you can try out both tools yourself using either the first 2 models in the settings, which use my api-keys under the hood (via a cloudflare worker) or provide your own keys,.. keys are not stored in anyway..only passed through to the AI api's via that custom cloudflare worker, the code of the worker can be found here : github.com/devhelpr/prompttoform-worker

Cool! I haven't been following LLM APIs recently so maybe you'll know... Is there a way to make them ingest/produce embeddings for graph vertices or edges? IIRC just before LLM explosion transformers were making some noise because of their ability to process graph-shaped data natively.

I'm asking because I see them being used to process tons of JSON data and it feels a little weird to make graphs go through text representation (maybe it helps with something but also seems pretty inefficient)

You can create graphs with llm's, thats what I do using that small generator.. but the layout is done without llms in that generator .. I am experimenting with different libraries like cytoscape and d3 forced layout.

do you have an example of what you mean exactly?

It's called graphical neutral network. They attach feature vectors to vertices and/or edges and can be used to build all sorts of graph processing primitives. They can use transformers under the hood so I guess that somebody already had to merge them with LLMs. Mixed modality text/graph data is possible. I was just wondering if there are any APIs that allow graphs as input or output (similar to how some LLMs allow audio or images).

I dont know.. and do you mean graph neural networks?

Ah yes

Thanks for bringing this up! I am going to dive more into GNNs , it sounds like a powerful combination with LLMs

Absolutely smitten! Thanks for sharing this!

I just made a tiny pull request to tweety to make the font size year configurable.

Also, curious if you’ve tried TabFS!?

Yeah, I love the idea of mapping abstract concepts to the filesystem and managing them using standard unix commands. It was one of the inspirations behind one of my other project: smallweb.run

I maybe went overboard @Achille Lacoin — I’ve added some very basic optional auth to tweety to make it a tad be more usable/safe.

Here’s the commit on my fork — if it is your jam I can create a PR for you.

This is using the browser in the same way that many people, including myself, use Emacs: as a flexible user interface to lots of other programs. A thin client.

In the abstract, I really like this. In practice, I find it too risky on today's browser oligarchy. In particular for a terminal. I see the terminal as the lowest-level UI on my computer. When everything else fails, for whatever reason, I still have the terminal. Making the browser a dependency of the terminal feels very wrong.

Yes, I know, using a Web terminal doesn't take the native terminal away. But it still feels as if an earthquake had messed up the software stack.

That said, I will definitely test this and I may end up using it. Not to replace the terminal, nor Emacs (my default command line interface), but as a way to integrate command line tools into browser workflows.

@Eli Mellen I added basic auth support in the latest release !

I'm also trying to rework Tweety as a chrome extension. It would remove the requirement to register tweety as a service, and add support for some interesting features (ex: managing tabs from the tweety cli)

programmingsimplicity.substack.com/p/one-paradigm-to-rule-them-all?r=1egdky

Ooh! I've been trying and failing to get a chrome extension going for a while...I could see something like this being really useful as kind of like a REPL for in-development plugins.

I've found it really tough to get into a flow where every minor change you need to re-package and then re-import the plugin, just to find you called the wrong method didn't request permissions for a certain thing...

Yeah definitely!

I can also see it becoming an alternative for userscripts.

I added history search to the mix:

I’m not sure I know exactly what you mean by user scripts?

But also the history search is very cool

Have you also tried interacting with chrome’s built in LLM?

I mean stuffs like violentmonkey.github.io

ahh i see

Does "chrome extension" mean that it doesn't work with Firefox any more?

It will also support firefox, I just started with chrome

I'm experimention with adding a way to trigger scripts from chrome context menus !

Another interesting api: chrome.notifications