[side-effect] The tool message dump is incomplete

[lvhan028]

Fix side effect brought by #4251

In the following example, the "parameters" part of tools is missed from the dump dict.

from openai import OpenAI
import json

tools = [{
    'type': 'function',
    'function': {
        'name': 'get_current_temperature',
        'description': 'Get current temperature at a location.',
        'parameters': {
            'type': 'object',
            'properties': {
                'location': {
                    'type': 'string',
                    'description': 'The location to get the temperature for, in the format \'City, State, Country\'.'
                },
                'unit': {
                    'type': 'string',
                    'enum': [
                        'celsius',
                        'fahrenheit'
                    ],
                    'description': 'The unit to return the temperature in. Defaults to \'celsius\'.'
                }
            },
            'required': [
                'location'
            ]
        }
    }
},
]



messages = [
    {'role': 'user', 'content': 'Today is 2024-11-14, What\'s the temperature in San Francisco now? How about tomorrow?'}
]

openai_api_key = "EMPTY"
openai_api_base = "http://0.0.0.0:23333/v1"
client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)
model_name = client.models.list().data[0].id
response = client.chat.completions.create(
    model=model_name,
    messages=messages,
    max_tokens=32768,
    temperature=0.8,
    top_p=0.8,
    # stream=False,
    stream=True,
    # extra_body=dict(spaces_between_special_tokens=False, enable_thinking=True),
    extra_body=dict(enable_thinking=True),
    tools=tools
    )
for res in response:
    print(res)

[Copilot]

Pull request overview

Fixes a regression where OpenAI-compatible tool definitions lose the parameters payload when being dumped/serialized.

Changes:

• Adjustes the Pydantic type of Function.parameters to allow preserving arbitrary JSON-schema-like objects in dumps.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

parameters is typed as object, which is overly permissive and allows non-mapping JSON values (e.g., a string/array) to pass validation even though OpenAI tool parameters is expected to be a JSON Schema object. Consider changing this to Optional[Dict[str, Any]] (or Optional[Mapping[str, Any]]) so request validation and generated OpenAPI schema more accurately reflect the expected shape while still preserving the full dict on model_dump().

Suggested change

	parameters: Optional[object] = None
	parameters: Optional[Dict[str, Any]] = None

[Copilot]

This change fixes a regression around tool parameters being dropped during dumps, but there doesn’t appear to be a unit test covering ChatCompletionRequest parsing + Tool/Function.model_dump() retaining nested parameters. Please add a regression test that constructs a request with tools[0].function.parameters set to a nested JSON schema dict and asserts it survives a model_dump()/model_dump_json() round-trip.

[windreamer]

Thank you for fixing my mistakes!