fix: warmup ACP subprocess eagerly to avoid 30s cold-start on status

The pool refactor made ACP client creation lazy, causing the first
/llm/health call to spawn + initialize + prompt Gemini in one request
(30s+). Now warmup() eagerly starts the subprocess on mcplocal boot.
Also fetch models in parallel with LLM health check.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/mcplocal/src/llm-config.ts

@@ -28,7 +28,9 @@ export async function createProviderFromConfig(
       const cfg: GeminiAcpConfig = {};
       if (config.binaryPath) cfg.binaryPath = config.binaryPath;
       if (config.model) cfg.defaultModel = config.model;
-      registry.register(new GeminiAcpProvider(cfg));
+      const provider = new GeminiAcpProvider(cfg);
+      provider.warmup();
+      registry.register(provider);
       break;
     }
 

src/mcplocal/src/providers/gemini-acp.ts

@@ -83,6 +83,18 @@ export class GeminiAcpProvider implements LlmProvider {
     this.pool.clear();
   }
 
+  /**
+   * Eagerly spawn the default model's ACP subprocess so it's ready
+   * for the first request (avoids 30s cold-start on health checks).
+   */
+  warmup(): void {
+    const entry = this.getOrCreateEntry(this.defaultModel);
+    // Fire-and-forget: start the subprocess initialization in the background
+    entry.client.ensureReady().catch(() => {
+      // Ignore errors — next request will retry
+    });
+  }
+
   /** Number of active pool entries (for testing). */
   get poolSize(): number {
     return this.pool.size;