Closed BurnsGG closed 1 year ago
You can probably change this line to Model: openai.GPT3Dot5Turbo
https://github.com/ian-kent/gptchat/blob/e706e89d1eb89a5bcceab754b6337d9520834d7e/chat_loop.go#L116
and give that a try.
I'm NOT a go programmer, but from what I can tell it looks like GPTChat uses https://github.com/sashabaranov/go-openai for openai api stuff. So you can check the docs there too.
Note: You're way more likely to hit token limits tho. GPT4 has much higher token length limits.
I've made a PR addressing this, fell free to scrutinize it.
To try it just fork and define the environment variable OPENAI_API_MODEL as "gpt-3.5-turbo", otherwise it defaults to GPT-4.
And @pale2hall is right, you can just replace the hardcoded openai.GPT4 below with openai.GPT3Dot5Turbo :
https://github.com/ian-kent/gptchat/blob/e706e89d1eb89a5bcceab754b6337d9520834d7e/chat_loop.go#L116 https://github.com/ian-kent/gptchat/blob/e706e89d1eb89a5bcceab754b6337d9520834d7e/module/memory/recall.go#L19
I've merged @NCatalani's PR 🙏
I don't think GPTChat works particularly well on GPT-3.5 though unfortunately, so hopefully you'll get a GPT-4 key soon 🤞
I already applied for the API since it was released and did not receive it yet. Seems like it's country based? so while waiting can we use GPT3.5