TokelyGUIDES
Connect models to your tools.
Let a model request a function, then execute it in your application.
Define a function
Choose a model with function calling support. Describe the function with a JSON schema and send it in tools. This Chat Completions example lets the model request a local weather lookup.
{
"model": "gpt-4o-mini",
"messages": [
{
"role": "user",
"content": "What is the weather in Helsinki?"
}
],
"max_tokens": 256,
"tools": [
{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a city",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string"
}
},
"required": [
"city"
]
}
}
}
]
}Execute and return the result
When message.tool_calls is present, validate the function name and arguments before executing code. Append the assistant message to history, then a tool message with the matching tool_call_id and your result as text. Send the updated messages for the final answer.
Your application executes the function. Allow intended operations and enforce your permissions. Up to 32 function tools are accepted per request.
A complete function call
This runnable Python example asks the model to call add_numbers. Your Python code validates the arguments, calculates the result and returns it to the model. It makes two generation requests when a function is called.
Install the SDK and set TOKELY_API_KEY as shown in Quickstart. Save this example as tools.py and run python3 tools.py. The model should explain that 12 plus 7 is 19; its wording may vary.
import json
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["TOKELY_API_KEY"],
base_url="https://api.tokely.me/v1",
)
messages = [{"role": "user", "content": "What is 12 plus 7?"}]
tools = [{
"type": "function",
"function": {
"name": "add_numbers",
"description": "Add two numbers.",
"parameters": {
"type": "object",
"properties": {
"a": {"type": "number"},
"b": {"type": "number"},
},
"required": ["a", "b"],
"additionalProperties": False,
},
},
}]
response = client.chat.completions.create(
model="gpt-4o-mini", messages=messages, tools=tools,
tool_choice={"type": "function", "function": {"name": "add_numbers"}},
max_tokens=256,
)
message = response.choices[0].message
if message.tool_calls:
messages.append({
"role": "assistant",
"content": message.content,
"tool_calls": [call.model_dump() for call in message.tool_calls],
})
for call in message.tool_calls:
if call.function.name != "add_numbers":
raise ValueError("Unexpected function")
arguments = json.loads(call.function.arguments)
if set(arguments) != {"a", "b"} or any(
type(value) not in (int, float) for value in arguments.values()
):
raise ValueError("Invalid function arguments")
result = arguments["a"] + arguments["b"]
messages.append({
"role": "tool",
"tool_call_id": call.id,
"content": json.dumps({"result": result}),
})
final = client.chat.completions.create(
model="gpt-4o-mini", messages=messages, max_tokens=256,
)
print(final.choices[0].message.content)
else:
print(message.content)Functions with Responses
Responses uses function definitions with name, description and parameters directly on the tool object. Continue with function_call and function_call_output items, matching their call_id. Provider-hosted web search, code execution and other built-in tools are not enabled.