Skip to content

Adopt ty-pre-commit hook; bump .ai-instructions (tiering + distill)#86

Merged
hmgaudecker merged 4 commits into
mainfrom
ai-instructions-ty-migration
Jun 15, 2026
Merged

Adopt ty-pre-commit hook; bump .ai-instructions (tiering + distill)#86
hmgaudecker merged 4 commits into
mainfrom
ai-instructions-ty-migration

Conversation

@hmgaudecker

Copy link
Copy Markdown
Collaborator

Migrate type checking to the official astral-sh/ty-pre-commit hook and pick up the ai-instructions overhaul.

prek run --all-files passes (incl. ty); full test suite green locally.

🤖 Generated with Claude Code

@codecov

codecov Bot commented Jun 15, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@hmgaudecker hmgaudecker changed the base branch from beartype to main June 15, 2026 15:33
- ty runs as the official astral-sh/ty-pre-commit hook (v0.0.49, --no-project),
  resolving imports from .pixi/envs/py314 via [tool.ty] environment.python.
- Remove the type-checking pixi env + feature (the old pinned ty + ty task).
- Track .ai-instructions on its main branch over HTTPS (.gitmodules) and bump it
  to c2701d4 (real tiering, distilled modules, ty-pre-commit boilerplate); add the
  AGENTS.md / GEMINI.md wrappers.
- Annotate `_get_sorted_dataset_merging_information`'s dict so ty can check it.
- Relock: pytask 0.6 (from #85), pinned `ty` dropped from the env.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@hmgaudecker hmgaudecker force-pushed the ai-instructions-ty-migration branch from cc8abd4 to 5dfccc3 Compare June 15, 2026 15:47
@hmgaudecker hmgaudecker requested a review from MImmesberger June 15, 2026 16:03
hmgaudecker and others added 3 commits June 15, 2026 19:23
* Apply boilerplate update for tier-b-research

- Switch .ai-instructions submodule URL to HTTPS, branch to main
- Update submodule to latest main
- Bump ruff hook v0.15.11 -> v0.15.13
- Add check-github-workflows hook
- Add per-rule descriptions to all per-file-ignores
- Bump setup-pixi v0.9.3 -> v0.9.5, pixi v0.62.2 -> v0.68.1
- Migrate CLAUDE.md content to AGENTS.md, shrink CLAUDE.md to @AGENTS.md
- Add GEMINI.md (@AGENTS.md) for Gemini CLI / roborev

* Bump .ai-instructions to e93e058 (add beartype rollout module)

* Bump .ai-instructions to 40e0ac6 (document claw callable-instance gotcha)

* Refresh CI/pre-commit pins and lock; pin ty to 0.0.23

- pixi 0.68.1 -> 0.69.0 in CI
- pre-commit: pyproject-fmt 2.21.1->2.21.2, ruff 0.15.13->0.15.14
- ty pinned to 0.0.23: 0.0.39 flags a real but out-of-scope
  type-narrowing issue in `_sort_dataset_merging_information`
- pixi.lock regenerated with pixi 0.69.0 so `pixi run --locked`
  is back-current

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Bump .ai-instructions to pick up sweep/pass/handler naming guidance

* Bump .ai-instructions to 82a2746 (jaxtyping sentinel-pickle upstream fix module)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Use keyword args for project-internal calls with 2+ positional args

Sweep applied across src/soep_preparation and tests. Rewrites only
calls whose target is a project-defined function, method, or class
with statically-known parameter names. Leaves stdlib, numpy, pandas,
jax, and other third-party calls alone.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Unpin ty; track latest

ty 0.0.39 narrows TypedDict-valued dict literals more strictly than 0.0.23,
so `_get_sorted_dataset_merging_information` needs an explicit annotation
on the accumulator dict for the narrowing to match the function's return
type.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Adopt ty-pre-commit hook; bump .ai-instructions to 80d7122

- ty runs as the astral-sh/ty-pre-commit hook (v0.0.49, --no-project),
  resolving third-party imports from .pixi/envs/py314 via
  [tool.ty] environment.python; drop the `ty` pixi task and dep.
- pre-commit.ci skips ty (no pixi/network); new GHA run-ty job covers it.
- AGENTS.md: ty now runs via `prek run --all-files`.
- Bump .ai-instructions submodule to 80d7122 (real tiering, distilled
  modules, ty-pre-commit boilerplate).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Re-pin .ai-instructions to merged main (c2701d4)

PR #9 was squash-merged, so the earlier 80d7122 pin is orphaned; c2701d4 is
the merged commit (includes the ty-jax-out-of-boilerplate cleanup).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
conda-forge superseded the `ca-certificates 2026.4.22` builds, so the locked
CI job (`pixi install --frozen`) could no longer load the lock ("refers to a
package that does not exist"). A targeted update is impossible because pixi
must first load the unloadable lock, so regenerate it from scratch against
current repodata. ca-certificates is now 2026.5.20; 47 tests pass.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Cover the module-selection and index-count sort directly, closing the patch
coverage gap on the type annotation added to that helper.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@hmgaudecker hmgaudecker merged commit 31d2992 into main Jun 15, 2026
6 checks passed
@hmgaudecker hmgaudecker deleted the ai-instructions-ty-migration branch June 15, 2026 19:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants