andrewvy / chromesmith Goto Github PK

This is useful for screenshot use-cases, where we want to load a page but only take a screenshot until all resources have finished loading.

This can be achieved by listening to all network events for a page to know when requests have settled.

There are a couple meanings to pooling, there's pooling of separate chrome processes, as well as pooling of targets within a single headless chrome process.

Some questions:

• Is there a benefit over single vs. multiple headless chrome instances?
• And off that fact, aren't pages/targets sandboxed in their own OS process?

While low-level API provided by CDP is available throughchrome-remote-interface, it may be nice to develop a higher-level DSL for common workflows. (Clicking elements, etc.)

Right now, each chrome instance will start its remote debugger port on 9222 + index. This will cause problems if there are multiple Chromesmith pools.

Looks like we will have to implement some sort of random port finder.

I think it would be very nice to be able to get the resource usage of the chrome instances. If we implement an interface for this in chrome-launcher, we will have to add support for our pooling system to be able to manage multiple instances and grab their metrics in a timely way.

Chromesmith should utilize user profiles to sandbox/clean-up between pages and sessions.

One of the goals of this library is to also expose GenStage functionality, batteries-included to load balance across the pool of processes + pages. This functionality is not required and is one of the tools available for developers to build their workflows.