Skip to content

Basics demos

Five minimal tasks that teach the core building blocks: navigate, wait, extract, screenshot, and inline JavaScript.

title

The smallest possible task — open example.com and return the page title.

curl -s -X POST localhost:8765/tasks/basics/title -d '{}'

task "basics/title"
    pool default
    timeout 15000
    transport rest

    status "Visiting example.com"
    goto "https://example.com"

    status "Reading title and heading"
    extract page from "html"
        title   text "title"
        heading text "h1"
        body    text "p"
    end
end

{
  "ok": true,
  "data": {
    "page": {
      "title": "Example Domain",
      "heading": "Example Domain",
      "body": "This domain is for use in documentation examples …"
    }
  }
}

Concepts: goto, extract, named results, status lines.

screenshot

Capture a viewport PNG of any URL. Returns base64 under data.png_b64.

curl -s -X POST localhost:8765/tasks/basics/screenshot -d '{}'
curl -s -X POST localhost:8765/tasks/basics/screenshot -d '{"url":"https://news.ycombinator.com"}'

task "basics/screenshot"
    pool default
    timeout 15000
    transport rest
    input url string default "https://example.com"

    goto "{{url}}"
    wait until "body" timeout 10000
    screenshot png_b64 selector "."
end

curl -s -X POST localhost:8765/tasks/basics/screenshot \
  -H 'Content-Type: application/json' \
  -d '{"url":"https://example.com"}' \
  | python3 -c '
import json, sys, base64
b = json.load(sys.stdin)["data"]["png_b64"]
open("/tmp/shot.png", "wb").write(base64.b64decode(b))
print("saved /tmp/shot.png")
'

Concepts: input schema, {{url}} templating, screenshot action.

inline-js

Run arbitrary JavaScript inline with templated arguments.

curl -s -X POST localhost:8765/tasks/basics/inline-js -d '{}'

The task uses an inline js step to read document.title and return structured data without a separate .js file.

Concepts: js, inline scripts, returning values from JS.

See also: JS modules for reusable scripts in scripts/.

save-html

Dump the fully rendered HTML to a server-side path.

curl -s -X POST localhost:8765/tasks/basics/save-html -d '{}'

Concepts: writing files, server-side output, rendered DOM vs source.

wait-then-click

Navigate, wait for an element, click it, then extract the result.

curl -s -X POST localhost:8765/tasks/basics/wait-then-click -d '{}'

goto "https://example.com"
wait until "button.submit" timeout 10000
click "button.submit"
extract result from "html"
    message text ".result"
end

Concepts: wait until, click, chaining steps.

What's next?