test rapids build backend

[Intron7]

No description provided.

[codecov-commenter]

⚠️ JUnit XML file not found

The CLI was unable to find any JUnit XML files to upload.
For more help, visit our troubleshooting guide.

[Intron7]

@coderabbitai review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

• New Features

  • Switched to RAPIDS build backend for automatic CUDA-suffixed wheel packaging and centralized dependency management.

• Documentation

  • Updated installation and contribution guides to reflect simplified CUDA dependency handling.

• Chores

  • Updated Python runtime version to 3.14 across CI and conda environments.

Walkthrough

The pull request migrates the build backend from scikit_build_core to rapids_build_backend, introducing a new dependencies.yaml for RAPIDS dependency management. Python version requirements are updated from 3.13 to 3.14 across CI and conda environments. The CI workflow removes inline CUDA-specific package mutation logic, delegating this responsibility to the build backend.

Changes

Cohort / File(s)	Summary
Build System Migration pyproject.toml	Switches PEP 517 backend from scikit_build_core.build to rapids_build_backend.build; moves nanobind and setuptools-scm to backend-specific config; simplifies optional dependencies by removing CUDA-specific extras and replacing with empty rapids placeholder.
Workflow Restructuring .github/workflows/publish.yml	Removes inline pyproject.toml mutation logic for CUDA versioning; shifts wheel builds to use CIBW_CONFIG_SETTINGS for CUDA architectures and explicit RAPIDS_DISABLE_CUDA=false; deepens git checkout for sdist with RAPIDS_DISABLE_CUDA=true.
Dependencies Configuration dependencies.yaml	New file mapping rapids optional extra with CUDA-specific dependency matrices (12.\* and 13.\*) installing cupy-cuda{12,13}x and RAPIDS components (cudf, cuml, cugraph, cuvs) with >=25.10 minimum version.
Python Version Updates ci/environment.yml, ci/environment_alpha.yml, conda/rsc_rapids_26.04_cuda12.yml, conda/rsc_rapids_26.04_cuda13.yml	Update Python runtime requirement from 3.13 to 3.14 in conda environment definitions.
Test Configuration hatch.toml	Shifts CUDA selection mechanism from feature-based (rapids-cu12/rapids-cu13) to direct extra-dependencies approach with cupy-cuda{12,13}x packages alongside RAPIDS extras.
Documentation Updates docs/contributing.md, docs/installation.md, docs/release-notes/0.15.0rc6.md, docs/release-notes/index.md	Updates Python version labels from 3.13 to 3.14; replaces wheel packaging documentation from inline CI script to RAPIDS build backend; revises source installation guidance for RAPIDS dependencies; adds release notes entry for version 0.15.0rc6.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	No pull request description was provided by the author, making it impossible to assess whether any description exists that relates to the changeset.	Add a pull request description that explains the migration to rapids-build-backend, the changes to build configuration, Python version updates, and the rationale for these changes.
Title check	❓ Inconclusive	The title 'test rapids build backend' is vague and uses generic phrasing that doesn't clearly convey the actual significant changes being made to the build system.	Revise the title to be more specific about the primary change, such as 'Switch to rapids-build-backend for wheel building' or 'Migrate build system to rapids-build-backend with dependencies.yaml'.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch rapids-build-system

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

🧹 Nitpick comments (1)

dependencies.yaml (1)

16-31: Optional: consider upper bounds on the published rapids extra.

The development pins in hatch.toml use <26.8 for cudf/cuml/cugraph/cuvs, but the user-facing extra generated from this file uses only >=25.10 with no upper bound, and cupy-cuda{12,13}x is fully unpinned. If RAPIDS 26.x introduces breaking API changes to cudf/cuml/cugraph/cuvs (or cupy ships a new major), pip install rapids-singlecell[rapids] --extra-index-url … could resolve to versions that haven't been validated against this release.

Mirroring the hatch upper bound (or at least bounding the next major) would keep the published extra aligned with what's actually being tested.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@dependencies.yaml` around lines 16 - 31, The published extra currently uses
open-ended pins for cupy-cuda12x/cupy-cuda13x and uses lower bounds only for
cudf-cu12/cudf-cu13, cuml-cu12/cuml-cu13, cugraph-cu12/cugraph-cu13,
cuvs-cu12/cuvs-cu13; update the matrix package entries to include an upper bound
that mirrors the hatch dev pin (e.g. <26.8 or at least <27.0) for
cudf/cuml/cugraph/cuvs and add a bounded pin or major-only upper bound for
cupy-cuda12x/cupy-cuda13x so the rapids extra resolves to tested major versions
(refer to the package names in the matrix blocks: cupy-cuda12x, cupy-cuda13x,
cudf-cu12, cuml-cu12, cugraph-cu12, cuvs-cu12 and their -cu13 variants).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@dependencies.yaml`:
- Around line 16-31: The published extra currently uses open-ended pins for
cupy-cuda12x/cupy-cuda13x and uses lower bounds only for cudf-cu12/cudf-cu13,
cuml-cu12/cuml-cu13, cugraph-cu12/cugraph-cu13, cuvs-cu12/cuvs-cu13; update the
matrix package entries to include an upper bound that mirrors the hatch dev pin
(e.g. <26.8 or at least <27.0) for cudf/cuml/cugraph/cuvs and add a bounded pin
or major-only upper bound for cupy-cuda12x/cupy-cuda13x so the rapids extra
resolves to tested major versions (refer to the package names in the matrix
blocks: cupy-cuda12x, cupy-cuda13x, cudf-cu12, cuml-cu12, cugraph-cu12,
cuvs-cu12 and their -cu13 variants).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: bec9e135-bf96-42e5-b1f4-bc01bc384e37

📥 Commits

Reviewing files that changed from the base of the PR and between 39f2ebf and 1a665f5.

📒 Files selected for processing (12)

• .github/workflows/publish.yml
• ci/environment.yml
• ci/environment_alpha.yml
• conda/rsc_rapids_26.04_cuda12.yml
• conda/rsc_rapids_26.04_cuda13.yml
• dependencies.yaml
• docs/contributing.md
• docs/installation.md
• docs/release-notes/0.15.0rc6.md
• docs/release-notes/index.md
• hatch.toml
• pyproject.toml