Extract

Extract API

Read any URL and get the page back as raw HTML, readability Markdown, or plain text - the read-a-page primitive an agent needs when the answer is behind a link rather than in a search result. Tier-priced by mode: normal and advanced cost 1 credit, ultra costs 2. Only a successful extraction is billed, so a dead link, a bot wall or a timeout costs nothing.

POST/api/v1/extract

Authorizations

Authorizationstringheaderrequired

Bearer authentication header of the form Bearer <token>, where <token> is your Scavio API key (e.g. Bearer sk_live_your_key).

Body

application/json

urlstringrequired

The page to read, 1-2048 characters. http and https only; a bare host like example.com is upgraded to https. Loopback, private, link-local and cloud-metadata hosts are rejected with a 400 before anything is fetched, so they cost nothing.

Example: https://en.wikipedia.org/wiki/Web_scraping

formatenum<string>default:markdown

How to return the page. Whatever you pick, the shape is the same: content is a single string and content_length is its length in characters.

html — The raw page exactly as served, chrome and scripts included.
markdown — A readability extraction: the main content as Markdown, most of the page furniture removed. The default.
text — That same Markdown flattened to plain text. The flattener is deliberately conservative, so snake_case identifiers and inline code survive intact.

modeenum<string>default:normal

Fetch tier, and the price-bearing parameter - this is what makes the call cost 1 credit or 2. Start at normal and step up only when a site refuses it.

normal — Plain datacenter fetch. 1 credit. Right for static pages, docs, articles and most blogs.
advanced — Full headless-browser render, for pages whose content is built by JavaScript. Also 1 credit.
ultra — The hardest-target tier, for pages behind serious anti-bot walls. 2 credits.

Request

from scavio import ScavioClient

client = ScavioClient(api_key="sk_live_your_key")

# extract is a TOP-LEVEL method, not a namespace.
page = client.extract("https://en.wikipedia.org/wiki/Web_scraping")
print(page["data"]["content"])

Response

"url": "https://en.wikipedia.org/wiki/Web_scraping",

"format": "markdown",

"mode": "normal",

"content": "- Meta: charset: UTF-8 - Title: Web scraping - Wikipedia ... [Web pages](https://en.wikipedia.org/wiki/Web_page "Web page") are built using text-based [markup languages](https://en.wikipedia.org/wiki/Markup_languages "Markup languages") ( [HTML](https://en.wikipedia.org/wiki/HTML "HTML") and [XHTML](https://en.wikipedia.org/wiki/XHTML "XHTML")), and frequently contain a wealth of data in text form. However, most web pages are designed for human [end-users](https://en.wikipedia.org/wiki/End-user_(computer_science) "End-user (computer science)") and not for ease of automated u...",

"content_length": 67988

"response_time": 2185,

"credits_used": 1,

"credits_remaining": 4820

}

from scavio import ScavioClient client = ScavioClient(api_key="sk_live_your_key") # extract is a TOP-LEVEL method, not a namespace. page = client.extract("https://en.wikipedia.org/wiki/Web_scraping") print(page["data"]["content"])