Is it possible to achieve this?

[bold hornet]

Hello, I'm investigating chatgpt & api for development purpose and integration with an enterprise app.
I'd need to know if it's technically possible to imbue a chatgpt "instance" with a specific personality that stays always consistent.
I've played with chatgpt app trying to accomplish something but it's quite fickle and many times it spit out long kind of disclaimer text that we don't want to output.
Just to put in context, one use case we may have is having a web app companion with a specific personality that can interact with our users. Also I'd like to know if it's possible to train it by giving our web app documentation (actually it's a set of markdown pages) in such a way that it could help users to have an advanced kind of help

[valid bear]

Not specifically ChatGPT but you can train Davinci which also uses GPT3 (Also known as GPT3.5)

Everything you're asking is exactly what OpenAI offers in their API documentation.

If you are receiving disclaimers then you are asking it to do something that A. It cannot without some help or B. It won't because of the potential issues that may follow

[mild pendant]

It would absolutely be possible! You could eather fine-tune the model or provide examples in your prompt. Just like @valid bear said, you cannot (yet) use the actual ChatGPT model to do this but you can train a GPT-3 model yourself.