Override the recommended AI assistant configuration

If your administrator configured a recommended AI assistant configuration for Che, replace it with your own Lola module. Your module is installed into your workspaces at startup in place of the recommended one.

Your configuration is stored in a Secret in your namespace. It applies to every workspace in the namespace and takes precedence over the recommended configuration.

Prerequisites
Procedure

  1. Create the Secret in your namespace:

    apiVersion: v1
kind: Secret
metadata:
  name: lola-config
  labels:
    controller.devfile.io/mount-to-devworkspace: "true"
    controller.devfile.io/watch-secret: "true"
  annotations:
    controller.devfile.io/mount-as: env
stringData:
  LOLA_REPO_URL: https://github.com/<user>/<module-repo>.git
  LOLA_REPO_REF: main
  LOLA_AGENTS: claude-code

    • Replace the URL and the ref with the values of your module.

    • Remove LOLA_REPO_REF to install the default branch of the repository.

    • A ref is used only with the URL of the same pair. The recommended ref does not apply to your module.

    • LOLA_AGENTS selects the AI assistants: one name, a comma-separated list such as claude-code,opencode, or all. Remove the key to leave the choice to the administrator, or to Claude Code when the administrator does not set it either.

    • A ConfigMap with the same keys works when the URL carries no credentials. Use a Secret when it does.

    • If the lola-config Secret already exists in your namespace, add these keys to it instead of creating another Secret, keep its other keys, and remove LOLA_DISABLED, or set it to anything other than true: while it is true, the init container stays off whatever else the Secret contains. See Disable the AI assistant configuration initialization.

Creating the Secret restarts the running workspaces in your namespace. To avoid that, add the controller.devfile.io/mount-on-start: "true" annotation: the Secret is then mounted at the next workspace start, which is soon enough, because the init container reads it only when the pod starts. See Mounting Secrets.

  1. Restart your workspace. The init container removes the content that it installed from the recommended module and installs your module.

Verification

  • In the workspace, /home/user/.lola/init-status reports origin=personal. The mode is switch when the recommended module was installed before, and install on a workspace that never ran the init container with a module:

    ok <timestamp> <module-name> origin=personal mode=switch warnings=0 agents=<assistant-list> url=<module-url>

  • The content of your module is installed in the directories of each selected assistant. For Claude Code, for example, /home/user/.claude/skills and /home/user/.claude/agents.

  • The skills, commands, and agents of the recommended module are gone from the assistants of the new set. When the new set leaves out Claude Code, MCP servers that the previous configuration added to /home/user/.claude.json can stay there.

Return to the recommended configuration

Remove LOLA_REPO_URL and LOLA_REPO_REF from the Secret, or from the ConfigMap that supplies them, and restart the workspace. Delete the whole resource only when it holds no key you need. Keep the other keys: LOLA_AGENTS still selects the assistants for the recommended module. The next start installs the recommended module again.

The init container protects your own files. It changes only the content that Lola installs, and it removes only the module that it installed itself: modules you add to /home/user/.lola/modules by hand stay. A hand-made MCP server that reuses the name of a managed server is the exception.
Additional resources