Skip to main content
Version: 1.0.5

Structured Data with ASI:One

ASI:One can return responses that conform to a JSON schema you define. This example shows how to extract structured information—a CalendarEvent—directly from natural language using tool calling and Pydantic for type-safe parsing.

from pydantic import BaseModel
from enum import Enum
import uuid
import json
from openai import OpenAI, pydantic_function_tool

# ASI1 API config
ASI_API_KEY = "<YOUR_ASI_API_KEY>"
ASI_BASE_URL = "https://api.asi1.ai/v1"

# Calendar event model
class CalendarEvent(BaseModel):
    name: str
    date: str
    participants: list[str]

# Init client
client = OpenAI(api_key=ASI_API_KEY, base_url=ASI_BASE_URL)

session_id = str(uuid.uuid4())
print("🔄 Sending calendar-event request to ASI1 experimental")
print(f"🆔 Session ID: {session_id}\n")

completion = client.beta.chat.completions.parse(
    model="asi1-experimental",
    messages=[
        {"role": "system", "content": "Extract the event information by calling the calendar_event function."},
        {"role": "user", "content": "Alice and Bob are going to a science fair on Friday."}
    ],
    tools=[pydantic_function_tool(CalendarEvent, name="calendar_event")],
    extra_headers={"x-session-id": session_id}
)

if completion.choices[0].message.tool_calls:
    event: CalendarEvent = completion.choices[0].message.tool_calls[0].function.parsed_arguments
    print("✅ CalendarEvent parsed: \n")
    print(json.dumps(event.model_dump(), indent=2))
else:
    print("⚠️ No tool call returned. Raw message:")
    print(completion.choices[0].message)

print(f"\n🆔 Session ID for debugging: {session_id}")

Expected Output​

The assistant calls the calendar_event tool and the parsed result is automatically converted into a CalendarEvent instance:

{
  "name": "science fair",
  "date": "Friday",
  "participants": [
    "Alice",
    "Bob"
  ]
}

You now have strongly-typed data ready for further processing (saving to a database, triggering notifications, etc.).

Next Steps​