Debug Fix beginner · 3 min read

How to handle function call results

Q: How to handle function call results

Use the function_call field in the OpenAI chat completion response to detect and extract function call results. Parse the arguments JSON string from response.choices[0].message.function_call.arguments to handle structured outputs properly.

Quick answer

Use the function_call field in the OpenAI chat completion response to detect and extract function call results. Parse the arguments JSON string from response.choices[0].message.function_call.arguments to handle structured outputs properly.

ERROR TYPE model_behavior

⚡ QUICK FIX

Parse the function_call.arguments JSON string from the response to correctly handle function call results.

Why this happens

When using function_call with the OpenAI API, the model returns the function call details inside the message.function_call object. The arguments field is a JSON string, not a Python dict, so if you treat it as a dict directly, you get errors or incorrect handling.

Example broken code:

from openai import OpenAI
import os

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

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Call the weather function"}],
    functions=[{
        "name": "get_weather",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            },
            "required": ["location"]
        }
    }],
    function_call="auto"
)

# Incorrect: treating arguments as dict directly
args = response.choices[0].message.function_call.arguments
print(args["location"])

This raises a TypeError because arguments is a JSON string, not a dict.

python

from openai import OpenAI
import os

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

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Call the weather function"}],
    functions=[{
        "name": "get_weather",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            },
            "required": ["location"]
        }
    }],
    function_call="auto"
)

# Incorrect: treating arguments as dict directly
args = response.choices[0].message.function_call.arguments
print(args["location"])

output

TypeError: string indices must be integers

The fix

Parse the arguments string as JSON to convert it into a Python dict before accessing its fields. This ensures you handle the structured output correctly.

python

from openai import OpenAI
import os
import json

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

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Call the weather function"}],
    functions=[{
        "name": "get_weather",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string"}
            },
            "required": ["location"]
        }
    }],
    function_call="auto"
)

func_call = response.choices[0].message.function_call
args = json.loads(func_call.arguments)
print(f"Function called: {func_call.name}")
print(f"Arguments: {args}")
print(f"Location argument: {args['location']}")

output

Function called: get_weather
Arguments: {'location': 'San Francisco'}
Location argument: San Francisco

Preventing it in production

Always validate and parse function_call.arguments as JSON before use. Implement error handling for JSON parsing failures to catch malformed responses. Use retries with exponential backoff for transient API errors. Consider fallback logic if the function call is missing or incomplete.

python

import json

def handle_function_call(response):
    try:
        func_call = response.choices[0].message.function_call
        if func_call and func_call.arguments:
            args = json.loads(func_call.arguments)
            # Process args safely here
            return func_call.name, args
        else:
            return None, None
    except json.JSONDecodeError:
        # Log error and fallback
        return None, None

Related errors

Error	Cause	Quick fix
TypeError: string indices must be integers	Accessing `arguments` as dict without JSON parsing	Parse `arguments` with `json.loads()`
JSONDecodeError	Malformed JSON string in `arguments`	Add try-except around `json.loads()` and validate input
KeyError	Accessing missing keys in parsed arguments	Check keys exist before access or use `dict.get()`

✅

Key Takeaways

Always parse function_call.arguments as JSON before accessing its contents.
Implement error handling around JSON parsing to avoid runtime crashes.
Use retries and validation to handle transient API or malformed response issues.

Verified 2026-04 · gpt-4o

Verify ↗