refactor(desktop): split kernel_commands/pipeline_commands into modules, add SaaS client libs and gateway modules

Split monolithic kernel_commands.rs (2185 lines) and pipeline_commands.rs (1391 lines)
into focused sub-modules under kernel_commands/ and pipeline_commands/ directories.
Add gateway module (commands, config, io, runtime), health_check, and 15 new
TypeScript client libraries for SaaS relay, auth, admin, telemetry, and kernel
sub-systems (a2a, agent, chat, hands, skills, triggers).

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

desktop/src/store/connectionStore.ts

@@ -463,8 +463,68 @@ export const useConnectionStore = create<ConnectionStore>((set, get) => {
           }
 
           if (!saasDegraded) {
-            set({ connectionState: 'connected', gatewayVersion: 'saas-relay' });
-            log.debug('Connected to SaaS relay');
+            // === SaaS Relay via Kernel ===
+            // Route LLM calls through SaaS relay (Key Pool) while keeping
+            // agent management local via KernelClient.
+            // baseUrl = saasUrl + /api/v1/relay → kernel appends /chat/completions
+            // apiKey = SaaS JWT token → sent as Authorization: Bearer <jwt>
+            if (isTauriRuntime()) {
+              if (!session.token) {
+                throw new Error('SaaS 中转模式需要认证令牌，请重新登录 SaaS 平台');
+              }
+
+              const kernelClient = getKernelClient();
+
+              // Fetch available models from SaaS relay
+              let models: Array<{ id: string }>;
+              try {
+                models = await saasClient.listModels();
+              } catch {
+                throw new Error('无法获取可用模型列表，请确认管理后台已配置 Provider 和模型');
+              }
+
+              if (models.length === 0) {
+                throw new Error('SaaS 平台没有可用模型，请先在管理后台配置 Provider 和模型');
+              }
+
+              // Use first available model (TODO: let user choose preferred model)
+              const relayModel = models[0];
+
+              kernelClient.setConfig({
+                provider: 'custom',
+                model: relayModel.id,
+                apiKey: session.token,
+                baseUrl: `${session.saasUrl}/api/v1/relay`,
+                apiProtocol: 'openai',
+              });
+
+              kernelClient.onStateChange = (state: ConnectionState) => {
+                set({ connectionState: state });
+              };
+
+              kernelClient.onLog = (level: string, message: string) => {
+                set((s) => ({
+                  logs: [...s.logs.slice(-99), { timestamp: Date.now(), level, message }],
+                }));
+              };
+
+              set({ client: kernelClient });
+
+              const { initializeStores } = await import('./index');
+              initializeStores();
+
+              await kernelClient.connect();
+
+              set({ gatewayVersion: 'saas-relay', connectionState: 'connected' });
+              log.debug('Connected via SaaS relay (kernel backend):', {
+                model: relayModel.id,
+                baseUrl: `${session.saasUrl}/api/v1/relay`,
+              });
+            } else {
+              // Non-Tauri (browser) — simple connected state without kernel
+              set({ connectionState: 'connected', gatewayVersion: 'saas-relay' });
+              log.debug('Connected to SaaS relay (browser mode)');
+            }
             return;
           }
           // Fall through to Tauri Kernel / Gateway mode