Function calling API — same shape as OpenAI, 18+ models

Function calling (also called "tool use") is how you turn an LLM into an agent: the model returns a structured request to call one of your functions, your code runs the function, and the result is fed back to the model. Plugsky implements the OpenAI shape, so LangChain, LlamaIndex, Vercel AI SDK, and the official OpenAI libraries all work unchanged.

The function calling loop

from openai import OpenAI
import json

client = OpenAI(base_url="https://api.plugsky.com/v1", api_key="sk-live-...")

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get current weather for a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"]
        }
    }
}]

def run_loop(user_msg):
    messages = [{"role": "user", "content": user_msg}]
    while True:
        r = client.chat.completions.create(model="plugsky-pro", messages=messages, tools=tools)
        msg = r.choices[0].message
        if not msg.tool_calls:
            return msg.content
        messages.append(msg)
        for call in msg.tool_calls:
            args = json.loads(call.function.arguments)
            result = get_weather(**args)  # your function
            messages.append({"role":"tool","tool_call_id":call.id,"content":json.dumps(result)})

print(run_loop("What's the weather in Dubai?"))

Tool definition

Tools are JSON Schema objects. Plugsky validates the parameters against your schema before sending the call back to your code:

• type: "object" with properties and required
• String, number, integer, boolean, array, enum, nested objects
• Optional description on every field — the LLM uses these to decide what to call
• Optional enum to constrain string values to a fixed set

Parallel calls and JSON mode

Plugsky supports:

• Parallel function calls: the model can return multiple tool calls in one response, and your code dispatches them concurrently
• tool_choice: force the model to call a specific tool, or "auto", or "none"
• response_format: {"type":"json_object"} for guaranteed JSON output
• Strict mode: with "strict": true on a tool, the model is forced to use exactly the schema you provide (no extra fields)

Building agents

Combine function calling with the AI agents API for a managed agent runtime with memory, retries, and audit logs. Or roll your own — the loop above works for any number of tools and steps.

Frequently asked questions