Issues with Ollama Integration and Model Persistence in AutoGen Studio (Latest Version)

[misterjice]

Hello AutoGen Studio team and community,

I'm encountering several issues with the latest version of AutoGen Studio when trying to integrate Ollama models. These issues did not occur in previous versions, and I'm hoping to get some guidance or solutions to resolve them. Below are the details of the problems I'm facing:

1. Ollama API Configuration Errors
   I'm trying to use Ollama with AutoGen Studio by setting the API base URL to http://localhost:11434/v1, which aligns with Ollama's OpenAI-compatible API endpoint. I've confirmed that the Ollama server is running locally (verified via curl http://localhost:11434/v1), and the model names I'm using match those listed in ollama list. However, I keep encountering errors when setting up models in AutoGen Studio. The errors are not always clear, but they prevent the models from being properly recognized or used.

2. Model Selection Not Persisting Across Agents
   When I create new agents in AutoGen Studio, the model selections I previously configured do not persist. I have to manually reselect the models for each new agent, which is time-consuming.
   I've tried saving the workflow explicitly, but the issue persists. This behavior did not occur in previous versions, where model selections carried over as expected.

3. "Illegal Environment Variable Name" Error
   When attempting to use AutoGen Studio, I frequently encounter the following error:

system illegal environment variable name
I've checked my system's environment variables (using set on Windows and printenv on macOS/Linux) and couldn't identify any obvious issues.
I'm unsure if this is related to AutoGen Studio's configuration files or a broader system issue, but it prevents me from proceeding with my workflows.

4. Unexpected API Key Error Despite Using Ollama
   When working in the team builder, I encounter the following error:

instantiation
Failed to instantiate component: The api_key client option must be set either by passing api_key to the client or by setting the
OPENAI_API_KEY environment variable
Suggestion: Check that the component can be properly instantiated with the given config

This error is unexpected because I'm using Ollama, which does not require an API key. I've tried the following workarounds:
Leaving the API key field blank.
Entering a dummy value (e.g., NULL or ollama).

In both cases, the error persists. I've also checked my environment variables and confirmed that OPENAI_API_KEY is not set, as I'm not using OpenAI.

This issue seems to suggest that AutoGen Studio is incorrectly requiring an API key for Ollama models, which was not the case in previous versions.

Additional Context
Ollama Version: latest
AutoGen Studio Version: latest
Operating System: Windoze 11
Everything worked as expected in previous versions of AutoGen Studio, so this seems to be a regression in the current release.
I've reviewed the AutoGen GitHub issues and discussions but couldn't find a clear solution for these specific problems.

Questions and Requests
Has anyone else encountered similar issues with Ollama integration in the latest version of AutoGen Studio?
Are there known workarounds for the model persistence and API key errors?
Could these issues be related to bugs in the current release, and if so, are there plans to address them in future updates?
If downgrading to a previous version is recommended, which version is known to work reliably with Ollama?

Any help or guidance would be greatly appreciated. I'm happy to provide additional details, logs, or screenshots if needed. Thank you for your time and support!