You can find a good Searxng instance or run your own (it can also run on a RPi)

Yes, this is actually very similar to the setup I've had for years - just a RPi to do the STT+TTS part, and hooks in the background to control the right services.

KOReader

OpenWakeWord (the solution I also use in my Platypush setup) is actually very good at hotword detection, especially when paired with libspeex for noise suppression.

The STT models from Vosk are actually quite good too. The 180 MB English model is light enough to run on a RPi and it's actually quite decent with clean accents. They also have a 1.8 GB model that is very accurate and that actually can still run fine on a RPi5 8 GB.

It doesn't fully work with Lemmy yet, but it should work from a Mastodon/Pleroma account.

Full federation with Lemmy is still an open point because it requires me to properly implement a Group actor for full threadiverse compatibility. From what I've seen so far even Akkoma hasn't implemented it properly.

The closest thing I know (but it only works in the browser) is the Firenvim extension

Anything that runs Android should run KoReader too - which IMO is still the best way to consume Wallabag content.

I have recently purchased a BOOX tablet with Android 14 and it works great with KoReader. But even my old Mobiscribe 1st gen with Android 7 was doing its job - it was too old for most of the recent Android apps, but KoReader works fine even on ancient Android versions.

At the current state it's only single-user, but I'm working on a multi-user implementation. It should be available as soon as I have a proper implementation for the ActivityPub Group actor (similar to what WriteFreely already does, where you have a Group @blog actor that aggregates post from multiple Person actors). That would also unlock full compatibility with Lemmy/NodeBB and other threadiverse implementations. But it takes a while to get done right - and a multi-user set up is also a design challenge when your selling point is to have a blog running on a single folder with Markdown files.

About a free instance, it's not currently available (you'll need your own domain and host and then you can just spin up a Docker container), but AFAIK YunoHost is considering adding it to their offering.

Nextcloud is my best pick. Of course, it does much more than calendars, so if you're looking for a calendar service that does only that it may be a bit overblown.

Big pros:

• Native WebDAV integration out-of-the-box - which means that you can easily export your calendars through WebDAV URLs in any supported app

• Support for importing external calendars

• If you also use other Nextcloud products (like mail, contacts or office) the integration betweren them is quite well designed

On Android then I use DAVx5 to synchronize calendars and contacts from Nextcloud with my phone, and Etar as a multi-provider calendar app.

I'll test it just in case by giving @fabio@manganiello.blog a shout

--EDIT--

Interesting, this actually looks like a bug in my implementation. It's not about Lemmy or Mastodon (at least in the case of receiving loose mentions), it's about the REPLY type that was ignored. Should be fixed now 🙂

There's an example with LaTeX here https://blog.platypush.tech/...

blog:http content is usually dynamically managed via asynchronous requests.

Your best bet is to use the Network tab and filter the actual request made by the browser when it tries to expand the resource.

Initially Webmentions and AP reactions were rendered on separate panels (with the Jinja rendering templates provided by the webmentions and pubby libraries respectively), but then I found that a bit confusing and cluttering, because there were basically two comment sections with two different time-sorted lists of threads. But if people prefer the other way around I could also add it as an option or a filter toggle 🙂

This isn't the right way forward. You don't oppose a thug that breaks sensible regulation by breaking that regulation too.

That's right, there's no pagination on the outbox right now - it's just a few text files anyway so I didn't bother to implement it yet. But it's on my (very short term) radar. The RSS/Atom feeds are already provided over /feed.rss and /feed.atom routes respectively though. About embedding the whole object vs. a reference, that can be configured via activitypub_object_type. It defaults to Note, so the whole content is included, but it can also be set to Article - in that case only a reference to the object is rendered, similar to what WriteFreely does in its latest implementation.