feat: add session updates and message dialogs

frontend/src/hooks/use-voice-preview.ts

@@ -18,6 +18,7 @@ export type VoicePreviewStatus = "idle" | "connecting" | "connected" | "failed";
 export type NetworkQuality = "unknown" | "good" | "fair" | "poor";
 
 const NETWORK_SAMPLE_INTERVAL_MS = 2_000;
+const SESSION_UPDATE_TIMEOUT_MS = 10_000;
 
 type NetworkStatsSample = {
   packetsSent: number;
@@ -46,7 +47,7 @@ export type ChatMessage = {
 };
 
 type AppMessage = Record<string, unknown> & { type?: string };
-type DynamicVariableValue = string | number | boolean;
+export type DynamicVariableValue = string | number | boolean;
 export type UserInputPart =
   | { type: "input_text"; text: string }
   | {
@@ -57,6 +58,15 @@ export type UserInputResult = {
   inputId: string;
   status: "accepted";
 };
+export type SessionUpdateRequest = {
+  dynamicVariables: Record<string, DynamicVariableValue>;
+};
+export type SessionUpdateResult = {
+  updateId: string;
+  status: "accepted";
+  changed: string[];
+  dynamicVariables: Record<string, DynamicVariableValue>;
+};
 export type ClientToolHandler = (
   argumentsValue: Record<string, unknown>,
 ) => unknown | Promise<unknown>;
@@ -67,6 +77,12 @@ type PendingUserInput = {
   timeout: number;
 };
 
+type PendingSessionUpdate = {
+  resolve: (result: SessionUpdateResult) => void;
+  reject: (error: Error) => void;
+  timeout: number;
+};
+
 function publicVariableSnapshot(
   value: unknown,
 ): Record<string, DynamicVariableValue> {
@@ -142,6 +158,13 @@ function newInputId(): string {
   return `input_${Date.now()}_${Math.random().toString(16).slice(2)}`;
 }
 
+function newUpdateId(): string {
+  if (typeof crypto !== "undefined" && "randomUUID" in crypto) {
+    return `update_${crypto.randomUUID()}`;
+  }
+  return `update_${Date.now()}_${Math.random().toString(16).slice(2)}`;
+}
+
 function readNetworkMetrics(report: RTCStatsReport): NetworkMetrics | null {
   const outbound: Partial<Record<"audio" | "video", number>> = {};
   const lost: Partial<Record<"audio" | "video", number>> = {};
@@ -246,6 +269,9 @@ export function useVoicePreview(
     new Map<string, { timestamp: string }>(),
   );
   const pendingUserInputsRef = useRef(new Map<string, PendingUserInput>());
+  const pendingSessionUpdatesRef = useRef(
+    new Map<string, PendingSessionUpdate>(),
+  );
   const clientToolHandlersRef = useRef(new Map<string, ClientToolHandler>());
   const endedByServerRef = useRef(false);
   const selectedDeviceIdRef = useRef("");
@@ -306,6 +332,11 @@ export function useVoicePreview(
       reject(new Error("连接已关闭，用户输入未完成"));
     });
     pendingUserInputsRef.current.clear();
+    pendingSessionUpdatesRef.current.forEach(({ reject, timeout }) => {
+      window.clearTimeout(timeout);
+      reject(new Error("连接已关闭，会话状态更新未完成"));
+    });
+    pendingSessionUpdatesRef.current.clear();
     networkStatsRef.current = null;
     onNodeActiveRef.current?.(null);
   }, []);
@@ -405,6 +436,33 @@ export function useVoicePreview(
           ),
         );
       }
+    } else if (
+      msg.type === "session-update-result" &&
+      typeof msg.update_id === "string"
+    ) {
+      const pending = pendingSessionUpdatesRef.current.get(msg.update_id);
+      if (!pending) return;
+      pendingSessionUpdatesRef.current.delete(msg.update_id);
+      window.clearTimeout(pending.timeout);
+      if (msg.status === "accepted") {
+        const dynamicVariables = publicVariableSnapshot(msg.dynamic_variables);
+        const changed = Array.isArray(msg.changed)
+          ? msg.changed.filter((name): name is string => typeof name === "string")
+          : [];
+        setSessionVariables(dynamicVariables);
+        pending.resolve({
+          updateId: msg.update_id,
+          status: "accepted",
+          changed,
+          dynamicVariables,
+        });
+      } else {
+        pending.reject(
+          new Error(
+            typeof msg.message === "string" ? msg.message : "会话状态更新失败",
+          ),
+        );
+      }
     } else if (
       msg.type === "assistant-text-start" &&
       typeof msg.turn_id === "string"
@@ -754,6 +812,45 @@ export function useVoicePreview(
     [],
   );
 
+  const updateSession = useCallback(
+    ({
+      dynamicVariables,
+    }: SessionUpdateRequest): Promise<SessionUpdateResult> => {
+      if (Object.keys(dynamicVariables).length === 0) {
+        return Promise.reject(new Error("至少需要更新一个动态变量"));
+      }
+      const transport = transportRef.current;
+      if (!transport || transport.state !== "connected") {
+        return Promise.reject(new Error("当前未连接语音服务"));
+      }
+      const updateId = newUpdateId();
+      return new Promise<SessionUpdateResult>((resolve, reject) => {
+        const timeout = window.setTimeout(() => {
+          pendingSessionUpdatesRef.current.delete(updateId);
+          reject(new Error("等待会话状态更新结果超时"));
+        }, SESSION_UPDATE_TIMEOUT_MS);
+        pendingSessionUpdatesRef.current.set(updateId, {
+          resolve,
+          reject,
+          timeout,
+        });
+        try {
+          transport.sendAppMessage({
+            type: "session-update",
+            schema_version: 1,
+            update_id: updateId,
+            dynamic_variables: dynamicVariables,
+          });
+        } catch (sendError) {
+          window.clearTimeout(timeout);
+          pendingSessionUpdatesRef.current.delete(updateId);
+          reject(new Error(errorMessage(sendError, "发送会话状态更新失败")));
+        }
+      });
+    },
+    [],
+  );
+
   const registerClientTool = useCallback(
     (functionName: string, handler: ClientToolHandler): (() => void) => {
       clientToolHandlersRef.current.set(functionName, handler);
@@ -789,6 +886,7 @@ export function useVoicePreview(
     supportsOutputSelection,
     sendText,
     sendUserInput,
+    updateSession,
     registerClientTool,
     connect,
     replaceVideoStream,