GitHub Linguist — Li usage evidence (WP3)¶

Documents current GitHub Search state for extension .li and the human-only steps to reach Linguist's usage bar for Li Langverse Li (not other ecosystems that happen to use .li).

Agents: maintain this doc's process and audit snapshots in lic; do not claim the threshold is met until Julian records passing searches below.

Current state (audit 2026-05-25)¶

Collision risk (.li)¶

The extension is not reserved for Li Langverse today. Indexed .li files may include:

• LemonOS and similar projects (interface / tooling paths under other orgs)
• Unclassified or misclassified blobs until Linguist lists Li
• Hobby one-off repos

Maintainers require ≥2000 indexed files per extension (multi-file-per-repo case) and a reasonable spread across unique user/repo pairs — see Linguist usage requirements. If one user dominates, add -user:… filters in the PR and in weekly audits.

Cannot merge Li into Linguist until the bar is met or a maintainer documents an explicit exception in the PR thread.

Goal¶

GitHub Search (and Linguist reviewers) must see thousands of public .li files that are unmistakably Li (def/proc, requires/ensures, import, lic toolchain layout), with many repos outside a single maintainer account.

Step-by-step — Julian (human)¶

1. Make li-langverse/lic searchable¶

1. Repository public (already true for lic).
2. Ensure .li is not gitignored at repo root and in packages (li-tests/, examples/, packages/*/src, packages/*/li-tests/).
3. Push real sources regularly; avoid force-pushing away history search indexes rely on.
4. Optional: add a short README badge or docs link so clones mirror Li layout (li.toml, packages/).

2. Seed public org repos¶

For each mirror or package repo you want counted:

1. Public GitHub repo under li-langverse (or trusted community orgs).
2. Include multiple .li files per repo (tests + src), not a single stub file.
3. Use normal Li idioms (contracts, imports) so heuristic/classifier training matches WP2 samples.
4. Prefer unique user/repo pairs over one mega-repo with all bytes.

Suggested seed order:

Do not flood GitHub with duplicate empty files or license-incompatible copies — reviewers spot artificial inflation.

3. Weekly search cadence (evidence log)¶

Every Monday (or pre-PR), Julian runs and saves URLs + screenshot or copied count in a private log (issue, gist, or PR draft). Suggested queries:

extension:li NOT is:fork
org:li-langverse extension:li NOT is:fork
extension:li "requires" NOT is:fork -user:li-langverse
extension:li "ensures" NOT is:fork
extension:li "import " NOT is:fork org:li-langverse

Record in a table:

When global ≥2000 and org slice shows hundreds+ with diverse repos, attach the same URLs to the Linguist PR body (G2 in playbook).

4. PR body evidence (after bar met)¶

Paste into upstream PR (template bullets):

• Primary: extension:li NOT is:fork — state count and confirm manual spot-check of random repos.
• Li-specific: extension:li "requires" NOT is:fork (and/or ensures, proc).
• If needed: -user:<your-account> to show distribution beyond language owners.
• Link SAMPLES_LICENSES.md commit and grammar MIT license.

What agents may do¶

• Update audit table in this file when given fresh search counts (no rounding up).
• Improve seeding docs/scripts in lic that help humans publish mirrors.
• Prepare WP6 patch bundle and playbook cross-links.

What agents must not do¶

• Claim usage threshold is met without Julian's logged search evidence.
• Mass-create junk public repos or commit to github-linguist/linguist.
• Replace Julian as PR author on Linguist.

Unblock Linguist PR¶

Related¶

• Playbook G0–G8
• PATCH_INSTRUCTIONS.md