This folder contains the modern local harness for framework comparison work. It keeps app loading checks separate from actual benchmark runs.

The comparable apps live in apps/:

• async-framework
• js-only
• react
• qwik-v1
• qwik-v2
• solid-v1
• solid-v2

async-framework uses the local root @async/framework browser source through the benchmark server, so its result version follows the package being tested. Qwik v2 and Solid v2 are beta package lines as of June 20, 2026. Their folders stay structurally aligned with the v1 apps except for package imports and build configuration required by those package lines.

cd benchmark
npm ci --registry=https://registry.npmjs.org
npm run install-apps --registry=https://registry.npmjs.org

If npm is configured to use a local registry on 127.0.0.1:4873, start that registry first or pass --registry=https://registry.npmjs.org.

Use this for cheap local verification. It builds the seven apps, starts or reuses the local server, loads each app page in Chromium, and verifies the shared controls are present. It does not click row-operation buttons, collect timings, or write trace data.

From the repository root:

pnpm run benchmark:apps

From this folder:

npm run apps

To only load already-built apps:

npm run app-health

Use this when you only need to know that the selected app pages load, shared UI is present, and browser requests complete without page errors. It does not click row-operation buttons, collect timings, or write trace data.

npm run benchmark:smoke

Use these only when you intend to run benchmark operations:

npm run benchmark:run -- --benchmark 01_ --iterations 1
npm run benchmark:results

By default, benchmark runs include async-framework plus the JavaScript-only, React, Qwik, and Solid apps. To select a subset:

npm run benchmark:run -- --framework async-framework react solid-v2 --benchmark 01_ --iterations 3

The benchmark runner uses Playwright and Chromium CDP trace events for CPU/script/paint metrics. JSON results are written under runner/results/, and raw traces are written under runner/traces/.

The results UI reads runner/results/latest.json through the benchmark server. Generate that file with a benchmark run first:

npm run benchmark:run -- --benchmark 01_ --iterations 1

Then start the benchmark server and open the results route:

npm start

Open http://localhost:8080/results.

Use PORT=8090 npm start to serve the UI on a different port. Smoke and app health checks do not write runner/results/latest.json; if no benchmark results exist yet, the UI will show a message asking you to run npm run benchmark:run.

The local benchmark server starts automatically when /ls is not already reachable on localhost:8080.