feat: add global device queue and readiness tracking

apps/showcase_ui/src/app/utils/tool-formatter.util.spec.ts

@@ -1,4 +1,15 @@
-import { getUniqueGenericTools } from './tool-formatter.util';
+import { cleanErrorMessage, getUniqueGenericTools } from './tool-formatter.util';
+
+describe('cleanErrorMessage', () => {
+  it('removes repeated LLM wrapper labels while preserving the provider reason', () => {
+    expect(cleanErrorMessage('LLM Error: LLM Request Error: 503 model overloaded'))
+      .toBe('503 model overloaded');
+  });
+
+  it('does not present an empty LLM wrapper label as an error reason', () => {
+    expect(cleanErrorMessage('LLM Error:')).toBe('Unknown error');
+  });
+});
 
 describe('getUniqueGenericTools retry aggregation', () => {
   it('groups recoverable LLM retries while preserving every attempt', () => {

apps/showcase_ui/src/app/utils/tool-formatter.util.ts

@@ -1022,7 +1022,11 @@ export function cleanErrorMessage(rawError: any): string {
     .replace(/^Pixel-level validation failed:\s*/i, '')
     .replace(/^Execution error:\s*/i, '')
     .replace(/^ServerError:\s*/i, '')
+    // LLM wrappers may add this prefix more than once. It is context, not the
+    // actual provider reason, and an empty prefix must not be shown as though
+    // it were a useful error message.
+    .replace(/^(?:LLM\s+(?:Request\s+)?Error\s*:\s*)+/i, '')
     .trim();
 
-  return fallback || errorStr;
+  return fallback || 'Unknown error';
 }