Dear repository maintainers,
I have noticed that when calling GPT-3.5/GPT-4, the completion results do not continue directly after the partial program, but instead output an independent and complete response. This makes it impossible to extend new nodes based on the given partial program. Therefore, can GPT-3.5/GPT-4 be used for this project, or is it necessary to use a locally deployed Huggingface model?
Thank you for your assistance.