Post

A new AI step is available in Airglow ✨ It lets your automations call @cocore.dev: you pick the model, write the prompt, choose the input data, and the answer flows into the next steps

23/9/26, 3:48 pm6 reposts19 likes

Hugo

@there-hugo.dev

You bring your own co/core key and save it in your Airglow settings. It's stored encrypted and only used for your own AI steps. No account yet? Sign in on cocore.dev with your atproto account and get 1M tokens for free!

Hugo

@there-hugo.dev

See it in action here 👇 airglow.run/u/bot.airglo...

airglow.runBriefly describe new automations created on Airglow | @bot.airglow.run | AirglowWhen a new automation is created on Airglow, ask a model on co/core (https://cocore.dev) to briefly describe the automation, and post the description to Bluesky.

Hugo

@there-hugo.dev

It's still in beta, so I'll keep an eye on what you build with it and keep iterating. Next up might be structured output data and a Jev-like classifier feature 👀

@byjp.me

👀 @cocore.dev are you considering supporting Jev-like classifiers on the models that the co/core client loads?

@byjp.me

This is so cool! Thank you! I’m thinking of implementing co/core in some of my atproto apps, is there a reason you chose the API key over writing the lexicon for inference directly? (It’s likely a bunch less work, and allows other open-ai compatible LLMs to hook in, but just checking!)