Future of the project (come and say hi)

[beyarkay]

So, loadshedding is back. I've got a full time job now, so am not able to regularly update the project with loadshedding information. It was a pretty thankless job at the time, but fun for a while. Also unpaid work is low priority work.

For those who don't know, this project (and all other loadshedding apps) are all updated manually. There are good reasons for this, which I can discuss if people are curious.

I'm happy to keep the code available and assist with people taking over the project, or guide new maintainers in how it works. But I'm no longer able to be the sole maintainer and updater.

Taking over the project would require watching the CPT/Eskom twitter accounts and editing a file when they release new schedules. Hosting the website/API I could probably re-setup without too much hassle.

If any businesses use this project, I'd be happy to maintain it for a fee, provided the information and source code remain FOSS.

It's definitely possible for AI to automate scraping twitter and creating the updates, and I'm pretty experienced so could make this happen, but I'm pretty busy so can't do it any time soon. If someone's looking for a cool project, I'm happy to mentor and provide guidelines/Definitions of Done on this.

[beyarkay]

Pinging potentially interested parties: @stevensricco @shaunkleyn @aidanhorn @declan-fitzpatrick @cliffbattco @gsccb @BernardAPedro @PVBiz @TerblancheS @Juandebruyn22 @ijacobs3 @ijacobs3 @spinza @Ohwelt @morgyorgy @overdrive2max @ZenitH-AT

(If you'd rather be left out of the discussion, please let me know)

[PVBiz]

Would like to understand how the process currently works, assume the same applies how ESP gets their data?

Can maybe assist with updating or looking at keeping it alive possibly

[morgyorgy]

I have extra server capacity and can help with hosting. My home automation rules for solar rely on this service, and they stop working without it—I imagine many others are in the same boat.

[ijacobs3]

I don’t mind assisting monitoring x for changes in loadshedding related matters ( grids , etc)

[aidanhorn]

I think I can fit in my contributions into my workflow. I would like us to focus on efficient project management though. Do you have thoughts on what apps we would use for communication? In my job, I use Google Chat, and I have my own professional Google Workspace account @econometrics.co.za on Business Standard, which lets me create a Google Chat Space, if desired. I also suggest a regular Google Meet meeting, such as once every quarter for half an hour.

Perhaps it would be best to plan out processes that should take place where humans are involved, which would make the pipeline more streamlined. 😊

I have an Ubuntu desktop which is running most of the time, including a 2kWh inverter, but the availability is only about 90% as I operate it from home.

[spinza]

I have moved on to using a plugin in on Home Assistant which is based on ESP. Requires an ESP API key though.

[DealingLace]

Big ups for the update, what a lekker project.

@beyarkay I'm deeply curious as to why it is a manual task, is it purely just to prevent API/web-scraping?

[Picalz]

@DealingLace @beyarkay Hi, I'm also curious. I've toyed around with scraping accessing x Eskom_SA & llama3 briefly and already looks super promising. Not much new data to work with currently - fingers crossed that doesn't change ;).

[DealingLace]

@DealingLace @beyarkay Hi, I'm also curious. I've toyed around with scraping accessing x Eskom_SA & llama3 briefly and already looks super promising. Not much new data to work with currently - fingers crossed that doesn't change ;).

Hey man, that sounds awesome. If possible I'd love to see how I could help! I can dedicate some server resources, but nothing GPU heavy unfortunately. Although, let me know what's required.

[Picalz]

@DealingLace Ahh nice! For now Id like to test more.
Its currently running on llama3 70b, but started with llama3.2 3b and both work. llama3 more consistent, but warranted for the larger more complex model.

Out of interest, this was output earlier this month when suspended:

{
    "status": "suspended",
    "stage": 0,
    "start_time": "2025-03-09T10:00:00",
    "end_time": "None",
    "additional_notes": "Coal operations at Kusile Power Station are at optimal levels",
    "source": "https://x.com/Eskom_SA/status/1898642678570443089",
    "certainty": 100,
    "model": "llama3:latest"
}

I've periodically had it running and logging all results (pulling multiple tweets at a time to test the queries relevancy filtering). In reality, it would be a little more simple by only feeding most recent and flagging change.

It will take some work to integrate with area and zone timetables (all though much of that work has already been done in this project 👍 @beyarkay ). COCT and other independents I haven't dug into yet, but I'm sure the process will be similar.

Recently, I've been looking into other online resources that have popped up over time - focusing on gov and public departments. My only issue is testing and validating. Its difficult to spring into action when loadshedding forces action elsewhere.