I have around 20 messages and when I tried to generate a new one, I encountered the error “this model maximum context length is 4096 tokens. However, your messages resulted in 4098 tokens.”
It’s kind of frustrating because I thought the CHAT-GPT API accepted unlimited messages, but apparently it doesn’t. How can my application start to accept unlimited messages or something like that? Without losing the context of the conversation or the CHAT-GPT remembering the first message sent?
Sorry, I’m Brazilian and not fluent in English and, therefore, some information may seem distorted in the page’s automatic translation. (OpenAI Docs)
But, getting back to my question, is there any way to circumvent this limit? Because, if so, it’s a replica of GPT-3. I used to do chatbot with GPT-3 and, to circumvent the limit, I just took the last 10 messages sent by the user. But, in CHAT-GPT, will I also need to do this?
There is no way to “get around the limit” but you can manage your token count just as you describe with GPT-3:
You must manage token count with the chat completion as well:
Yes, you need to manage tokens.
