Persistently Setting System Instructions or Code of Conduct for GPT-OSS:20B

Open Models
1

Hello OpenAI team,

I am currently running GPT-OSS:20B within an Ollama container on a Debian system. I would like to know if there is a way to impart system instructions or a code of conduct to the model persistently, so that the model follows them automatically without needing to be provided with these instructions on every single API call.

From my understanding, I can include system instructions in each API request, but I am looking for a solution where I don’t have to repeat them every time. Is it possible to configure GPT-OSS:20B in a way that it “remembers” or internalizes these instructions? If so, could you please explain how this can be achieved?

Thank you very much for your cooperation!

2

It is not a “persistent setting”, but you can affect the way the AI model behaves with fine-tuning. That is the alternative to in-context learning - placing text input which would always need to be present.

1 Like