This shows the two-step flow used by the current Cookbook: the model returns a function call first, you execute it locally, then send back a function_call_result for the final answer.

​
1) Initial request (expose your functions)

curl -X POST "https://api.incredible.one/v1/chat-completion" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $INCREDIBLE_API_KEY" \
  -d '{
    "model": "small-1",
    "stream": false,
    "messages": [
      {"role": "user", "content": "What is 25 + 17?"}
    ],
    "functions": [{
      "name": "add_numbers",
      "description": "Add two numbers together",
      "parameters": {
        "type": "object",
        "properties": {
          "a": {"type": "number", "description": "First number"},
          "b": {"type": "number", "description": "Second number"}
        },
        "required": ["a", "b"]
      }
    }]
  }'
The response will include a type: "function_call" item with a function_call_id and the arguments for add_numbers.

​
2) Send back the function result

Execute add_numbers(a, b) locally, then include both the function call item and a function_call_result in a follow-up request: 
curl -X POST "https://api.incredible.one/v1/chat-completion" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $INCREDIBLE_API_KEY" \
  -d '{
    "model": "small-1",
    "stream": false,
    "messages": [
      {"role": "user", "content": "What is 25 + 17?"},
      {"type": "function_call", "function_call_id": "CALL_ID_FROM_STEP_1", "function_calls": [{
        "name": "add_numbers",
        "input": {"a": 25, "b": 17}
      }]},
      {"type": "function_call_result", "function_call_id": "CALL_ID_FROM_STEP_1", "function_call_results": [42]}
    ],
    "functions": [{
      "name": "add_numbers",
      "description": "Add two numbers together",
      "parameters": {
        "type": "object",
        "properties": {
          "a": {"type": "number"},
          "b": {"type": "number"}
        },
        "required": ["a", "b"]
      }
    }]
  }'
The assistant will return the final natural-language answer using the function result.
View source on GitHub → 5_function_calling.py