# Highlights - https://youtu.be/CojlV7-cYzA?si=TqRmls30LOZE2Wv-&t=222 - ![[Pasted image 20240520154714.png]] - Here, "Gossiper" (pink node with a mouth speaking) notifies *Listener*s (mostly ears) - We see that only the cat listener reacts, the others drop the message - This might happen if I say "I noticed the front litter box was used" because that currently does not set a timer (green nodes), such notes are not highly integrated - https://youtu.be/CojlV7-cYzA?si=z_EMk09kHRf6afRb&t=232 - ![[Pasted image 20240520154930.png]] - Here, we have a transcription that - undergoes natural language processing (with Rasa) - undergoes entity extraction (formalized structure, no longer English) - is added to a report that... - ...includes an [[Cat sifting report example|aggregate summary]] - (the food tracking stuff is probably because of the per-minute bucketing, it's not related) - This is a good example of [[Idea Provenance]] - if anything goes wrong, I can audit each step taken # Transcription - \[00:00:00\] My name is Michael Seydel and what you see here is a graph or network that may remind you of a social network visualization. - \[00:00:08\] In this case, you're looking at the neurons and synapses of my voice-driven digital external brain. - \[00:00:15\] I want to briefly demo two use cases. - \[00:00:18\] The first is around rudimentary distress detection. - \[00:00:22\] I'll record a voice note on my phone that will take a moment to sync, transcribe, and then propagate. - \[00:00:30\] "I'm feeling distress." - \[00:00:35\] I haven't optimized for speed, so Syncthing has to notice the new audio file in the background and then sync it before anything interesting happens. - \[00:00:45\] I need to show that... - \[00:00:55\] Alright. - \[00:00:57\] I know that was pretty fast, but basically we saw two things. - \[00:01:02\] The simpler one is that my lights flashed in response to distress detection. - \[00:01:10\] And a push notification is on the way, though it's not visible. - \[00:01:14\] The other is that this visualization that was the digital neurons communicating in real time. - \[00:01:21\] The second use case is around my cat's litter use. - \[00:01:25\] I basically do a bit of natural language processing to build a report that summarizes the litter use for the day... - (Rasa open source) - \[00:01:32\] And let me switch over to today's. - \[00:01:37\] So this is today's litter box report. - \[00:01:41\] We see a summary and then the data that the summary is composed of with links back to the source data. - \[00:01:47\] And if I use Obsidian, I can see all the contents of that source note. - \[00:01:56\] This particular report has a problem, though. - \[00:01:58\] I happen to know that it's missing some data. - \[00:02:01\] If we go over and look at cat transcriptions for the day, we can see all of them because they were filtered. - \[00:02:07\] But we can also see the ones that were not automatically integrated. - \[00:02:11\] And earlier I went and I highlighted this particular use. - \[00:02:16\] That was at 4.01am. - \[00:02:20\] So I'm going to go ahead here and I'm going to update this data. - \[00:02:25\] And now note that I have not fixed the summary, but... - \[00:02:33\] Yeah, here my food reminder went off as well, but you see a distress detection push notification. - (not on the screen 🙄) - \[00:02:39\] And now here I'm going to go ahead and make another note. - \[00:02:42\] I am going to say... - \[00:02:46\] "I just sifted the front litter box and found a pee clump." - \[00:02:55\] So again, it'll take a moment to process everything, but... - \[00:03:01\] When the new data comes in, what we should see is a correct output. - \[00:03:06\] We shouldn't see my lack of updating the summary be a problem. - \[00:03:16\] Alright, there we go. - \[00:03:18\] So we can clearly see that the note that I just made is there. - \[00:03:23\] It has all the details like the other ones do. - \[00:03:27\] And then if we hop back to the graph... - \[00:03:31\] So we didn't see the graph operating in real time, but if I do a refresh... - \[00:03:39\] If I use this slider to go backwards, it's using... - \[00:03:45\] Just the historical data bucketed by the minutes. - (for the day) - \[00:03:48\] I expected some stuff on the right side. - (maybe a bug, e.g. the frames might be reversed 🤷) - \[00:03:52\] Yeah, essentially that is what the graph looks like when one of these litter reports is updated. - \[00:04:00\] There's other stuff going on here that I'll get into more detail later, but I just wanted to share that much for now.