Browse Code
ai-wright adds AI steps to any Playwright test. ai.act carries out UI actions from a plain-language objective, ai.verify makes vision-driven assertions that fail the step when the model is not confident enough, and ai.extract pulls structured data from the page. Screenshots are annotated with Set-of-Marks overlays and combined with DOM element maps, so the model can work through complex UIs. Keep most of each test as deterministic Playwright code and use AI only on the brittle steps.
It is for teams writing Playwright end-to-end tests who want AI help on dynamic interfaces such as chat widgets, canvas dashboards and multi-step wizards, without vendor lock-in. Bring your own OpenAI, Google Gemini or Anthropic Claude key, or use a TestChimp key. Install with npm install ai-wright, then import { ai } from 'ai-wright' in your tests. Maintained by TestChimp (https://testchimp.io), the QA platform for coding agents.
Features
- ai.act: run one or more UI actions from a natural-language objective
- ai.verify: vision-driven assertions with a configurable confidence threshold
- ai.extract: return structured data as strings, string arrays, ints or int arrays
- Set-of-Marks screenshot overlays plus DOM element maps for precise targeting
- Pre-action planning, such as dismissing modals before the main step
- Waits for page stability and retries failed commands automatically
- Bring your own LLM: OpenAI, Google Gemini or Anthropic Claude
- Optional TestChimp key so token costs are covered by your TestChimp account
- Pluggable provider interface for adding other LLMs, including local models
- Works inside existing Playwright tests; mix AI steps with regular code
- Open source under the MIT License
