Gemini is usually too verbose.
Instead of adding “be terse” to each prompt, I was wondering if there is an equivalent to ChatGPT “customize how GPT responds”?
Welcome to the forum. The system instruction can be effective, if you include something like “brief” to indicate that the response should be less verbose. Once set, you then don’t need to repeat it in each prompt.
System Instructions don’t even work
Can you expand more on that? So far I’ve been using it and I was able to effectively control the response format and process.
I am relying fully on API however, not the SDK.
Thank you, this works well but doesn’t stay on new chat. Is there a way to keep it as default?
To make the system instruction appear permanent, set it in the GenerativeModel class: generative-ai-python/docs/api/google/generativeai/GenerativeModel.md at main · google-gemini/generative-ai-python · GitHub The interaction using either generate_content or start_chat will adopt what is set there.
Hope that helps.
Simple, i did an entire block list in the System Instructions, Testament, Stark Constrat, Malevolent are all meant to be blocked, yet this is not working, this is a bug that i don’t tolerate, so each times this happens, i do a Send Feedback, douzens and douzens, yet no bug fixing yet