rejot-dev/fragno/web

Web

The web provider reads a page with web.extract or saves the file behind a URL with web.download. Read "/static/codemode/providers/web.d.ts" first, then follow the branch you need.

Read a page

1. Select the source: a url supplied by the user or provided html. Complete when the exact

source and supported input fields are known.

1. Choose the output shape:

• content for rendered HTML;
• markdown for readable page text.

Complete when one action matches how the result will be used.

1. Call web.extract:

``js async () => await web.extract({ action: "markdown", input: { url: "https://example.com" }, }); ``

Inspect the result branch selected by action. Complete when extraction succeeds and the returned content has been checked.

Download a file

1. Choose the path under /workspace. Its extension sets the file's type, so take the name from the

integration's metadata or the URL, such as /workspace/downloads/logo.png. Complete when the path ends in the extension of the file's real type.

1. Call web.download:

``js async () => await web.download({ url: link, path: "/workspace/downloads/logo.png", }); ``

Complete when it returns the path and sizeBytes of the saved file.

1. Recover from a failure by its code:

• WEB_DOWNLOAD_FAILED carries the server's answer. A temporary link that answers 403 or 404 has

usually expired: get a fresh one from its source and download again.

• WEB_DOWNLOAD_URL_REFUSED: ask for a public https link.
• WEB_DOWNLOAD_TOO_LARGE: the file exceeds what a workspace file holds. Tell the user, with the

link to open it directly.

Complete when the file is saved or the user knows why it cannot be.