Closed sandeepvootoori closed 1 year ago
This will also probably be very helpful for people who want to use GPT 4 for planner and use GPT 3.5 for functions. They can just instantiate the kernel with GPT 4 and setup all functions with 3.5
Looping in @markwallace-microsoft who is looking at this now :)
This work is being tracked via this issue: https://github.com/microsoft/semantic-kernel/issues/3041
I've added this issue as a related issue so will close this now.
Please allow passing of model/deployment name through config.json of the skills. The reason for this is it will make it easy for development and make it easy to use different models based on needs of a function. If the model is not present in the config, kernel should execute the function using kernel model instance.