Snapshots

snapshot() gives you a cleaned HTML view of the current page. It is designed for exploration, not for pixel-perfect rendering.

Two modes

Mode	Use it for
action	Clicking, hovering, typing, and scrolling
extraction	Reading page content and building extraction schemas

SDK:

const actionHtml = await opensteer.snapshot("action");
const extractionHtml = await opensteer.snapshot("extraction");

CLI:

opensteer snapshot action --workspace demo
opensteer snapshot extraction --workspace demo

What the output looks like

Each visible target is annotated with a c counter:

<main c="1">
  <h1 c="2">Example Domain</h1>
  <p c="3">This domain is for use in illustrative examples.</p>
  <a c="4" href="https://www.iana.org/domains/example">More information...</a>
</main>

You use those counters for SDK element targets and CLI positional element arguments:

await opensteer.click({ element: 4 });

opensteer click 4 --workspace demo

Important rule

Counters are only valid for the current page state.

Always re-snapshot after:

• navigation
• opening a new page
• form submission
• any action that changes the DOM enough to move elements around

What snapshot() returns

The high-level SDK returns the HTML string only.

It does not return the full internal snapshot object. If you need a visual image, use computerExecute({ action: { type: "screenshot" } }).