When max_token, how to continue the response context?

When I send a request asking it to rewrite or extend the length of an article for me, sometimes it may exceed the token limit, and it stops in an incomplete state. Just like the ‘continue’ feature on the ChatGPT UI page, is there a similar way to achieve this with the API?

I tested on playground, initially set the maximum length to 8.
Got an incomplete response. Adjusted the maximum length and told it to “please continue”. Seems to work.

You don’t need to send another user message, you can just submit the AI response as “assistant” and it will continue writing. With a completion engine, you don’t even have to do that, just submit the whole built-up prompt again without the “AI:”

1 Like

You’re right! Nice!

Playground test: Haiku Interrupted

If the total user+assistant interaction exceeds 4096 tokens, is it not possible to continue with the conversation by using the ‘resend assistant’ method to obtain the following content?