Core Concepts

Recordings

Short product videos (e.g., Loom) that demonstrate the desired user flow and success criteria. Fume parses these to infer intents, UI structure, and assertions.

Add a short clip showing a typical recording and the detected intents overlay.

Test Generation

Fume converts recordings into Playwright tests by combining:

• Semantic understanding of the video narration and on‑screen actions
• Runtime probing of your app to resolve stable selectors and states
• Best‑practice patterns for waiting, retries, and assertions

Output is readable Playwright code you own and can version control.

Runners

You can execute tests:

• In Fume Cloud (managed, zero setup)
• In your CI (GitHub Actions, GitLab, custom runners)

Provide a matrix of supported environments and Node/Playwright versions.

Self‑healing

When UIs change, Fume repairs test code by:

1. Falling back to a browser‑use agent to complete the flow during a run
2. Updating selectors and steps post‑run and opening a change review

This keeps suites healthy without manual babysitting.

Ownership and portability

Generated tests are plain Playwright. You can run, edit, and keep them even if you stop using Fume.