Fully migrate off Gemini: shared Ollama client (gemma4:26b @ shadow-death), health reports llm backend, drop @google/genai

server.ts

@@ -10,6 +10,7 @@ import { generateMCPToolFromPrompt } from './src/server/aiGenerator.js';
 import { mcpWorkflowManager } from './src/server/mcpWorkflows.js';
 import { mcpSecurityManager } from './src/server/mcpSecurity.js';
 import { SystemLog, ServerStats } from './src/types.js';
+import { OLLAMA_HOST, OLLAMA_MODEL } from './src/server/llmClient.js';
 
 // Inject memoryStore reference into tools module
 setMemoryStoreRef(memoryStore);
@@ -65,7 +66,6 @@ addLog({
 app.get('/api/v1/system/status', (req: Request, res: Response) => {
   const uptimeSeconds = Math.floor((Date.now() - startTime) / 1000);
   const avgLatency = totalRequests > 0 ? Math.round(totalLatencySum / totalRequests) : 0;
-  const apiKey = process.env.GEMINI_API_KEY;
   const servers = mcpRegistry.getServers();
   const activeTools = mcpRegistry.getAllActiveTools();
 
@@ -81,7 +81,9 @@ app.get('/api/v1/system/status', (req: Request, res: Response) => {
     totalMCPServersCount: servers.length,
     avgLatencyMs: avgLatency,
     memoryUsageMb: Math.round(process.memoryUsage().heapUsed / 1024 / 1024),
-    hasGeminiKey: Boolean(apiKey && apiKey !== 'MY_GEMINI_API_KEY'),
+    llmProvider: 'ollama',
+    llmModel: OLLAMA_MODEL,
+    llmHost: OLLAMA_HOST,
   };
 
   res.json(stats);