Welcome to stuntd discussions

[bladedevoff]

Hey! I'm the maintainer of stuntd, a proxy that learns your LLM's typed decisions and answers the confident ones locally with a small head.

Where to post:

• Q&A for questions about setup, training or numbers you don't understand
• Show and tell if you run it on something real, even a toy project
• Ideas for things that aren't ready to be an issue yet
• Issues for bugs, with the X-Stuntd header and the stuntd version if you can

0.1.2 is on the way: decisions with several fields, the Anthropic Messages API, retraining in the background and lazy checkpoint load. I'll post here when it's out.