You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As far as I can tell, uniqueItems (currently used by Inscryption) isn't widely supported by the available structured output grammar engine thingies:
guidance has an image with a feature matrix for what I assume are the most common solutions (I'm not really in the business of picking structured output libraries)
llama.cpp gives some reasoning as to why they don't support it
The spec doesn't mention the keyword as forbidden. Does Neuro support this?
The text was updated successfully, but these errors were encountered:
Neuro gets told your specification, along with the command name and the command's description. The returned response from Neuro can be literally anything, including invalid json:
The data parameter comes directly from Neuro, so there is a chance it might be malformed, contain invalid JSON, or not match the provided schema exactly.
You are responsible for validating the JSON and returning an unsuccessful action result if it is invalid.
So really, your question could be re-phrased as "Does Neuro understand uniqueItems", and the answer to that is who really knows what a large language model "knows" at all, but for all intents and purposes don't count on it to be correct.
As far as I can tell,
uniqueItems
(currently used by Inscryption) isn't widely supported by the available structured output grammar engine thingies:The spec doesn't mention the keyword as forbidden. Does Neuro support this?
The text was updated successfully, but these errors were encountered: