If you're doing this, it's also valuable to perhaps set up some groups data :)

I suppose this is now reviewable. Guide below.

How to test

First run pip install -r requirements.txt because this adds random data package mimesis as a dependency.

1. Run the new simulate_matrix Django command to run a simulation (e.g. DEBUG=1 ./manage.py simulate_matrix --start 2022-03-02 --end 2022-05-15 --seed xyz --n-clusters 20) and see its output
2. Run the server with DEMO=1 (e.g. DEMO=1 ./bin/start) and log in for the full demo experience

Approach

What happens when a user enters the demo environment?

Whenever a user signs up/logs in (they are the same here), they are:

• either logged into the existing account matching the email – if they've used the demo before
• or a Hedgebox simulation is first ran, its results are saved, and only then the user is plopped into PostHog – if it's their first time with the demo

How does the simulation work?

Each individual simulation is a matrix (abstractly, a Matrix instance, and concretely in our case, a Matrix subclass HedgeboxMatrix instance). A matrix consists of clusters (Clusters), which in our case (HedgeboxCluster) are companies. And a cluster effectively is a grid of people (SimPersons). Those people are simulated session-by-session synchronously in an outwards spiral pattern. This whole architecture allows for pretty efficient modeling of relations inside groups (for Group Analytics), such as people recommending the product to each other, or invitations. Currently only browser clients can be simulated.

A Matrix also has a set_project_up() method, which instruments the whole project with insights, dashboards, cohorts, etc.

The results of a Matrix are saved using MatrixManager, which does not know how the simulation works, only how to save it.

TODOs

This system has MUCH more potential than the previous one, but it's not as robust as it could yet. Here are some enhancement opportunities:

• make agents invite each other within a company, with $groupidentify
• make agents upgrade/downgrade plan for their org
• add some more insights
• add an annotation or two
• fix events showing up as stale
• add another dashboard
• add recently viewed insight entries
• document simulation approach more in code

Oh right, I didn't explicitly point that out, but the manage.py command is just a dry run. This isn't hooked up to the setup_dev command yet, so for the data to be ingested you need to use option 2 @mariusandra:

Run the server with DEMO=1 (e.g. DEMO=1 ./bin/start) and log in for the full demo experience

Oh, apologies, that's on me. I tried that earlier, but it didn't work as I had too much nothing:

Now I do see this:

.. but no data other than my own local plugins and clicking around. I'm not even sure what project things should appear under.

So, for the actual demo experience, you need to log out and sign in with a different email. That should set you up with fresh data.

Hmm, this should take a few seconds since the events are now ingested async via Kafka, but I'll look into it. 👀 There's a couple of approaches where this could be pre-generated, though I think it's more fun if everyone gets a random environment of their own – provided of course this takes like less than a 10 seconds, where the wait would be OK (we could just show an approximate "Preparing the world" progress bar).