PyGPT
A desktop AI assistant with an inline Jev System One plugin.
Added to Jevfast
How it uses Jev
PyGPT exposes an inline jev_evaluate(state, questions) tool. Its plugin documents a support-refund classification request with a Choice over refund and other, and a separate Noul damaged-item example; the current assistant supplies the actual state and typed questions.
What Jev decides
Documented refund-intent exampleExample answers · not a recorded Jev response · Source ↗
Question 1 · intent
YOUR APP
INSTRUCTION
Classify support intent.
STATE
The documented example state is {message: “Refund never arrived.”}; it is an illustrative example, not a recorded live model result.
JEV · CHOICE
- refund
- other
App workflow
- Enable the inline plugin
A PyGPT user configures the TypeSafe key and Jev model for the Jev / System One plugin.
- Issue a bounded decision
The current assistant invokes jev_evaluate with structured state and named Noul, Choice or Score questions; the plugin validates and posts the request.
- Use the typed result
The plugin returns raw Jev JSON to the current model or surrounding tool workflow for classification, routing or verification.
Why it is interesting
A desktop AI assistant can call Jev as a bounded judgment tool during an otherwise generative workflow, retaining raw typed answers for routing and verification.