Add CodSpeed continuous performance benchmarks

[codspeed-hq]

This adds continuous performance measurement to the repository with CodSpeed, so parser and renderer regressions show up on every pull request.

What was added

Benchmark suite (benchmarks/*.bench.ts, Vitest bench)

The suite focuses on comark's own hot paths rather than comparisons with other parsers (the existing mitata scripts in benchmarks/ are left untouched and remain the place for cross-library comparisons). 28 benchmarks in total:

• parse.bench.ts — parsing small, medium and large documents, plus the effect of autoClose: false, registerDefaultPlugins: false, and creating a fresh parser per call.
• render.bench.ts — renderHtmlFromDocument, renderAnsiFromDocument and renderMarkdown on pre-parsed documents, plus one combined parse + render pass.
• streaming.bench.ts — autoCloseMarkdown on partial and complete documents, streaming-mode parsing, and re-parsing a full sequence of growing chunks (the LLM streaming path).
• plugins.bench.ts — per-plugin cost for toc, emoji, punctuation, footnotes, security and summary, individually and combined, against a baseline.

Fixtures live in benchmarks/fixtures.ts and are deterministic so results stay comparable across runs.

Tooling

• @codspeed/vitest-plugin added as a root dev dependency (through the catalog) and wired into vitest.config.ts. The plugin is a no-op outside of a CodSpeed run, so local vitest bench and the existing test/ suite are unaffected.
• benchmark.include scoped to benchmarks/**/*.bench.ts, matching how test.include is already scoped.
• New pnpm bench script.
• .github/workflows/codspeed.yml runs the benchmarks on main, on pull requests and on workflow_dispatch, using CodSpeedHQ/action@v5 in simulation mode with OIDC authentication. The Node/pnpm setup mirrors ci.yml.
• CodSpeed badge added to the README.

Dependency notes

• @codspeed/vitest-plugin has not widened its vite peer range to v8 yet. Without a workaround pnpm resolves the root workspace back to vite 7, which would silently downgrade vitest and nuxt at the root. A packageExtensions entry in pnpm-workspace.yaml widens that peer range, and vite is now an explicit root dev dependency pinned to the catalog version (^8.1.4) so the resolution stays on vite 8.
• The lockfile diff is larger than the change suggests: @codspeed/core pulls in axios, which changes a number of hashed peer-resolution keys. No package versions are downgraded.

Verification

Run locally with the CodSpeed CLI in simulation mode (codspeed run --mode simulation -- pnpm bench): all 28 benchmarks were collected and reported successfully. pnpm lint and pnpm typecheck pass.

Next steps

• Merge this PR so the baseline is recorded on main; the first pull requests afterwards will get performance reports.
• Optionally trigger the workflow manually (workflow_dispatch) to backfill historical data.
• Consider adding the Shiki highlighting plugin to the suite once you are comfortable with the extra CI time it costs.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
comark	Ready	Preview	Aug 28, 2026 7:48am
comark-json-render	Ready	Preview	Aug 28, 2026 7:48am
comark-nextjs	Ready	Preview	Aug 28, 2026 7:48am
comark-nuxt	Ready	Preview	Aug 28, 2026 7:48am
comark-svelte	Ready	Preview	Aug 28, 2026 7:48am
comark-sveltekit	Ready	Preview	Aug 28, 2026 7:48am
comark-twoslash	Ready	Preview	Aug 28, 2026 7:48am
comark-vue	Ready	Preview	Aug 28, 2026 7:48am

[pkg-pr-new]

Open in StackBlitz

comark

npm i https://pkg.pr.new/comark@389

@comark/angular

npm i https://pkg.pr.new/@comark/angular@389

@comark/ansi

npm i https://pkg.pr.new/@comark/ansi@389

@comark/html

npm i https://pkg.pr.new/@comark/html@389

@comark/nuxt

npm i https://pkg.pr.new/@comark/nuxt@389

@comark/react

npm i https://pkg.pr.new/@comark/react@389

@comark/svelte

npm i https://pkg.pr.new/@comark/svelte@389

@comark/vue

npm i https://pkg.pr.new/@comark/vue@389

commit: 0a3bdec

[codspeed-hq]

Congrats! CodSpeed is installed 🎉

🆕 28 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• medium document with a fresh parser instance: 11.3 ms
• medium document without auto-close: 7.9 ms
• medium document without default plugins: 5.9 ms
• large document: 91.9 ms
• medium document: 10.3 ms
• small document: 3.5 ms
• all of the above combined: 14.2 ms
• baseline (default plugins only): 14.6 ms
• emoji: 18 ms
• footnotes: 9.3 ms
• punctuation: 9.4 ms
• security: 11.9 ms
• summary: 9.2 ms
• toc: 10 ms
• medium document: 11.7 ms
• large document: 38.5 ms
• medium document: 8.6 ms
• large document: 40.6 ms
• medium document: 14.3 ms
• small document: 912.5 µs
• ...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---