polarizer

CLI to transform playwright-cli snapshot YAML into compact, LLM-friendly text (or JSON).

Install

npm install -g @waynest/polarizer
# or (no install)
npx -y @waynest/polarizer --help

Quick start

Prereqs:

• Node.js (>= 20)
• playwright-cli installed and working

Run on a saved snapshot:

polarizer --input path/to/page.yml --filter interactive --text context --format text

CLI usage

polarizer --help

Common options:

• --input <file>: snapshot YAML file (default: stdin)
• --format <text|compact|json>: output format
• --filter <all|interactive>: keep everything or interactive elements only
• --text <none|context|all>: control non-interactive text inclusion
• --max-tokens <n>: approximate token budget (adds a truncation marker when used)
• --base-url <url>: resolve relative /url: values to absolute URLs

Diff two snapshots:

polarizer diff --prev before.yml --curr after.yml --format compact

End-to-end (real page) validation

Hacker News:

npm run e2e:hn

This writes artifacts under artifacts/e2e/hn-*/, including:

• raw.yml (playwright snapshot)
• polarized.txt (full output)
• polarized.llm.txt (token-limited output)

Standalone quality judge (Codex):

npm run judge:hn

The judge report is saved to artifacts/e2e/hn-*/judge.md.

GitHub repo (microsoft/playwright):

npm run e2e:github-playwright
npm run judge:github-playwright

WSJ homepage (may return an anti-bot interstitial depending on network conditions):

npm run e2e:wsj
npm run judge:wsj

Diff validation (controlled page mutation via TodoMVC):

npm run e2e:diff

Tests / checks

npm test
npm run typecheck
npm run lint

Development

npm install
npm run build