v0.2.0 — production hardening + structural support

v0.2.0 — production hardening + structural support

First post-initial release. All improvements driven by real-world deployment.

DOM-race hardening

• Null-safe document.body access in extractor (was crashing mid-navigation)
• Explicit waitForFunction(() => document.body) before mutation observers attach
• Body-existence guards inside observer + scroll scripts

Type-import alignment

• import type { Page } now from playwright-core (was playwright)
• Drops the cross-package nominal-type mismatch consumers had to cast through
• Peer dep flips playwright → playwright-core
• DOM lib added to tsconfig so arrow functions referencing document type-check cleanly

Semantic table support

• <table> recognized + rendered as GitHub-flavored markdown
• <thead> / first-row-of-<th> treated as headers; <tbody> rows or direct <tr> walked
• Pipes/newlines in cells escaped; rows padded to column count
• New BodySegment kind: { kind: 'table', headers?, rows, caption? }
• 200 rows × 30 cols cap for snapshot-size safety

Shadow DOM piercing

• Walks el.shadowRoot.children before tag-specific handling
• Open shadow roots only (closed ones are unreachable to script by browser design)
• Catches modern web components (Salesforce Lightning, MS 365, Lit/Polymer apps)

Iframe handling

• Same-origin iframes: walks contentDocument.body inline
• Cross-origin iframes: emits [IFRAME:n] tag with src URL + title
• New IFRAME_OPEN / IFRAME_CLOSE boundary tags around inlined content

Build

• typescript added to devDependencies
• playwright-core added to devDependencies (for build-time types)
• CI uses npm install (cross-platform compatible) instead of npm ci