Hi all,
I have a simple setup which uses the function calling example from the API documentation (What’s the weather in boston). I have it configured to send to model: gpt-3.5-turbo-0613.
The only change vs the example is that I’ve converted it to node.js.
Weirdly, I’m getting an empty response to the call:
{ totalTokens: 0, finishReason: ‘’, completionText: ‘’ }
If I expand the demo to allow a conversation to occur, the rest of the calls work fine and I get the expected response. However, if I ask it what the weather is like anywhere (therefore triggering the possibility of returning the function) I get the empty response.
As such I see that it is correctly identifying when it needs to return a function as the reponse, but then returning me a blank response rather than one documenting the function to call.
Has anybody else experienced this?
Many thanks!
Luke