OpenAI client libraries with GitHub Models

[achandmsft]

Select Topic Area

Question

Body

Creating this discussion to track issues around using the official OpenAI client libraries for Python, NET/C#, JavaScript/TypeScript, etc with GitHub AI models

[kristapratico]

Issue: API version required when using structured outputs with GH models

I’m using the Python OpenAI library with GH models and found that it requires API version when making calls that use structured outputs.

Example code and error below:

class Step(BaseModel):
    explanation: str
    output: str
 
class MathResponse(BaseModel):
    steps: List[Step]
    final_answer: str
 
client = openai.OpenAI(
    base_url="https://models.inference.ai.azure.com/",
    api_key=os.environ["GH_TOKEN"],
)
 
completion = client.beta.chat.completions.parse(
    messages=[
        {"role": "user", "content": "solve 8x + 31 = 2"},
    ],
    response_format=MathResponse,
    model="gpt-4o"
)

Error:

openai.BadRequestError: Error code: 400 - {'error': {'code': 'BadRequest', 'message': 'response_format value as json_schema is enabled only for api versions 2024-08-01-preview and later'}}

I can workaround this by passing in the API version as a default query parameter in the client, but I shouldn't have to do this:

default_query = {"api-version": "2024-08-01-preview"}

client = openai.OpenAI(
    base_url="https://models.inference.ai.azure.com",
    api_key=os.environ["GITHUB_TOKEN"],
    default_query=default_query
)

[Johnballboa]

Issue: API version required when using structured outputs with GH models

I’m using the Python OpenAI library with GH models and found that it requires API version when making calls that use structured outputs.

Example code and error below:

class Step(BaseModel):
    explanation: str
    output: str
 
class MathResponse(BaseModel):
    steps: List[Step]
    final_answer: str
 
client = openai.OpenAI(
    base_url="https://models.inference.ai.azure.com/",
    api_key=os.environ["GH_TOKEN"],
)
 
completion = client.beta.chat.completions.parse(
    messages=[
        {"role": "user", "content": "solve 8x + 31 = 2"},
    ],
    response_format=MathResponse,
    model="gpt-4o"
)

Error:

openai.BadRequestError: Error code: 400 - {'error': {'code': 'BadRequest', 'message': 'response_format value as json_schema is enabled only for api versions 2024-08-01-preview and later'}}

I can workaround this by passing in the API version as a default query parameter in the client, but I shouldn't have to do this:

default_query = {"api-version": "2024-08-01-preview"}

client = openai.OpenAI(
    base_url="https://models.inference.ai.azure.com",
    api_key=os.environ["GITHUB_TOKEN"],
    default_query=default_query
)

Please help me im verry blur about that issue

[Johnballboa]

Select Topic Area

Question

Body

Creating this discussion to track issues around using the official OpenAI client libraries for Python, NET/C#, JavaScript/TypeScript, etc with GitHub AI models

How i can i do because i dont have a brain for my issues now