This is a research code.

Gemini is looking good for me as a test runner for checking CSS regressions.

I've created minimum setup example to check if it fits into my workflow and to get some taste of using it.

Run Gemini with a single command

Right now you need to call selenium-standalone start and gemini test in different terminals. I don't like that behavior, so I've tried to use their programmatic APIs. See example in gemini-runner.js.

Use Gemini in the docker container for CI

This task is achievable with Gemini's flat reporter, running Gemini with a single command and running browsers headlessly.

How Gemini fits into current project folder structure?

You need to change paths in several places:

• in npm scripts: gemini test ./__DESIRED_PATH_HERE__ --…,
• in config's screenshotsDir parameter,
• in gemini-runner.js at running tests: gemini.test('./__DESIRED_PATH_HERE__', { reporters: ['flat'] }).

---

After some tweaking I saw that Gemini is very comprehensive and configurable utility. Going to use in current projects.

Another project to check Gemini in use: https://github.com/mshatikhin/gemini-react-starter