Function calling lets a model ask your code to do something: look up an order, check the weather, query a database. The model never runs code itself; it returns a structured request, you run it, and you send the result back.
The loop
- You send the user's message plus a list of tools (name, description, JSON schema of arguments).
- The model replies with a tool call: which function and what arguments.
- Your code runs the function.
- You send the result back as a
toolmessage. - The model writes the final answer using the result.
Complete example (OpenAI format)
import json, os
from openai import OpenAI
client = OpenAI(base_url="https://devsaio.lol/v1", api_key=os.environ["DEVSAIO_API_KEY"])
def get_weather(city: str) -> dict:
# call a real weather API here
return {"city": city, "temp_c": 31, "conditions": "sunny"}
tools = [{
"type": "function",
"function": {
"name": "get_weather",
"description": "Get current weather for a city",
"parameters": {
"type": "object",
"properties": {"city": {"type": "string", "description": "City name"}},
"required": ["city"],
},
},
}]
messages = [{"role": "user", "content": "Should I wear a jacket in Karachi today?"}]
r = client.chat.completions.create(model="claude-sonnet-4-6", messages=messages, tools=tools)
msg = r.choices[0].message
if msg.tool_calls:
messages.append(msg)
for call in msg.tool_calls:
args = json.loads(call.function.arguments)
result = get_weather(**args)
messages.append({"role": "tool", "tool_call_id": call.id, "content": json.dumps(result)})
r = client.chat.completions.create(model="claude-sonnet-4-6", messages=messages, tools=tools)
print(r.choices[0].message.content)Tips for reliable tool use
- Write good descriptions. The model decides when to call a tool from its description.
- Keep schemas small and strict. Use
enumfor fixed choices and mark required fields. - Validate arguments before running anything; never trust them blindly, especially for actions like refunds or deletes.
- Loop until there are no more tool calls, with a maximum number of steps to avoid runaway loops.
- Return errors as results (
{"error": "city not found"}) so the model can recover gracefully.
Function calling is the foundation of AI agents. Most strong models on DEVSAIO support it; check the model's capabilities on the model page.
Claude, GPT, Gemini, DeepSeek, image and video models through one OpenAI- and Anthropic-compatible endpoint. Prepaid, pay only for what you use.