slots. This pull request moves 0.12.0.dev2 to the final 0.12.0 on the same line. Nothing in #14541 is repeated here.

Every place the version appears. All six declarations plus the three lockfiles are in the diff. A repository-wide search for 0.12.0.dev and 0.12.0-dev returns nothing.

File	Key	Value
Cargo.toml:59	workspace.dependencies.aisimulate-core	=0.12.0
lib/bindings/python/Cargo.toml:63	dependencies.aisimulate-core	=0.12.0
pyproject.toml:17	project.dependencies	==0.12.0
benchmarks/pyproject.toml:42	project.dependencies	==0.12.0
container/deps/requirements.aisimulate.txt:5	requirement line	==0.12.0
tests/dependencies/test_aiconfigurator_consistency.py:96	expected value	=0.12.0
Cargo.lock, lib/bindings/python/Cargo.lock, lib/bindings/kvbm/Cargo.lock	[[package]] aisimulate-core	0.12.0

container/context.yaml holds no AISimulate key, so there is no per-device default to split. There is no uv.lock, and no Helm chart, operator manifest, or recipe names the version.

Pin, not floor. Every Python declaration uses == and every Rust declaration uses =. uv pip install aisimulate==0.12.0 on macOS arm64 installed exactly 0.12.0. PyPI lists 0.1.0.dev1, 0.1.0.dev2, 0.12.0.dev1, 0.12.0.dev2, 0.12.0, and 0.12.0 is the latest. The crates.io checksum for aisimulate-core 0.12.0 is 7d266db788afdb866421aad39f8fbc7096dba9517105ff7d770ee83ad70f82bb, which matches all three lockfiles.

Both versions run the calling code. The hazard from the earlier bump was that version 0.12 rejects a raw version string. I installed 0.12.0.dev2 and 0.12.0 side by side and called the perf-database entry points that this repository uses, which are get_database, get_supported_databases, get_latest_database_version, get_version_slots, and resolve_query_version. For all 28 system and backend pairs, the slot name current resolved to a loaded database on both versions, with zero errors, and the two outputs are byte identical.

Control, to prove the probe can tell the cases apart: a raw version such as 9.9.9 raises ValueError with an old-style raw version query message, None returns no database, and previous raises where no such slot exists. Those results are also byte identical across the two versions.

No call site passes a raw version. DEFAULT_BACKEND_VERSIONS in lib/bindings/python/src/dynamo/_internal/aic.py:19 maps every backend to current, and resolve_backend_version only keeps a version that the caller supplied.

Measured test selection. Marker expression pre_merge and unit and gpu_0 over tests/dependencies/ collects 23 of 29 tests. All five tests in test_aisimulate_consistency.py and the AIC test in test_aiconfigurator_consistency.py are in the selected set. Running them on this branch gives 7 passed.

The same 7 tests on the merge base with 0.12.0 installed give 1 failed, 6 passed. The failure is test_installed_aisimulate_matches_the_declared_release, which expected 0.12.0.dev2. The guard therefore pins the version and does not pass both ways.

Where verification stopped. I ran everything on macOS arm64 with Python 3.12. I did not build a container image and I did not run the planner, profiler, or replay suites, because those need the compiled Dynamo bindings. benchmarks/pyproject.toml is the one declaration that test_aisimulate_consistency.py does not compare, so a future bump can miss it silently. That gap exists on the base branch too, so it is not a defect in this pull request.

[dmitry-tokarev-nv]

There are several mentions of aisimulate==0.1.0.dev2 and aisimulate==0.1.0.dev1 in the docs I think we should bump those too.
e.g.

• https://github.com/ai-dynamo/dynamo/blob/main/docs/fern/pages/developer-guide/knowledge-base/modular-components/ai-simulate-experimental/sweeper-experimental/quickstart.md?plain=1#L15
• https://github.com/ai-dynamo/dynamo/blob/main/docs/fern/pages/developer-guide/knowledge-base/modular-components/ai-simulate-experimental/sweeper-experimental/dynamo-integration.md?plain=1#L20-L33

[saturley-hall]

Using the NVIDIA-hosted wheel directly here is fine as an interim solution. The current concern with publishing the complete artifact to public PyPI is its size. Once the required data is pulled separately instead of being embedded in the wheel, the smaller wheel can be published publicly without that registry-size concern. We should keep --only-binary=:all: so this stage continues to guarantee that it is staging an actual wheel rather than the public source-distribution stub.

[github-actions]

🌿 Fern Docs Preview: https://nvidia-preview-01a0b13c-7bba-740e-bd48-61db4a9737f9.docs.buildwithfern.com/dynamo/dev

[dmitry-tokarev-nv]

Re-approved at a50038afcc3a126beb0e18f61629bc444312c6b8. The two new commits are a clean fast-forward on top of the commit I approved before, and the merge base with main did not move.

The version pin did not change. All six declarations and all three lockfiles still say 0.12.0, and the crate checksum still matches.

My earlier ask about the docs is addressed. Every install command now reads aisimulate==0.12.0. The remaining 0.1.0.dev1 lines are deliberate notes that mark a retired API, not stale pins.

The Dockerfile change is correct and it is the real fix for the red image builds. Public PyPI holds only a 20 KB source stub for 0.12.0, so --only-binary=:all: could never find a wheel there. This was not index lag. The NVIDIA index serves the real wheels, and --only-binary=:all: stays in place as you asked.

Measured: push shape, pin, index contents, and the test

Push shape, from the compare API: status: ahead, ahead_by: 2, behind_by: 0, merge_base_commit equal to the approved commit. Merge base with main is 0f061b772b8a2acd6beb08c9887e8a0acb318309 before and after the push.

Declared version at the live head, by key:

Location	Key	Value
Cargo.toml:59	aisimulate-core	=0.12.0
lib/bindings/python/Cargo.toml:63	aisimulate-core	=0.12.0
pyproject.toml:17	aisimulate	==0.12.0
benchmarks/pyproject.toml:42	aisimulate	==0.12.0
container/deps/requirements.aisimulate.txt:5	aisimulate	==0.12.0
Cargo.lock	aisimulate-core	0.12.0, 7d266db7…f82bb
lib/bindings/python/Cargo.lock	aisimulate-core	0.12.0, 7d266db7…f82bb
lib/bindings/kvbm/Cargo.lock	aisimulate-core	0.12.0, 7d266db7…f82bb

The crates.io index gives the same checksum for aisimulate-core 0.12.0, and the crate is not yanked.

Index contents. https://pypi.org/pypi/aisimulate/0.12.0/json lists one file, aisimulate-0.12.0.tar.gz, 20458 bytes, no wheel. The old command therefore failed for every platform:

$ pip download --only-binary=:all: --no-deps --dest ./dist2 "aisimulate==0.12.0"
ERROR: Could not find a version that satisfies the requirement aisimulate==0.12.0
       (from versions: 0.1.0.dev2, 0.12.0.dev1, 0.12.0.dev2)

The new command works. Python 3.12, macOS arm64:

$ pip download --only-binary=:all: --no-deps --no-index \
    --find-links https://pypi.nvidia.com/aisimulate/ --dest ./dist "aisimulate==0.12.0"
Saved ./dist/aisimulate-0.12.0-cp311-abi3-macosx_11_0_arm64.whl

The downloaded file hashes to bf446cd872088fccc8f830b438aabc1ac8cfed8ec657b466c7c674a8936103ab, which is the hash in the index anchor. The index also carries cp311-abi3 wheels for manylinux_2_28_x86_64 and manylinux_2_28_aarch64, so both build architectures are covered.

tests/dependencies/test_aisimulate_consistency.py passes at the live head, 5 of 5, with aisimulate==0.12.0 installed. The new assertion pins: removing the --find-links line from the template makes test_container_stages_the_published_aisimulate_wheel fail.

[saturley-hall]

This was determined a candidate for admin merging as it was a low risk change as the diff between 0.12.0.dev2 and 0.12.0 was negligible and one of the frontend architectures succeeded so it is unlikely there is a frontend container-specific failure mode.
@nvda-mesharma