What exactly does 'system content' do?

API
1

I found the code sample quoted below, in the documentation. The most important part of it is

messages = [
    {"role": "system", "content": context},
    {"role": "user", "content": question}
]

What exactly does this do, in terms of e.g. what would be the equivalent ChatGPT input? Is ‘system content’ syntax sugar for something you could write as a prompt, or is there something else going on?

from openai import OpenAI

client = OpenAI()

Replace ‘your_openai_api_key’ with your actual OpenAI API key

def ask_question_with_context(context, question):
messages = [
{“role”: “system”, “content”: context},
{“role”: “user”, “content”: question}
]

response = client.chat.completions.create(model="gpt-4",
messages=messages)

return response.choices[0].message.content

Example usage

context = “”"
The country of Eldoria is a fictional land located in the heart of the Pacific Ocean. Known for its lush landscapes,
exotic wildlife, and rich cultural heritage, Eldoria has a population that speaks Eldorian, a language native to the
island. The country operates as a democratic republic and is famed for its unique architecture and vibrant festivals.
Its capital city is Florkin.
“”"

question = “What is the capital of Eldoria?”
answer = ask_question_with_context(context, question)
print(answer)

2

It’s called the “system prompt” and guides the way you want the chat completion responses to behave.

For example, in the context of a chatbot, you could write something like:

“You are a bot that talks like Winston Churchill, you are always cracking jokes”

etc. etc.

2 Likes