Guides

Tool calls

Both models call tools in OpenAI’s format: you describe functions, the model asks for one, you run it and send back the result.

Checked 26 October 2026

The loop

Send tools: each one a function with a name, a description and JSON Schema parameters.
When the model wants one, the reply’s finish_reason is tool_calls and message.tool_calls lists them, with arguments as a JSON string.
Run each one, and add a tool message with its tool_call_id and the result.
Send the conversation again. Repeat until finish_reason is stop.

Python

import json
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://nokeep.ai/v1",
    api_key=os.environ["NOKEEP_API_KEY"],
)

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Today's weather in a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]

def get_weather(city):
    return {"city": city, "high_c": 21, "sky": "clear"}

messages = [{"role": "user", "content": "Should I bring a jacket in Lisbon today?"}]
while True:
    reply = client.chat.completions.create(model="deepseek-v4.1-flash-uncensored", messages=messages, tools=tools)
    message = reply.choices[0].message
    messages.append(message)
    if not message.tool_calls:
        print(message.content)
        break
    for call in message.tool_calls:
        result = get_weather(**json.loads(call.function.arguments))
        messages.append({"role": "tool", "tool_call_id": call.id, "content": json.dumps(result)})

tool_choice

Model	Takes
`deepseek-v4.1-flash-uncensored`	`auto`, `none`, `required`, or a named function. Also `parallel_tool_calls`.
`glm-5.3-flash-uncensored`	`auto` or `none`. It picks its own tools; anything else is `400 unsupported_tool_choice`.

Good to know

Every tool message needs the tool_call_id it answers, or the call is refused with 400 invalid_messages.
Tool schemas travel with every request and count as input tokens.
The models don’t run anything themselves. Your code runs the tools, so it decides what they can reach.

Get an API key Try the free test key: sk-nokeep-test

The loop

tool_choice

Good to know

What’s on your mind?