[ovflowd]

Description

This PR replaces tinyglobby and glob-parent in @doc-kit/core with a small #utils/glob.mjs on Node.js's built-in fs.globSync (stable since 22.17 and 24.0; doc-kit targets Node.js 24, per .nvmrc):

• findFiles(patterns, { cwd, ignore }) returns the same POSIX, cwd-relative paths as tinyglobby, absolute patterns included, and leaves out directories.
• globParent(pattern) returns the directory before the first glob segment.

The ast and ast-js generators and two tests use them, and tinyglobby also leaves packages/node's devDependencies. Both packages stay in the lockfile, since Vite and our dev tooling use them, so this drops two direct runtime dependencies of @doc-kit/core rather than lockfile entries.

Validation

node --run test (754/754), node --run lint and node --run format:check are clean, and the frozen-lockfile install passes. The new glob.test.mjs covers matching files without directories, ignore, absolute patterns, a URL as cwd, and globParent for glob patterns, plain paths and the .// fallbacks.

Compared against tinyglobby 0.2.17 and glob-parent 6.0.2 on nodejs/node's own files, both find the same files and the same parent for every pattern:

Pattern	Files
doc/api/*.md (ignoring quic.md)	69
lib/*.js	56
lib/**/*.js (ignoring lib/internal/**)	73
an absolute <checkout>/doc/api/*.md	70
doc/api/{fs,path}.md	2

Related Issues

Came out of the dependency audit in #1178.

Check List

• I have read the Contributing Guidelines and made commit messages that follow the guideline.
• I have run node --run test and all tests passed.
• I have check code formatting with node --run format:check & node --run lint.
• I've covered new added functionality with unit tests if necessary.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
api-docs-tooling	Ready	Preview	Oct 26, 2026 6:08pm UTC

[codecov]

Codecov Report

❌ Patch coverage is 99.26471% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 93.47%. Comparing base (ec7751f) to head (fe73cbd).

Files with missing lines	Patch %	Lines
packages/core/src/generators/ast/generate.mjs	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1180      +/-   ##
==========================================
+ Coverage   93.44%   93.47%   +0.02%     
==========================================
  Files         278      280       +2     
  Lines       27820    27947     +127     
  Branches     2790     2811      +21     
==========================================
+ Hits        25997    26124     +127     
  Misses       1801     1801              
  Partials       22       22              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

api-links Generator

Performance estimate _{(single CI run)}

• Generation time: 5.6% slower (1.07 s → 1.13 s)
• Peak memory: 0.3% lower (289.13 MB → 288.23 MB)

json Generator

Performance estimate _{(single CI run)}

• Generation time: 13.3% faster (7.46 s → 6.47 s)
• Peak memory: 9.0% higher (1.18 GB → 1.29 GB)

legacy-html Generator

Performance estimate _{(single CI run)}

• Generation time: 68.9% slower (10.58 s → 17.87 s)
• Peak memory: 1.1% lower (1.79 GB → 1.77 GB)

legacy-json Generator

Performance estimate _{(single CI run)}

• Generation time: 25.7% faster (9.04 s → 6.72 s)
• Peak memory: 3.2% higher (1.15 GB → 1.19 GB)

llms-txt Generator

Performance estimate _{(single CI run)}

• Generation time: 5.2% slower (8.10 s → 8.52 s)
• Peak memory: 6.5% lower (1.19 GB → 1.11 GB)

orama-db Generator

Output size: 1 file changed · net +98.00 B

File size details

File	Main	PR	Change
orama-db.json	29.24 MB	29.24 MB	+98.00 B (+0.0%)

Performance estimate _{(single CI run)}

• Generation time: 4.6% faster (7.35 s → 7.01 s)
• Peak memory: 5.2% lower (1.28 GB → 1.21 GB)

web Generator

Performance estimate _{(single CI run)}

• Generation time: 10.4% slower (22.42 s → 24.76 s)
• Peak memory: 2.2% higher (1.90 GB → 1.94 GB)

[cloudflare-workers-and-pages]

🚀 Deploying Preview to Cloudflare 🚀

Preview Deployments by commit

Status	Deployment URL	Commit	Updated (UTC)	See this deployment's details
Build: Success ✅ View logs ↗	No Preview URL. Enable ↗	fe73cbd	2026-10-26T18:09:30.118Z	View logs ↗
Build: Success ✅ View logs ↗	No Preview URL. Enable ↗	fd5420a	2026-10-26T13:08:59.869Z	View logs ↗

[avivkeller]

Does this match everything?

[ovflowd]

Yup, I checked it against Node's own fs.globSync on real files (Node 24 and 26): it catches everything that's glob syntax there, */**, ?, [...] classes, {a,b}/{1..3} and all five extglobs. It's actually a bit closer than glob-parent was, a?c is a glob for fs.globSync (glob-parent missed it) and (a|b) is matched literally (glob-parent treated it as one). The only over-match is an unmatched { like a{b, which just makes the parent one level higher, same as glob-parent did.

While at it, I noticed fs.globSync reads \ as a path separator (no escapes), so doc\api\*.md gave . as the parent. Fixed in fe73cbd, with tests for these cases :)