Function calling

Chat completions accept the usual OpenAI tools and tool_choice fields and return tool_calls with finish_reason: tool_calls. Tool definitions count toward the input-token reservation, so large schemas reserve more credit up front.

curl https://api.refuseless.dev/v1/chat/completions \
  -H "Authorization: Bearer $REFUSELESS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3-refuseless",
    "messages": [{"role": "user", "content": "What is the weather in Paris?"}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the weather for a city",
        "parameters": {
          "type": "object",
          "properties": { "city": {"type": "string"} },
          "required": ["city"]
        }
      }]
  }'

The assistant turn comes back with content: null and a tool_calls array. Execute the function yourself, append a tool-role message with the result, and call again — the standard OpenAI loop, unchanged. The legacy /v1/completions endpoint has no tool channel; use chat for tools.

Start

API reference

Guides

Function calling

Function calling

​Function calling

Function calling