Don't overwrite default model with blockrun/auto

[achiurizo]

When using the openclaw to handle installation of clawrouter, you can run into an issue where blockrun/auto is set as the model and this creates connection issues with the model as the process is still being setup. This requires manual intervention to get back on track.

By not overriding the model, the openclaw agent can continue the process until completion and instead we prompt the user or agent to switch via /model blockrun/auto

[1bcMax]

Hey @achiurizo, thanks for this PR!

I understand the concern about timing issues during installation. We recently shipped some changes (v0.6.0-v0.6.4) that should address this:

1. waitForProxyHealth() - register() now waits up to 3s for the proxy to be healthy before returning. This ensures the proxy is ready before OpenClaw starts routing requests:

const healthy = await waitForProxyHealth(port);
if (!healthy) {
  api.logger.warn('Proxy health check timed out...');
}

2. Universal free fallback - Even if the wallet is empty, ALL requests now fall back to nvidia/gpt-oss-120b (free model) instead of failing. This means blockrun/auto works immediately without funding.

The design intent is zero-friction onboarding: install → smart routing works immediately (using free model if wallet is empty).

Questions:

• Are you still seeing timing issues with the latest version (0.6.4)?
• If so, can you share the specific error or behavior you're encountering?

If waitForProxyHealth() isn't sufficient, we could extend the timeout or add additional safeguards. Let me know!