Set up simulation of sending events to queue

This is a nice, but slightly different type of project than the usual stuff on PPH. I hope you would like to do this.

I need to set up a prototype for a client in a few days and I need help. The prototype is meant to be a setup of an event stream which receives events (they can be simulated, fake, synthetic) from 1 or more 'black boxes' (assume 1 'producer' for the prototype). These black boxes (event producers) do things like facial recognition, proximity detection, etc. inside retail stores. These black boxes send various events (see attached list) to a queue which feeds them into a recommender system (which, for now, is also a black box).
So events come from a black box and go to a black box (for the prototype). What I need is to simulate the producing black box, the queue and some reports on how it performs. The idea is - and this might be different than you would expect - that I will receive several reports from several tests, showing that a basic set up does not deliver great performance, and then, with some tweaks, it becomes better. So it's not about getting it right the first time, it's about showing improvements.

There might be more follow up projects coming after we do this successfully (for example building the recommender itself), but now the only part that I would need to set up is to create the 3 things described above (an event simulator, sending those events through a queue and 'a consumer' that receives them) for the event types which are attached.

The architecture I propose is attached. You first see a blue architecture that provides an overview, but we now only need to create the 3 orange elements:
1. generate events with the formats as attached (each camera would generate many events per second)
2. send these events to an API
3. the API receives the events in a queue.

DO NOT TAKE THE EVENT DEFINITIONS IN THE ATTACHMENT TOO LITERALLY! It's about setting up the structure. All content (the events) is only fake for now anyway. So do not focus or spend a lot of time on the format of the time field or the format of the storeID... AT THIS MOMENT, ANY FORMAT IS OK. My definitions are just a general guideline to help you get an idea. The point is to have an event generator that spits out many events in some variations and a queue that processes them to a consumer. Later, we might change the event format anyway.

It all must be done in Python. I do not have other preferences. Maybe the queue can be in Faust (Kafka/Zookeeper), but if there is something else you can do faster in this first prototype, I would prefer the speedier solution.

The idea is that you will include minimally 3 hrs to walk me through the code and the tests in (one or more) video calls. I will need to understand what you did. Although I'm pretty technical; I'm not a developer, so you would need to talk me through it step by step. You can include 3 hrs extra on top of your estimated building time.

How much time would this cost you and are you able to do this in the coming days?

If this will be fun, we will make next steps to build the actual recommender, which I hope will be even more fun ;-)

Let me now if you have any questions. I look forward to your reply.

Regards,

Jeroen