✨ Point the site at a local effectionx checkout

[taras]

Motivation

The site clones thefrontside/effectionx into build/clones and reads its packages from main, so there is no way to see a package you are working on — a new package, a renamed one, an edited description — until the work is pushed and merged. Everything that renders /x, the header, and llms.txt goes through useClone(), which only knows how to clone and git reset --hard.

Approach

initClones takes a checkouts option, a map of owner/repo to a directory to use in place of a clone, and main.tsx wires EFFECTIONX_DIR to thefrontside/effectionx:

EFFECTIONX_DIR=../effectionx deno task dev

A mapped repository never reaches cloneOrRefresh, so it is never cloned, fetched or reset — the site reads whatever branch and uncommitted work are on disk, and cannot disturb a checkout you are working in. Paths are resolved and checked before the clones directory is emptied, so a path that is not there fails at startup with cannot use /path as a local checkout of thefrontside/effectionx: no such directory rather than on the first request that needs the repository. Each mapping is logged at startup.

Verified against a real checkout on a feature branch: /x listed its 26 packages, /x/task-buffer rendered v1.4.0, the version on that branch, where origin/main has 1.3.3 — so the page is reading the working tree, not a clone. build/clones/ stayed empty, and the checkout came through with the same HEAD and a byte-identical git status, untracked files included. Pointing EFFECTIONX_DIR at a throwaway two-package directory showed only that directory's package in llms.txt and on /x.

www/deno.json's test task gets --allow-env, because importing lib/clones.ts pulls in the process helpers, which read OSTYPE at module load. That is the same one-line change as #1241; whichever merges second merges cleanly, since both make the line identical. Worth knowing for anything added to www later: the root deno task test that CI runs picks up the www workspace, and it grants --allow-env but not --allow-write, so the test here sticks to pure resolution and touches no files. Both entrypoints pass — deno task test in www, and the root one at 39 passed.

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/effection@1242

commit: 7ba3c74

[codspeed]

Merging this PR will not alter performance

✅ 6 untouched benchmarks

---

_{Comparing tm/local-effectionx-checkout (7ba3c74) with v4 (37a6177)}

[github-actions]

🚀 Deploy Preview Ready!

• Stable Review Link: https://pr-1242--effection.netlify.app
• Latest Commit Snapshot: https://6ab94f1e2dcca9ab1138929c--effection.netlify.app

[cowboyd]

I'm going to go ahead and approve this with the proviso that we should 100% be using configliere for argument parsing and variable binding precisely for this use case.