Few-shot examples with energy conservation are used in prompts when chatting ungrounded with the models

[nhwkuhns]

Bug Details

When chatting ungrounded, directly with the model, it will occasionally give responses based on the few-shot examples used for the grounded chats.

Steps To Reproduce

1. Start an ungrounded chat
2. Send the prompt "test test"

What is the expected behavior?
The model would typically reply with something like "it appears you are trying to test something. If you give me more context I can help"

Screenshots

Information Assistant details

Please provide the following details. You can simply include a screenshot of your Info panel as well.

GitHub branch: main
main
Version or Latest commit:
v1.1.1 commit e60ed3e
What region is your Azure Open AI Service in?
EastUS
What ChatGPT model are you using?
gpt-4o
model name: (i.e. gpt-3.5-turbo, gpt-4)
gpt-4o
model version: (i.e. 0613)
2024-05-13
What embeddings model are you using?
BAAI/bge-small-env-v1.5

Additional context
Add any other context about the problem here.

If the bug is confirmed, would you be willing to submit a PR?

• Yes
• No

[dayland]

@nhwkuhns , I have put in a PR for this. It will get updated in our next release.