When using ChatGPT, it is very good at keeping the context of requests, I can ask it to generate 4 colors, and then ask for 4 more and it understand. With the completions api, I can’t figure out how to do this.
For example I am asking for 3 colors, fine I get 3 colors, then I ask for 3 more, and most of the time I will get the same colors, and some of the colors again.
ChatGPT is essentially GPT3.5
When you use the completions endpoint, the available models only use up to GPT3. It would look like this:
model: "text-davinci-003",
If you want to obtain the functionality of ChatGPT and interact with it via an API, I would recommend checking out this topic where they discuss open source solutions.
You need to include the previous response in the next query so it knows what it has said so far
Like this
Q: previous question
A: previous answer
Q: give me 4 more examples
A:
You might want to rotate the last 3 or 4 interactions above your final question. This is how you code a chatbot with gpt and davinci