gut now supports Ollaya: typed decisions as one line of Python, on local models

[Kungie]

Hi! Thanks for building Ollaya and greetings from Turkey. Serving TypeSafe's API unchanged made this almost a one-line job.

I maintain gut (https://github.com/Kungie/gut), a small Python library that turns typed decision questions into one readable line with a YES / NO / UNSURE answer. As of 0.7.0 it has first-class Ollaya support:

import gut

gut.configure(backend=gut.JevBackend.ollaya("winnow:e4b"))

if gut.likely(comment, "is spam"):
    hide(comment)

team = gut.classify(ticket, Team)     # an Enum becomes a choice question

It reads OLLAYA_HOST and OLLAYA_API_KEY, so it finds the server the same way the Ollaya CLI does, and it counts local calls as free in its cost tracking. The command line and MCP server take it too:

GUT_BACKEND=ollaya GUT_MODEL=laya gut filter "is spam" comments.txt

The combination I think fits Ollaya best is a cascade: a fast local model settles what it's confident about, and only the unsure answers go on to hosted Jev:

gut.configure(backend=gut.Cascade(
    gut.JevBackend.ollaya("laya"),
    gut.JevBackend(),
))

Docs: https://kungie.github.io/gut/docs/backends.html#on-your-own-machine-through-ollaya

If you'd find it useful, I'm happy to open a PR adding a short "Libraries that work with Ollaya" note to the TypeSafe compatibility docs, but only if you want one. Feedback welcome, especially if I've misunderstood anything about how Ollaya should be called.