agentic nostr
Discussion
What am I looking at
Agents on nostr. I gave up on DVMs after about 10 attempts. They are too complex for me to understand. I'm going for something simpler.
The cool part is that any of those agents can also become DVMs, where they can also respond to job requests (Kinds 5000-5999). If it's just text like in the image, you could probably just use Kind 5050.
You may like ezdvm (https://github.com/dtdannen/ezdvm). Super quick to put any python code behind a DVM. Once your DVM is running, you can test it using the playground at https://dvmdash.live/playground/ (need to have a NIP-07 browser extension running). Planning to launch v0.1 on pypi very soon!
With a bit more basic infrastructure DVMs should be much much easier to use
That is very cool, thanks. It might be helpful to have some text to say what kind 5050 is. But I can look it up.
I know. Really, we need wikifreedia pages for each of these. I need to setup a good wikifreedia editor to start pumping these out lol. For now you can see examples by looking at the kind browser here: https://dvmdash.live/kind/ and this site describes some of them https://www.data-vending-machines.org/
Thanks. I had almost given up on DVMs but your work has encouraged me to give it some more time. The main problem is that the whole concept is too BIG. I am sure that once you understand it, it seems very easy. Or if you use a framework. But another way to learn something is just piece by piece. And from that point of view its very hard to find an *easy* jumping in point. Your site has helped on that front. But for someone wanted to know the concepts and how to make one from scratch it's quite opaque. Even should it start on NIP-89 or NIP-90 or which kinds to use, it's all magic numbers that if you know them, you know. But not beginner friendly. I will try again though. I think maybe I can implement NIP-89 for a bot that answers simple questions. I'll read up on the kinds too. tl;dr high learning curve is a barrier to beginners