How to Intermediate · 3 min read

How to chain structured output calls

Q: How to chain structured output calls

Use the response_format parameter in AI API calls to define expected output schemas, then pass outputs from one call as inputs to the next to chain calls. This enables reliable multi-step workflows with typed data between calls.

Quick answer

Use the response_format parameter in AI API calls to define expected output schemas, then pass outputs from one call as inputs to the next to chain calls. This enables reliable multi-step workflows with typed data between calls.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the openai Python SDK and set your API key as an environment variable.

Run pip install openai
Set export OPENAI_API_KEY='your_api_key' on Linux/macOS or setx OPENAI_API_KEY "your_api_key" on Windows

bash

pip install openai

Step by step

This example demonstrates chaining two structured output calls using the OpenAI SDK. The first call extracts user info as a structured JSON object. The second call uses that output to generate a personalized greeting.

python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Step 1: Extract structured user info
response1 = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Extract name and age from: 'Alice is 30 years old.'"}],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "type": "object",
            "properties": {
                "name": {"type": "string"},
                "age": {"type": "integer"}
            },
            "required": ["name", "age"]
        }
    }
)
user_info = response1.choices[0].message.content

# Step 2: Use structured output from step 1 to generate greeting
response2 = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": "You are a friendly assistant."},
        {"role": "user", "content": f"Generate a greeting for a person named {user_info['name']} who is {user_info['age']} years old."}
    ]
)
print("Extracted user info:", user_info)
print("Greeting:", response2.choices[0].message.content)

output

Extracted user info: {'name': 'Alice', 'age': 30}
Greeting: Hello Alice! It's great to know you are 30 years old.

Common variations

You can chain structured outputs asynchronously using async Python with the OpenAI SDK. Also, you can use different models like gpt-4.1 or Anthropic's claude-3-5-sonnet-20241022 with similar structured output parameters. For streaming, handle partial outputs but chaining requires waiting for full structured output.

python

import asyncio
import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

async def main():
    response1 = await client.chat.completions.acreate(
        model="gpt-4.1",
        messages=[{"role": "user", "content": "Extract city and country from: 'Paris, France'"}],
        response_format={
            "type": "json_schema",
            "json_schema": {
                "type": "object",
                "properties": {
                    "city": {"type": "string"},
                    "country": {"type": "string"}
                },
                "required": ["city", "country"]
            }
        }
    )
    location = response1.choices[0].message.content

    response2 = await client.chat.completions.acreate(
        model="gpt-4.1",
        messages=[{"role": "user", "content": f"Describe the weather in {location['city']}, {location['country']}."}]
    )
    print("Location:", location)
    print("Weather description:", response2.choices[0].message.content)

asyncio.run(main())

output

Location: {'city': 'Paris', 'country': 'France'}
Weather description: Paris, France typically enjoys mild spring weather with occasional showers and pleasant temperatures.

Troubleshooting

If the structured output is missing or malformed, verify your response_format JSON schema matches the expected output format.
Ensure the model supports structured output (e.g., gpt-4o or newer).
If chaining fails, check that you correctly extract and pass the structured output between calls.

✅

Key Takeaways

Use the response_format parameter to define expected output schemas for reliable data extraction.
Chain calls by passing structured outputs as inputs to subsequent calls to build multi-step workflows.
Use async calls for concurrency but await full structured output before chaining.
Validate your JSON schema to avoid malformed outputs.
Choose models like gpt-4o or gpt-4.1 that support structured output.

Verified 2026-04 · gpt-4o, gpt-4.1, claude-3-5-sonnet-20241022

Verify ↗