Tlaloc-Es/killpy

Motivation --min-size landed in #32 and works well. While reviewing it a handful of small rough edges came up that are not worth blocking a release over, but are each a self-contained improvement. They are grouped here so they can be picked off individually — any one of them is a reasonable first contribution. Where killpy/commands/_utils.py — SizeParamType, _SIZE_PATTERN, _SIZE_UNITS killpy/commands/{list,delete,stats,doctor,find}.py — the --min-size option declarations killpy/commands/stats.py — the --history early return Proposed approach Each item is independent; a PR fixing just one is welcome. stats --history accepts --min-size and ignores it. --history returns stored totals before any scan happens, so scan filters cannot apply. Today the value is parsed and validated, then dropped silently. It is currently documented as a caveat; raising a usage error when the two are combined would be more honest. Note --path behaves the same way, so whatever we decide should probab

Motivation On Windows the only install paths today are pip/pipx/uv, which assume a working Python setup. Scoop and winget reach developers before that point — but both work best with a self-contained executable. Proposed approach Two-step discussion: (1) decide whether to ship a standalone binary per release (PyInstaller or shiv, built and attached to the GitHub Release by the publish workflow); (2) if yes, add a Scoop bucket manifest and a winget manifest pointing at that artifact. Textual TUIs work in Windows Terminal, so the blocker is packaging, not runtime. Input from Windows users is especially welcome. Acceptance criteria Decision recorded on shipping a standalone Windows executable If approved: release workflow attaches the binary, and Scoop/winget manifests are published and documented

Motivation Arch users install CLI tools from the AUR by default, and an AUR package is a community-maintainable discovery surface. Proposed approach Publish a killpy PKGBUILD built from the PyPI sdist using python-build + python-installer (standard Python packaging guidelines). Dependencies map to python-click, python-rich, and python-textual. Include .SRCINFO and keep the version pinned to the latest release. Co-maintainers welcome — if you package for Arch and want to adopt it, say so here. Acceptance criteria killpy installable from the AUR (e.g. yay -S killpy) Install documented in README/docs

Motivation Conda users are exactly the audience with multi-GB envs directories, and conda install -c conda-forge killpy keeps them inside their own toolchain. A conda-forge feedstock is also a permanent, community-maintained distribution channel. Proposed approach Generate a recipe with grayskull pypi killpy and submit it to conda-forge/staged-recipes. Runtime dependencies (click, rich, textual) are already available on conda-forge, so the recipe should be mostly mechanical. After the feedstock is created, releases are bumped automatically by the conda-forge bot. Acceptance criteria Feedstock merged and conda install -c conda-forge killpy / mamba install killpy works Install documented in README/docs

Motivation brew install is the default install path for a large share of macOS and Linux developers, and a formula is a permanent discovery surface. A disk-cleanup tool is a natural fit for Homebrew's audience. Proposed approach Start with a personal tap (Tlaloc-Es/homebrew-killpy) hosting a formula built from the PyPI sdist (Python formula with virtualenv resources, e.g. generated via brew update-python-resources). Once install numbers and notability criteria are met, propose the formula to homebrew-core. Open to other approaches. Acceptance criteria brew install Tlaloc-Es/killpy/killpy works on macOS and Linux Formula updates on new releases (can be automated in the publish workflow) Install section of the README/docs lists the Homebrew path