[BUG] Azure OpenAI for reasoning models.

[adityamcodes]

Describe the bug
When trying to leverage o3, o4-mini, claude thinking, etc - the system uses the older versions of langchain which break with o3 & o4-mini.

• Specifically tested with AzureOpenAI instance

To Reproduce
Steps to reproduce the behavior:

• leverage Azure OpenAI with o4-mini or o3
• integrate with anything (swarmrouter, hhcs, etc)
• run any kind of swarm

Expected behavior
should work lol

Screenshots
will add later when time available

Additional context
This is an extremely high priority for us and our clients. If this functions, we will shift to an enterprise plan

[kyegomez]

@adityamcodes you can use this with litellm and the Agent class Agent(model_name="azure/o1-pro") etc

Can read more here:

https://docs.litellm.ai/docs/providers/azure#azure-batches-api

also need envs like this

import os
os.environ["AZURE_API_KEY"] = "" # "my-azure-api-key"
os.environ["AZURE_API_BASE"] = "" # "https://example-endpoint.openai.azure.com"
os.environ["AZURE_API_VERSION"] = "" # "2023-05-15"

optional

os.environ["AZURE_AD_TOKEN"] = ""
os.environ["AZURE_API_TYPE"] = ""

[kyegomez]

Here's some new docs we added

https://docs.swarms.world/en/latest/swarms/models/agent_and_models/

[adityamcodes]

last time i tried it broke bc of default params setup in litellm (i think), but its been a few weeks so let me try it

[kyegomez]

@adityamcodes hmmmm let me know if it works

[adityamcodes]

yeah it didnt end up working, was very finicky

[kyegomez]

Can you share your example code please?

[adityamcodes]

yeah lemme try and find it

[adityamcodes]

`from swarm_models.popular_llms import AzureOpenAILLM
import os

from dotenv import load_dotenv

load_dotenv()

assert os.environ["AZURE_API_KEY"]
assert os.environ["AZURE_API_BASE"] # "https://example-endpoint.openai.azure.com"
assert os.environ["AZURE_API_VERSION"] # "2023-05-15"

agent = AzureOpenAILLM(model_name="azure/o4-mini",
api_key=os.environ["AZURE_API_KEY"],
azure_endpoint=os.environ["AZURE_API_BASE"],
api_version=os.environ["AZURE_API_VERSION"],
)

agent.run("reply hello")
`

openai.NotFoundError: Error code: 404 - {'error': {'code': '404', 'message': 'Resource not found'}}

env's are correct

[adityamcodes]

python3.11,
Name: swarms
Version: 7.8.4
Name: swarm-models
Version: 0.3.2

[adityamcodes]

can't seem to do from swarms import Agent

[adityamcodes]

ugh got it,

tediously

agent = Agent(model_name="azure/o4-mini", temperature=1, llm_args={"drop_params": True})

[beyto1974]

The error I got on swarms 7.8.4:

Error generating response: litellm.UnsupportedParamsError: azure does not support parameters: ['top_p'], for model=o4-mini. To drop these, set `litellm.drop_params=True` or for proxy:

`litellm_settings:
 drop_params: true`
. 
 If you want to use these params dynamically send allowed_openai_params=['top_p'] in your request.

The following gives same error:

agent = Agent(model_name="azure/o4-mini", temperature=1, llm_args={"drop_params": True})

The following is working:

import litellm

litellm.drop_params = True

agent = Agent(model_name="azure/o4-mini")

With these .env variables:

AZURE_API_BASE=https://<my-ai-resource>.openai.azure.com/openai/deployments/<my-deployment-name>
AZURE_API_KEY=<my-deployment-key>