No loading indicator during local AI model load — first prompt shows "Agent error"

When starting the OpenClaw Windows node configured to use a local AI model, sending the first prompt while the model is still loading/starting results in an "Agent error" being shown to the user. There is no visual feedback (spinner, "model loading…" status, progress bar, etc.) indicating that the model is still initializing, so the error is confusing — it looks like a failure rather than a transient startup state.

Steps to reproduce:

1. Start OpenClaw Windows node with a local AI model configured.
2. Immediately send a prompt while the model is still loading.
3. Observe "Agent error" with no indication that this is due to the model still starting up.

Expected behavior:

• The UI should show a clear loading/starting state for the local model (e.g. "Model is loading, please wait…") instead of allowing prompts to fail silently with a generic "Agent error".
• Ideally, prompt submission should be disabled/queued until the model is ready, or the error message should explicitly explain the model is still loading and suggest retrying.

Environment: OpenClaw Windows node, local AI model backend.

No source files, tests, or entry points are named. Start by locating the Windows node's local-model initialization and prompt-submission UI, then reproduce the first-prompt case while loading; done means users see an explicit loading state and prompts no longer produce a misleading generic error during initialization.