[mbil]

Thanks for the feedback. I just updated the article. Basically I plug those extracted values into a templated response.

[babyshake]

Have you had a chance to experiment with whether GPT-3 prompts can be used to dynamically construct the response? For example, "if the messages describes a frontend role, say I am not interested, and if it is a backend role say I am interested."

[mbil]

I haven’t explored it any further yet, but that seems like a useful extension.