@mayurajs/cloudflare-quick-action

Cloudflare Browser Run's quick actions for Mayura — one call renders a page in Cloudflare's browser and gives back what was asked for — as a client and as tools, from any runtime with no dependency.

npm install mayura @mayurajs/cloudflare-quick-action

import { cloudflareQuickActions, quickActionTools } from '@mayurajs/cloudflare-quick-action';

const actions = cloudflareQuickActions({ accountId, apiToken });
const markdown = await actions.markdown({ url: 'https://example.com/' });

const tools = quickActionTools(actions, { origins: ['https://docs.example.com'], screenshot: true });
// cloudflare.markdown, cloudflare.links, cloudflare.scrape and cloudflare.screenshot, permission web:cloudflare:read.

apiToken needs the Browser Rendering - Edit permission.

The client

cloudflareQuickActions({ accountId, apiToken }) has markdown, content (the rendered HTML), links, scrape (elements by CSS selector), screenshot (PNG), pdf, json (structured data by Cloudflare's model, billed as Workers AI), and crawl.start, crawl.status and crawl.cancel. Each takes the page's url, and optionally allowRequestPattern: regular expressions, outside which the page loads nothing. A failure names only its HTTP status, never what Cloudflare wrote.

Tools

| Tool | Permission | Option | | --- | --- | --- | | <name>.markdown, <name>.links, <name>.scrape | web:<name>:read | always | | <name>.screenshot: the page as an image | web:<name>:read | screenshot: true | | <name>.json: structured data, by prompt and optional JSON Schema | web:<name>:extract | json: true | | <name>.crawl: up to maxCrawlPages pages of a site, as markdown | web:<name>:crawl | crawl: true |

• origins is required: the sites the tools may read, as createBrowsers takes them, or 'all'. A URL outside them is refused, and comes back as the tool's result so the model can choose another. The page Cloudflare renders is also kept to them: it loads no scripts, images or frames from elsewhere (allowRequestPattern).
• A crawl stays on its site, polls until done, keeps only pages read within origins, and is cancelled when its call is or after crawlTimeoutMs (5 minutes), so it stops spending.
• Content is cut at maxPageBytes (64 KiB). costMicros charges each tool's calls to the run's budget; Cloudflare bills browser time on its side (its free plan allows 10 browser-minutes a day, and REST calls slowly).
• Page content is the tools' result for the model to read: treat it as untrusted text.
• Options: name (cloudflare), origins, screenshot, json, crawl, maxPageBytes, maxCrawlPages (10, at most 100), crawlTimeoutMs, costMicros.

For a browser to drive, see @mayurajs/browser-cloudflare. Apache-2.0.