A voice assistant comes to mind. Ideally, it'd be local, but if you don't have the hardware you'll go with the cloud, in which case, the fastest, the better.
A Jellyfin or Plex server can be had for real cheap using used hardware. A Ryzen 3 build can be found for next to nothing here in Brazil[1], so I imagine it’d be even cheaper in the US.
Add an old Quadro card for hardware decoding, or go with an Intel CPU for Quick Sync, throw some IronWolf drives inside, install your favorite Linux distro, and you’re off to the races.
Yes, managing a server is more work than just signing up for Netflix or whatnot, but it’s definitely worth the effort.
[1]: A quick search shows me a Ryzen 3 3200G build with 16 GB of RAM for $200, and electronics are super expensive in Brazil.
Personally, I got my first invite by signing up for a seedbox accepted by the tracker. Then I got invites to other trackers from the same group by being a good seeder.
While waiting for another developer to finish something for a project I’m working on, I’m working on an MCP server in Go that provides a Home Assistant voice assistant with basic functionality that is missing from the default feature set, like telling jokes, flipping a coin, choosing between N options, making equations, and other things I consider essential for a voice assistant.
I’ll probably go further and include things like Apple Weather support[1], and Plex, YouTube, and Kagi search, if I have the time.
[1]: Because in Home Assistant you can only ask about the weather in the location you have configured, not in other locations. At least that’s how it works here at home.
Not seeing the refusals everyone is talking about, but I’ve only spent a few hours with it so far.
Had it review a password generator library I wrote to see if the passwords have biases and review how cryptographically secure the code is and had it review a registration/login flow for security issues, as two security examples, and it did just that.
Overall, I like the model so far, but not enough to pay past my subscription to keep it. Once it’s out of the subscription, I’m done with it.
I pay about $80 for a 1Gbps symmetrical connection in Brazil.
I work in photography in my spare time, and a single photoshoot can easily pass 2k raw photos, which I backup to three remote locations. Plus, I download every movie and TV show we consume at home, have games to play, and work as a programmer.
Honestly, if I could afford a 10 gig switch and routers (we have three at home, in mesh), I’d go higher.
Quality of life matters. If I can make a website or a download just a little bit faster, I’ll.
I self-host most of the things my family and I use on servers at home, and I document everything. Hopefully, the documentation will help.
I have a “death instruction manual” document with all the data and information my wife will need to access my digital life, including the documentation mentioned above, and contact information for people I trust to help her through the process of either managing those servers, or selling them for a good amount of money (there is about 90TB of storage in those).
I also try to teach her, and in time my son (he is too young right now), to manage those servers, but that is more of a long-term goal. She’s comfortable with Linux, so that’s a start.
This seems pretty awesome; being able to use an 8B model for tool calling would be perfect.
Interested in using this for Home Assistant using a Mac Mini as my server. Does it run on MacOS?
How is the latency when using the proxy? I’m using Claude Haiku 4.5 for my voice assistant right now and it’s pretty fast, but if I could keep the LLM local, it’d be even better.
I haven’t created a ticket for OpenLibrary yet, but it’s on my mental todo list. I’ll create the ticket for multiple new extractors today.
I’d love to help improve other services. I plan on charging for Librario at some point, but I’ll offer a free version and offer free API keys for projects like Calibre and others.
Right now the meeting happens on the fly and then is cached. In the future I imagine the finished merge will be saved as JSON to the database, depending on which is more expensive, the merging or a database call.
Merging on the fly kinda works for the future too, for when data change or for when the merging process changes.
No idea what the future will hold. The idea is to pre-warm the database after the schema has been refactored, and once we have thousands of books from that, I’ll know for sure what to do next.
TLDR, there is a lot of “think and learn” as I go here, haha.
No, I decided pretty early on to make it database specific instead of more generic, so we do use some PostgreSQL features right now, like their UUIDv7 generation.
But once the database refactor is done, I wouldn’t say no to a patch that made the service database agnostic.