API reference

`BrowserAgent`

Core Playwright wrapper with LLM-driven navigation.

import { BrowserAgent } from "slapify";

const agent = new BrowserAgent({ model: "gpt-4o" });
await agent.launch();

const result = await agent.task("Click the login button and submit the form");
console.log(result);

await agent.close();

Constructor options

Option	Type	Default	Description
`model`	`string`	`"gpt-4o"`	LLM model identifier
`headless`	`boolean`	`true`	Run without visible browser window
`maxSteps`	`number`	`30`	Maximum agent steps per task

Methods

Method	Returns	Description
`launch()`	`Promise<void>`	Start the browser
`task(goal)`	`Promise<TaskResult>`	Execute an autonomous task
`navigate(url)`	`Promise<void>`	Navigate to a URL
`screenshot()`	`Promise<Buffer>`	Capture current page screenshot
`close()`	`Promise<void>`	Close the browser

`runPerfAudit`

Run a Lighthouse-style performance audit against a URL.

import { runPerfAudit, BrowserAgent } from "slapify";

const agent = new BrowserAgent();
await agent.launch();

const result = await runPerfAudit("https://example.com", agent);
console.log(result.scores);
// { performance: 92, accessibility: 87, bestPractices: 100, seo: 83 }

await agent.close();

`AuditResult`

Property	Type	Description
`scores`	`object`	Performance, accessibility, best practices, and SEO scores (0–100)
`metrics`	`object`	LCP, FID, CLS, TTFB, and other Core Web Vitals
`report`	`string`	Full HTML report string

`TaskResult`

Returned by agent.task():

Property	Type	Description
`success`	`boolean`	Whether the goal was achieved
`steps`	`StepRecord[]`	Array of steps taken with screenshots
`elapsedMs`	`number`	Total wall-clock time