ValueError
azure.openai._function_calling.FunctionCallingSchemaError
Stack trace
ValueError: Function calling schema is invalid: 'parameters' field must be a valid JSON Schema object
File "/usr/local/lib/python3.9/site-packages/azure/ai/openai/_function_calling.py", line 123, in _validate_function_schema
raise FunctionCallingSchemaError("Function calling schema is invalid: 'parameters' field must be a valid JSON Schema object") Why it happens
Azure OpenAI requires the function calling schema to strictly follow JSON Schema standards. If the 'parameters' field or other parts of the schema are missing, malformed, or not valid JSON Schema objects, the SDK raises this error to prevent sending invalid requests to the API.
Detection
Validate your function schema JSON against the official JSON Schema specification before passing it to the Azure OpenAI client. Log schema validation errors or catch ValueError exceptions during client calls.
Causes & fixes
The 'parameters' field in the function schema is missing or not a valid JSON Schema object.
Ensure the 'parameters' field is a properly formatted JSON Schema object with correct types and required fields as per Azure OpenAI documentation.
The function schema contains unsupported or incorrectly typed fields not allowed by Azure OpenAI.
Remove unsupported fields and verify all fields conform exactly to the JSON Schema draft supported by Azure OpenAI.
Passing the function schema as a string instead of a Python dictionary.
Pass the function schema as a Python dict object, not as a JSON string, so the SDK can validate and serialize it correctly.
Code: broken vs fixed
import os
from azure.ai.openai import OpenAIClient
client = OpenAIClient(os.environ["AZURE_OPENAI_API_KEY"])
functions = [
{
"name": "get_weather",
"description": "Get the weather forecast",
"parameters": "{\"type\": \"object\", \"properties\": {\"location\": {\"type\": \"string\"}}}" # Incorrect: parameters as string
}
]
response = client.chat.completions.create(
engine="gpt-4o",
messages=[{"role": "user", "content": "What's the weather in Seattle?"}],
functions=functions # This line triggers the ValueError
)import os
from azure.ai.openai import OpenAIClient
client = OpenAIClient(os.environ["AZURE_OPENAI_API_KEY"])
functions = [
{
"name": "get_weather",
"description": "Get the weather forecast",
"parameters": {
"type": "object",
"properties": {
"location": {"type": "string"}
},
"required": ["location"]
} # Fixed: parameters as dict with valid JSON Schema
}
]
response = client.chat.completions.create(
engine="gpt-4o",
messages=[{"role": "user", "content": "What's the weather in Seattle?"}],
functions=functions # Fixed: valid schema passed
)
print(response.choices[0].message.content)Workaround
Catch the ValueError when creating the chat completion, then manually validate and fix the function schema JSON before retrying the request.
Prevention
Always define function calling schemas as Python dictionaries conforming strictly to JSON Schema standards and validate them with a JSON Schema validator before passing to Azure OpenAI.