Hey, fedi.

I've got a talk lined up for .

On Sunday, 31 July 2022 at 0200 hours UTC (7:00pm Pacific), I'll be doing a talk on using to parse user input.

The Summer School talk format is 15 minutes presenting/45 for questions, so this'll be a .

More details to come as @SummerSchool posts details.

Please boost?

"To help some of the newcomers make connections: name 5-7 things that interest you but aren't in your profile, as tags so they are searchable. Then boost this post or repeat its instructions so others know to do the same."

Hi, new folks. o/

I guess I should repost my . I recommend that you post one as well, so that we can get to know you.

I'm a weirdo of many trades. My day job involves and by way of . People way smarter and more skilled than I say I'm a , so I guess that's what I am. I love and do . I'm old fashioned enough to still have a blog. I'm not doing too well right now, because reasons, so I don't really have the compute cycles to do much more than hit-and-run posting here. That should change in the weeks to come.

Building out a test environment on Audra to tinker with that AHT-20 sensor.

In reworking some of my bots to use instead of , I'm looking at using a single unlisted room instead of chats to individual accounts (1:1 chat, not MUC).

This changes the command parser substantially. Probably more than I think.

What I'm considering is whether to keep the commands looking like "Bot, do this thing..." or switching to "Do this thing..." and the bots present look at the message, see if it's from their owner (me, in this case), and if so proceed to parse the command. The latter means that failed parses would fall into the same case as mis-parses, and result in a storm of error responses from every bot that doesn't have commands like that.

Of course, bots could easily just not send error responses, but that messes up the "there really is an error" handling case.

I'm also toying with the idea of having both use cases implemented for usability's sake (plus, it adds the "send commands directly to the bot" case).

What do you think, ?

Rhasspy: A personal assistant written in that is designed with self-hosted utility software (speech recognition, speech to text, natural language processing, and speech synthesis) primarily, instead of outsourcing it to untrusted third parties.


I make stuff that crunches numbers. Got tired of red team, now sysadmin/defense.

My imaginary friends are better conversationalists than I am, and are written in and .

Lots of augmentations of the hardware and software kind. Chances are you might have a conversation with one of them and not realize it. About a third of me is prosthetic these days.

I suck as an electrical engineer but sometimes I can patch stuff together.

I like picking locks. Still trying to get good with padlocks.

I assume you've been told about Homestuck.

I like to be well informed. The downside of this is that I don't sleep well anymore.

I love music.

Yes, I use Arch. I don't care what you use as long as it does what you need.

Okay, this is weird...

I have a server written in that can accept HTTP PUT requests on a particular URI:


I also have a client written in Python which sends messages to the server on that URI with PUT requests:


Here's what's messing with me:

On my x86-64 boxen, this works just fine. It does what you'd expect, reliably, on demand. On my Raspbian machines, though, the exact same request doesn't work. The JSON doesn't deserialize.

Has anyone ever seen this before?

Show older

A bunch of technomancers in the fediverse. This arcology is for all who wash up upon it's digital shore.