I want to introduce you all to creating, running, and refactoring python notebooks with the Noteable ChatGPT plugin.
We’ve got much more to do very soon. For now, here’s a very tired me from the plugin night deadline (two days for us to get this going!)
We’re quickly learning a lot about what GPT-4 “thinks” of our API. I’ll love to share some of that insight in a new thread.
Some planned upcoming work:
- Creating a default project if the user doesn’t say what project they want to work in. We’d expose changing this to ChatGPT as well.
- Split notebook creation from kernel creation. We assumed we should wait for the kernel to come up before we return. ChatGPT takes long enough putting in IDs that we don’t have to wait for the kernel come up. The model is still typing while we’re waiting for the kernel. It’s better to let ChatGPT do its chatterboxy thing while we provision resources and create documents.
- Process minimal output. Right now our payloads are too much for GPT 4 to both handle and recall the rest of the session. We need some
reprs intended for llms. I started doing that in
genaiand I’d like to make it broader across the Scientific Computing ecosystem.
Set up an account on app.noteable.io and let me know what you think. We’d love to have users help us with the roadmap!