Hello all!
This is an idea I’d like to share with you to discuss.
As a soon-to-be former user of the Assistants API, I’d find useful to have a “Prompt” entity, simmilar to what an “Assistant” currently is, that could be invoked whenever creating a model response.
Such “Prompt” entity would have every related parameter (such as system prompt, tools, temperature, etc.), all configurable through the OpenAI Platform webpage, so that API requests could just have a parameter “prompt” with the ID of an specific prompt, that would override sending the promp parameters explicitly on the body request.
Just as an example, what I’m proposing would look like this:
{
“prompt_id”: “prompt_somePromptId”,
“input”: [
{
“role”: “user”,
“content”: [
{
“type”: “input_text”,
“text”: “Some input…”
}
]
}
]
}
I hope to be clear, but I’m open to discussion!!
BR,
Christian