feat(agent): integrate Langfuse observability

Add OpenTelemetry-based Langfuse tracing to the AI agent. Every
generateText call now emits spans with tool calls, token usage, and
metadata (userId, conversationId) to Langfuse for monitoring.

apps/api/src/app/agent/agent.service.ts

@@ -544,7 +544,15 @@ If you detect any inconsistencies in the data, flag them clearly to the user.`,
           }
         })
       },
-      maxSteps: 5
+      maxSteps: 5,
+      experimental_telemetry: {
+        isEnabled: true,
+        functionId: 'ghostfolio-agent',
+        metadata: {
+          userId,
+          conversationId: convId
+        }
+      }
     });
 
     // --- Verification Layer ---

apps/api/src/main.ts

@@ -20,6 +20,22 @@ import helmet from 'helmet';
 import { AppModule } from './app/app.module';
 import { environment } from './environments/environment';
 
+// Initialize Langfuse OpenTelemetry tracing (if configured)
+if (process.env.LANGFUSE_SECRET_KEY) {
+  try {
+    const { NodeSDK } = require('@opentelemetry/sdk-node');
+    const { LangfuseSpanProcessor } = require('@langfuse/otel');
+
+    const sdk = new NodeSDK({
+      spanProcessors: [new LangfuseSpanProcessor()]
+    });
+    sdk.start();
+    console.log('Langfuse tracing initialized');
+  } catch (error) {
+    console.warn('Langfuse tracing not available:', error.message);
+  }
+}
+
 async function bootstrap() {
   const configApp = await NestFactory.create(AppModule);
   const configService = configApp.get<ConfigService>(ConfigService);