Open olsonperrensen opened 8 months ago
Keep in mind that the model used for answering any questions is gpt-3.5-turbo
I'd suggest you either use an env var to add flexibility, or change the model to gpt-4
As far as I can tell they merged the generative UI demo - which was for a stock trading bot - with this one The prompt for the AI is currently:
You are a stock trading conversation bot and you can help users buy stocks, step by step.
You and the user can discuss stock prices and the user can adjust the amount of stocks they want to buy, or place an order, in the UI.
Messages inside [] means that it's a UI element or a user event. For example:
- "[Price of AAPL = 100]" means that an interface of the stock price of AAPL is shown to the user.
- "[User has changed the amount of AAPL to 10]" means that the user has changed the amount of AAPL to 10 in the UI.
I'm pretty confused about this whole change - should this be reverted to the original prompt?
As far as I can tell they merged the generative UI demo - which was for a stock trading bot - with this one The prompt for the AI is currently:
You are a stock trading conversation bot and you can help users buy stocks, step by step. You and the user can discuss stock prices and the user can adjust the amount of stocks they want to buy, or place an order, in the UI. Messages inside [] means that it's a UI element or a user event. For example: - "[Price of AAPL = 100]" means that an interface of the stock price of AAPL is shown to the user. - "[User has changed the amount of AAPL to 10]" means that the user has changed the amount of AAPL to 10 in the UI.
I'm pretty confused about this whole change - should this be reverted to the original prompt?
Ya, I don’t know what the demo was previously about, but it’s now specifically about stock trading as you’ve mentioned, perhaps to demonstrate the use of system messages. It would have made for a better template if the system messages were left to the user to configure.
Hi, this has happened numerous times since yesterday:
The chat AI always replies like that for some reason?
Any changed committed recently that explain this new behavior?