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
{{ message }}
Repository navigation
Support OpenAI Decisions API through the classifier interface
#3166
OpenAI released the Decisions API in public beta on October 6. It returns typed answers for classification, choice, and scoring tasks.
With the classifier interface and Jev support introduced in #2758, would it make sense to add an OpenAI adapter under agentscope.classifier? This would let users with OpenAI credentials use the same interface for classification and model routing.
I propose keeping the contribution focused on:
An adapter based on ClassifierModelBase, reusing OpenAICredential.
Mapping supported questions and answers to the existing classifier types.
Tests, a usage example, and companion documentation.
A couple of API differences may need discussion first, particularly refusal responses and image inputs, which the current classifier types may not fully represent.
Is this already planned by the core team, and would a community contribution be welcome? I’d be happy to work on it once the scope and direction are agreed.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
OpenAI released the Decisions API in public beta on October 6. It returns typed answers for classification, choice, and scoring tasks.
With the classifier interface and Jev support introduced in #2758, would it make sense to add an OpenAI adapter under
agentscope.classifier? This would let users with OpenAI credentials use the same interface for classification and model routing.I propose keeping the contribution focused on:
ClassifierModelBase, reusingOpenAICredential.A couple of API differences may need discussion first, particularly refusal responses and image inputs, which the current classifier types may not fully represent.
Is this already planned by the core team, and would a community contribution be welcome? I’d be happy to work on it once the scope and direction are agreed.
All reactions