v0.1.0 is here — and it hasn't slowed down since

Our last post said v0.1.0 was coming soon. It's not soon anymore — v0.1.0 shipped on September 4th, allwright's first minor version after more than sixty 0.0.x releases. And rather than pausing to admire the number, we've since cut four more releases in three days. This is where things actually stand at v0.1.4.

What v0.1.0 itself was

Worth being honest about this the same way we were in the last post: v0.1.0 didn't ship a pile of new surfaces the moment the tag was cut. It was the line in the sand we said it would be — the point where what had already landed in patch releases (web and Android automation, five client languages, one API shape, project scaffolding) added up to something worth calling a real milestone, instead of another 0.0.x bump.

The version number finally meaning something is exactly the point. What's changed since is everything that's landed underneath that number in the few days after.

What's landed since: v0.1.1 – v0.1.4

v0.1.1 — Web accessibility snapshots

Web pages in all five clients gained accessibility_snapshot / AccessibilitySnapshot / accessibilitySnapshot, returning a structured document of every node's role, name, states, properties, and children — including cross-origin iframes — as JSON (default) or YAML:

const tree = JSON.parse(await page.accessibilitySnapshot());
const yaml = await page.accessibilitySnapshot({ format: "yaml" });

Accessible-name and role computation is Allwright-owned code, informed by the W3C accessibility specs and by studying Playwright's behavior as a reference — no vendored Playwright code, no dom-accessibility-api dependency. See the web plugin's accessibility documentation for the full contract and its limits.

v0.1.2 — Snapshot modes, including an AI-ready one

A snapshot now has a mode alongside its format:

const snapshot = JSON.parse(await page.accessibilitySnapshot({ mode: "ai" }));
// For a node in the snapshot:
await page.click(`[aria-ref="${node["aria-ref"]}"]`);

References are opaque, stable while the element/role/name stay unchanged, and reset on navigation — useful for driving a page from a snapshot instead of a hand-written selector.

v0.1.3 — Semantic web locators, and excluding matches

Pages and locators picked up Playwright-style semantic builders across all five clients — getByRole, getByText, getByLabel, getByPlaceholder, getByAltText, getByTitle, and getByTestId — chainable with CSS/XPath and filterable by has, hasNot, hasText, hasNotText, and visible:

await page.getByRole("listitem")
  .filter({ has: page.getByRole("heading", { name: "Beta", exact: true }) })
  .getByRole("button", { name: "Buy", disabled: false })
  .click();

Alongside them, locator.not(otherLocator) excludes matching elements (Python not_, Go Not, Rust not(&other)), and Vitest assertions gained matching negation for both web and Android — expect(locator).not.toHaveText("Loading") and expect(locator).not().toBeVisible(), retrying like every other assertion. See the web locator reference for the full builder and filter set.

v0.1.4 — Accessibility snapshots reach Android

The same accessibility_snapshot capability web got in v0.1.1 is now available on Android, sharing the same versioned document structure and JSON/YAML encoding as web — one accessibility model across both surfaces instead of two.

const tree = JSON.parse(await androidApp.accessibilitySnapshot());

Where to go next

• Read the Changelog for the full, continuously updated release history.
• Check Availability for the current, capability-by-capability picture of web and Android automation.
• Read the road to v0.1.0 for where things stood right before this milestone landed.
• Star or watch the repo — at this pace, v0.1.5 won't be far behind.