Hi All,
I'm challenging myself to integrate WxO with Guardium AI using the gateway feature, so I can show customers and partners how it's possible to guarantee security for agents and LLMs. The idea is to demonstrate how a platform like Guardium AI gives security teams the right tools and places responsibilities in the proper areas of the company.
One of the most powerful features in Guardium AI is its ability to act as a secure gateway for LLM connections. Instead of allowing developers to call public LLM APIs directly, we can give them a controlled proxy URL to use inside the company. This shifts security, compliance, and audit responsibilities to the right teams and significantly reduces risk.
For this lab, I’m using the ADK to create a new connection and add a new model that points to the Guardium AI proxy. I'm working with the OpenAI provider, but I'm running into an issue when using the orchestrate models add command — specifically in the --provider-config section, even though everything seems to follow the documentation.
Here’s the command I'm using:
openai-gpt-5 is the connections a created, as the follow:
When I execute myscript, i have the following pydantic_core error, certainly due to it is using opanai provider:
If I remove the list in "response_headers" inside the command to create the new model, the model is create, but has connection error from Guardium AI:
Alltrue is the solution behinf Guardium AI.
Anyone interested in make this happen with me?
BY THE WAYIf I remove th eentire "Provider_config" the new model works, nut connecting OpenAI thru the default provider in the system.