Greetings to all one of you.
If you remember, last month I asked that those of you who read the blog, and/or viewed the images to let me know. And yes as expected, bupkiss. Zero, zilcho, nada. Well….
I won’t belabor the point any further…
January was fairly busy.
Single in January
In December I signed up to participate in the Single in January Challenge on my favorite photo website https://www.cameradie.org. The Single in Challenges require the participants to post one photo a day from the same camera/lens competition. I’ve done this twice before (they are typically every 8 months) and at the end of both times, I’ve declared “never again”.
Like many on the site have said, it can be challenging getting a shot every day to post. It certainly was for me.
The last two challenges, I had picked cameras/processes that I don’t normally, which made it even more “challenging”.
This time, I picked the camera/lens I use most often to reduce the stress a bit, the Fuji X-T2 with the much aligned XF16-80 zoom.
I tend to do post processing irregularly, so typically the shots I posted were SOOC. Photo Mate R3 was used to download the jpg files from the camera to my Android tablet. At first I was doing some adjustments, but with the number of images in the “group”, PH slowed to the point of being nearly useless.
Typically at the end of these, I say “never again”, but I know that when the next one is announced, I’ll join in. Certainly the weather will be more receptive.
Artificial Intelligence
We chatted before about my interactions with ChatGPT. Unfortunately these things tend to be a bit of a rabbit hole, and Charly, as I like to call it, was definitely so. In any event Charly was able (sometimes with a a lot of wailing an gnashing of teeth) to help me do some things that it would have taken me many months to accomplish.
We built a python tool to offset the time of a GPX track to coordinate with an actual photo taken during the time covered by the track. This was to assist in GeoReferencing images with where they were taken. I’ve been using digikam to do this in conjunction with my downloaded Google timeline, with only marginal success due to differences
Typically I’ll clean the GPX file up, getting rid of the spurious “returns to home” and increasing in frequency in GPX Track Editor.
The real challenge, was getting the time offset correctly. Typically I neglect to change the time in my camera to suit the locality, and I think Google does some weird things with the time, so this can be difficult.
Using ChatGPT, I have “vibe-coded” a tool the allows the user to select an image from a group and then select the location where it was taken on a map, and then the time sequence is adjusted for the entire track.
Thusfar the tool has proven to be useful.
After having more than one Charly session go sideways, I decided to investigate other AI tools. I’ve looked at Anthropic’s Claude.ai, and Perplexity At first, I challenged each of them to give me a “unbiased” review of strengths and weaknesses of each. That was amusing, and better than I would have expected in terms of being biased toward the reviewers capabilities.
I’m still evaluating, but a couple of things have emerged:
Claude.ai “seems” to be much better at “Vibe Coding” which is much of what I’ve been doing. Charly and I had struggled for a bit assembling a python script to convert a folder into a Subversion Repository. We had stood up a Subversion server in the homelab using a shared volume on the Synology NAS. Linux permissions were a real bugaboo, but we got the server installed. What I wanted next was a script that I could assign to a right-click menu in Windows Explorer to convert an existing folder into a Subversion Repository. After a lot of too-ing and fro-ing, Charly threw it’s arms up and declared the task impossible. For laughs, I fed the info into Claude.ai and it created a functional script in one iteration. That was impressive.
Unfortunately Claude.ai is not so impressive at other things. I was wanting to look at why the comments section of the Wall Street Journal were not viewable in my browsers. We got to the point where the cause seemed to be filtering done on a network level in my Ubiquiti Dream Machine SE, via the Unifi software. Asking specifically where to find certain items in Unifi, Claude sent me to non-existent places in the software. When I requested that it provide help with the latest version of the software, this was the response:
“However, I should clarify – I don’t have the ability to update my own knowledge or references. My training data has a cutoff, and while I try to provide accurate information, UniFi’s interface does change with updates.“
I found that quite curious, and of very limited use when dealing with modern IT infrasctructure.
A gotcha with the free version of Claude.ai is there is a fixed limit of conversation in any given session. Once the limit is hit, no further interaction is possible, and there is no ability to archive the work done to that point. The free version of Charly does not have this brick wall. So, when in a session in Claude, it’s wise to ask where you are in the relation to the conversation limit.
My first real usage of Perplexity was literally last night as I asked it some questions about planning a trip to Japan in the fall. A good dialog was had thusfar, but that is the extent of my testing.
My “Pro” subscription to ChatGPT is to be cancelled next week. At this point, I’m not certain that I’m going to subscribe to one of the others.
Journalling
Last year, I had started using an Android based journal-ling package on my tablet, to document the things that were going on in my life. It felt that the days were slipping by with little to no record of the activity. Several months of entries were made, but it was just not working. The app was tailored more for “feelings” than ease of use, and several times I would make entries that didn’t get saved automatically. The author and I corresponded a bit, and he tweaked some things, but by the time that was done, I had stopped using the app.
A daily Moleskine Diary replaced the Android app, and I have been using that since early January. Moleskine diaries have been my go to for many years. I had kind of gotten away from them when I was working in a secure facility, but coming back to them this year has been like finding an old friend. Warm and comforting.
Home Assistant
HASS and I are becoming less antagonistic nowadays. I’m becoming a little more conversation with the arcane YAML lingo. A developed intruded C.A.F.E. which is a “visual” methodology for creating and editing automations in HASS.
It described as such:
“C.A.F.E. is a visual flow editor that brings Node-RED-style power to Home Assistant without the external engine. It transpiles your visual diagrams into 100% compliant, native Home Assistant logic stored directly in the core system.“
The beauty is it works from native YAML automations and doesn’t reside outside like Node-RED.
When I first saw the announcement, I jumped into it, only to find some pretty spectacular bugs. So, we went back and forth quite a bit, and version 0.4.2 seems to be quite stable and useful. At this point, I’ve only used it to view, and tweak one or two of my existing automations. Like many bleeding edge FOSS developments, documentation is all but non-existent, so I need to dive in with something new and work through the traps.
The other development in the HASS arena is the use of vibe-coding. Twice now, I’ve gotten either Charly, or Claude to write YAML for me that drastically reduced the time necessary to develop automations. I’ve not really tried to create a Dashboard from scratch since the one I created for the audio interface.
Claude was able to help me wrap my head around HASS Templates. The syntax of HASS Templates probably makes sense in an internet development mode, but to me, a self taught procedural programmer, it’s the functional equivalent of speaking Klingon. I’m definitely not conversant, but it makes a little more sense.
Complicating all this is HASS is deprecating previous templating formats for a new format structure. I got this notice that I had 5 templates that in six months would become inoperative. Thank goodness there was plenty of notice, but this also generated plenty of churn as I had not created the templates in question. There were translations of the exiting templates provided, and it turned out to be fairly easy to replace the existing templates to the new format. Kudos HASS foundation on that one.
It’s funny, but having this automation capability has helped on a number of small ways to “ease the burden” if you will.
About 6 months ago, BambuLabs pushed out an update which turned on the chamber fan in the printer at the end of a 3D printing process when using ABS filaments. It was deemed necessary due to the off gassing of the filament during printing, but they neglected to insert a stop, so the fan would just run forever. It was fairly trivial to create an automation in HASS which turned the fan off 5 minutes after the print was completed. The other ways to do this involved updating every ABS printer profile to insert the fan off code. And of course, all this gets overwritten when another update is pushed. So, HASS provided a very elegant, simple solution.
3D Printing
It’s amazing how much usage I’ve gotten out of the BambuLabs P1S printer last year. The status indicated some 450 hours of usage in the last year. That’s incredible. I doubt that I had used the Wanhao Duplicator 6 that much in it’s entire lifetime. I’ve certainly used more filament than ever before.
Being able to just create stuff to solve the little first world problems is really cool. The latest adventure was printing a replacement for the “turbo” button on my Ryobi battery powered leaf blower. The original one broke literally the first time I used the blower when it fell from where it was hung. Without the turbo feature, it was just about useless. After I replaced it, the capability was amazing.
I got a wild hair and decided to see if I could create a two position trigger for the blower, and do away with the separate button. Unfortunately, there is way too much “stuff” in the handle to do this, so I made a lock for the turbo button instead. I’ll keep thinking about the two stage trigger though.
Project Management
Over the years, I’ve kinda just thrown projects together under a shared drive on the NAS. Not much discipline, definitely a mumble jumble. It’s become clear that this was not workable, as I was losing things due not being filed. So, I finally bit the bullet and decided to clean things up.
Years ago, I had used Subversion to manage the work I did for Outland Technology This was used primarily so that Buddy could synchronize his files with the work I was doing for him. It was all hosted on an Infrant ReadyNAS and worked very well.
Doing some looking around for open source revision control, GIT is currently the fav amongst virtually everyone on the innerwebs. Doing some digging, GIT doesn’t work well with large files, and definitely doesn’t work well with binary files, and most importantly GIT pushes you to use GITHUB to archive files.
Not comfortable with any of these shortcomings, I decided to stand up a SVN server in the homelab under Docker Swarm. As discussed earlier, this was not without it’s foibles, but it’s done, and at this juncture, it’s working well. There have been several instances of my developments going sideways, with little ability to return to a state of “sanity”. This provides that capability.
Photography
Looking through the archives, none of the three sets of images for the “Single in” challenges that I’ve participated in have been processed. Yikes. That’s a lot of images. So, that is definitely high on the priority list, and the subsequent creation of albums in Immich.
Japan
The kids and I have decided to take a trip to Japan in November. I have just started researching the trip, and if I’m honest, it’s a bit overwhelming at this juncture. A session in Perplexity was created to help with the thought process, and that’s been useful thus far. I’ll post more as it develops.
Well, I think that’s enough yapping for now. Again, I’ll ask you to drop me a line. Just say HEY!
Cheers
