Add WSO2 Model Provider as an LLM connection option

[RadCod3]

Current Limitation

Currently in MI users must BYOK for all the available LLM Connection options. In BI, if a user is logged into BI Copilot, they can add and use a "WSO2 Default Model Provider" which re-uses the same LLM Connection used for the copilot.

The user flow for this in BI is as follows,

1. Log into BI Copilot.
2. Click Cmd + Shift + P and invoke the command Ballerina: Configure default WSO2 model provider.

• This creates a Config.toml file with the following structure

[ballerina.ai.wso2ProviderConfig]
serviceUrl = "https://copilotendpoint..."
accessToken = "asdf...."

3. Add an AI Chat Agent artifact which by default uses the WSO2 model provider.
4. Run the program.

Suggested Improvement

Similar to this, we can add an option like WSO2_AI in the LLM Connection options and use the same auth used in the MI Copilot.

Version

3.1.2

[RadCod3]

Update: 27/01/2026

• WSO2_AI is shown as an LLM connection option when using Chat, RagChat and Agent mediators.
• The option is greyed out if a user is not logged into MI Copilot.

• When adding WSO2_AI we do not show any configuration fields in the UI other than connection name.

• Under the hood however, we add the following values to config.properties after user clicks Add

WSO2_AI_SERVICE_URL:string
WSO2_AI_ACCESS_TOKEN:string

• We do not add the access token or service url to the .env file and instead dynamically inject them to the environment when a user runs the integration via the extension.
• This was done so that user need not worry about the access token expiring, as before injecting we refresh the token if needed.
• When deploying however, user must set the necessary environment variables, which can be added to the .env of the project by invoking the vscode command MI: Configure Default WSO2 Model Provider
• Once added, in the UIs for Chat, Rag Chat and Agent, the Model Name field is hidden if the user has selected WSO2_AI is the LLM connection.

• Tested with the scenarios under https://mi.docs.wso2.com/en/latest/get-started/build-first-ai-integration/

• FE PR: Add support for WSO2_AI model provider in MI vscode-extensions#1322

• Connector PR: Add WSO2 AI model provider option wso2-extensions/mi-module-generative-ai#41