Our first impressions of Eve, and some advice from building with AI/SDK

Great read, this one and the one about Satoru.

I'm trying to go from support bug report/request to PR with Slack in between for decision-masking -- roughly.

I want it to be able to test in their cloud browser when necessary, take screenshots (you can wire into github pr through slack image uploads apparently), but most importantly have an isolated db for its app server with 1-3 records from prod that it can manipulate. Trying to figure out the right approach with Mongo Atlas.

Enterprise

Yo - you'll want to look into Agent Browser. I'm not sure if it's possible to run it with Eve but we've used this locally to do exactly what you're asking for. We also use it to post videos/images to GitHub PR's.

You may also want to use Vercel Blob storage to avoid blowing up your git repo size or DB (if it's practical or useful). If you end up pursuing this and getting stuck, just shout and we can collab

reply to #2

hey i agree with you

Jonoroboto:

Hey folks how’s it going, me again.

With the release of eve, I thought we’d write up what we’ve learned so far, through AI/SDK and our previous experience of building a bunch of agents. At the moment we’re working on the blog as we’ve launched it, because as you might imagine, we are constantly learning new things about the framework, but I’d say we’ve got a fairly good handle on it.

Without further ado:

https://robotostudio.com/blog/building-agents-on-eve

We will keep updating this as we build new AI examples, so feel free to check it every once in a while. That being said, is anybody building anything with the new framework? It would be great to share notes if you find something interesting.

Also the demo example in the blog above is what keeps our page updated. If you want to see a real world example of where it’s actually done the work itself, check the FAQs for contentful and tell me if you can spot the AI written one…

That's really great, I like it.