Continuation of model response with hardcoded previous session

API
, , ,
1

Hello,

i have the goal of continuing a model response with a hardcoded supplied history. It is easiest if i demonstrate how this may look like:

from openai import OpenAI
client = OpenAI()

response = client.chat.completions.create(
  model="gpt-5",
  messages=[
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "hello what is a monkey"
        }
      ]
    },
    {
      "role": "assistant",
      "content": [
        {
          "type": "text",
          "text": "A monkey is a primate in the simian group that is not an ape. They’re intelligent, mostly social mammals found in"
        }
      ]
    }
  ],
  response_format={
    "type": "text"
  },
  verbosity="medium",
  reasoning_effort="medium"
)

print(response)

I now envision that the model continues the initial response. So something in the realm of:

”tropical forests, savannas, and mountains.”

But unfortunately i get a completely new response, as if the initial response had never occured (perhaps it also noticed that there was an incomplete response and decided to generate a new one as the other one seems truncated).

Is there any way i get this continuation without special continuation tokens, extra instruction prompts or similar? I’d like to stay as close as possible to the “raw intelligence” without prompt overhead as im using this technique for benchmarks.

best regards,

Platinium