Roadmap

Skiritai is evolving rapidly. Here's what's on the horizon.

Visual Perception Layer

Current AI exploration relies on DOM analysis and CSS selectors. The next major evolution adds visual perception — the agent will "see" the page like a human tester.

Visual-based AI Exploration

• Screenshot understanding — the agent interprets page screenshots to identify UI elements by their visual appearance, not just DOM structure
• Canvas & WebGL support — interact with interfaces that lack accessible DOM (charts, games, media players)
• Layout-aware navigation — understand spatial relationships between elements for more natural interaction patterns

Multimodal Model Support

• Vision-Language Models — leverage GPT-4o, Claude 3.5 Sonnet, Gemini and other VLMs for richer page understanding
• Native multimodal models — support models with built-in vision capabilities, reducing the need for separate DOM analysis steps
• Model-agnostic perception — the perception layer adapts to whichever model is configured, automatically selecting the best strategy

Visual Regression Detection

• Cross-run screenshot comparison — automatically detect unexpected UI changes between test executions
• Diff highlights — generate visual diffs showing exactly what changed on the page
• Baseline management — maintain approved screenshot baselines and flag deviations

---

Multi-Platform Testing

Skiritai currently supports Web testing via Playwright/Chromium. We plan to extend the same Explore → Replay workflow to additional platforms.

Current: Web

• Playwright-based browser automation
• 14 built-in tools (navigate, click, fill, scroll, etc.)
• Persistent browser sessions via CDP

Planned: Mobile (iOS / Android)

• Appium or browser-use mobile integration
• Same BaseCase API — write test cases once, run on mobile devices
• Touch gesture support (tap, swipe, pinch)
• Real device and emulator/simulator support

Planned: API Testing

• HTTP request tools for the AI agent (GET, POST, PUT, DELETE)
• JSON schema validation and response assertions
• Mix API calls with browser steps in a single test case
• Authentication flow support (OAuth, API keys, JWT)

Under Investigation: Desktop

• Playwright Electron support for Electron apps
• OS-level automation for native desktop applications
• Cross-window interaction patterns

The Vision

The goal is a unified test framework where the same Explore → Replay workflow works across Web, Mobile, and API — write once, test everywhere.

class CrossPlatformCase(BaseCase):
    async def web_login(self):
        await self.ai.action("Login to the web app")

    async def api_check(self):
        await self.ai.action("Verify the user profile API returns correct data")

    async def mobile_notification(self):
        await self.ai.action("Check the push notification appears on mobile")

---

Recently Completed

Feature	Description
Visual Reports	Vue 3 + Ant Design standalone HTML report with screenshots, assertions, and step details
ai.screenshot()	Capture named screenshots during test execution
ai.verify()	Assertion API for natural-language verification checks
@max_steps	Per-step agent recursion limit decorator
Failure Policies	@on_failure(SKIP/RETRY) for resilient multi-step flows