fix: attach Google Meet realtime bridge

docs/plugins/google-meet.md

@@ -355,6 +355,8 @@ Defaults:
 - `realtime.toolPolicy: "safe-read-only"`
 - `realtime.instructions`: brief spoken replies, with
   `openclaw_agent_consult` for deeper answers
+- `realtime.introMessage`: short spoken readiness check when the realtime bridge
+  connects; set it to `""` to join silently
 
 Optional overrides:
 
@@ -371,6 +373,7 @@ Optional overrides:
   },
   realtime: {
     toolPolicy: "owner",
+    introMessage: "Say exactly: I'm here.",
   },
 }
 ```
@@ -409,7 +412,16 @@ VM. In both cases the realtime model and `openclaw_agent_consult` run on the
 Gateway host, so model credentials stay there.
 
 Use `action: "status"` to list active sessions or inspect a session ID. Use
-`action: "leave"` to mark a session ended.
+`action: "speak"` with `sessionId` and `message` to make the realtime agent
+speak immediately. Use `action: "leave"` to mark a session ended.
+
+```json
+{
+  "action": "speak",
+  "sessionId": "meet_...",
+  "message": "Say exactly: I'm here and listening."
+}
+```
 
 ## Realtime agent consult
 
@@ -434,6 +446,12 @@ voice session. The voice model can then speak that answer back into the meeting.
 The consult session key is scoped per Meet session, so follow-up consult calls
 can reuse prior consult context during the same meeting.
 
+To force a spoken readiness check after Chrome has fully joined the call:
+
+```bash
+openclaw googlemeet speak meet_... "Say exactly: I'm here and listening."
+```
+
 ## Notes
 
 Google Meet's official media API is receive-oriented, so speaking into a Meet
@@ -453,9 +471,9 @@ For clean duplex audio, route Meet output and Meet microphone through separate
 virtual devices or a Loopback-style virtual device graph. A single shared
 BlackHole device can echo other participants back into the call.
 
-`googlemeet leave` stops the command-pair realtime audio bridge for Chrome
-sessions. For Twilio sessions delegated through the Voice Call plugin, it also
-hangs up the underlying voice call.
+`googlemeet speak` triggers the active realtime audio bridge for a Chrome
+session. `googlemeet leave` stops that bridge. For Twilio sessions delegated
+through the Voice Call plugin, `leave` also hangs up the underlying voice call.
 
 ## Related
 

extensions/google-meet/index.test.ts

@@ -205,6 +205,7 @@ describe("google-meet plugin", () => {
       voiceCall: { enabled: true, requestTimeoutMs: 30000, dtmfDelayMs: 2500 },
       realtime: {
         provider: "openai",
+        introMessage: "Say exactly: I'm here and listening.",
         toolPolicy: "safe-read-only",
       },
       oauth: {},
@@ -284,7 +285,7 @@ describe("google-meet plugin", () => {
       properties: {
         action: {
           type: "string",
-          enum: ["join", "status", "setup_status", "resolve_space", "preflight", "leave"],
+          enum: ["join", "status", "setup_status", "resolve_space", "preflight", "leave", "speak"],
         },
         transport: { type: "string", enum: ["chrome", "chrome-node", "twilio"] },
         mode: { type: "string", enum: ["realtime", "transcribe"] },
@@ -520,11 +521,16 @@ describe("google-meet plugin", () => {
   });
 
   it("joins Chrome on a paired node without local Chrome or BlackHole", async () => {
-    const { methods, nodesList, nodesInvoke } = setup({
-      defaultTransport: "chrome-node",
-      defaultMode: "transcribe",
-      chromeNode: { node: "parallels-macos" },
-    });
+    const { methods, nodesList, nodesInvoke } = setup(
+      {
+        defaultTransport: "chrome-node",
+        defaultMode: "transcribe",
+        chromeNode: { node: "parallels-macos" },
+      },
+      {
+        nodesInvokeResult: { payload: { launched: true } },
+      },
+    );
     const handler = methods.get("googlemeet.join") as
       | ((ctx: {
           params: Record<string, unknown>;
@@ -669,6 +675,7 @@ describe("google-meet plugin", () => {
             name: string;
             args: unknown;
           }) => void;
+          onReady?: () => void;
           tools?: unknown[];
         }
       | undefined;
@@ -680,6 +687,7 @@ describe("google-meet plugin", () => {
       submitToolResult: vi.fn(),
       acknowledgeMark: vi.fn(),
       close: vi.fn(),
+      triggerGreeting: vi.fn(),
       isConnected: vi.fn(() => true),
     };
     const provider: RealtimeVoiceProviderPlugin = {
@@ -756,6 +764,7 @@ describe("google-meet plugin", () => {
     inputStdout.write(Buffer.from([1, 2, 3]));
     callbacks?.onAudio(Buffer.from([4, 5]));
     callbacks?.onMark?.("mark-1");
+    callbacks?.onReady?.();
     callbacks?.onToolCall?.({
       itemId: "item-1",
       callId: "tool-call-1",
@@ -772,6 +781,9 @@ describe("google-meet plugin", () => {
     expect(sendAudio).toHaveBeenCalledWith(Buffer.from([1, 2, 3]));
     expect(outputStdinWrites).toEqual([Buffer.from([4, 5])]);
     expect(bridge.acknowledgeMark).toHaveBeenCalled();
+    expect(bridge.triggerGreeting).toHaveBeenCalledWith("Say exactly: I'm here and listening.");
+    handle.speak("Say exactly: hello from the meeting.");
+    expect(bridge.triggerGreeting).toHaveBeenLastCalledWith("Say exactly: hello from the meeting.");
     expect(callbacks).toMatchObject({
       tools: [
         expect.objectContaining({
@@ -808,6 +820,7 @@ describe("google-meet plugin", () => {
             name: string;
             args: unknown;
           }) => void;
+          onReady?: () => void;
           tools?: unknown[];
         }
       | undefined;
@@ -819,6 +832,7 @@ describe("google-meet plugin", () => {
       submitToolResult: vi.fn(),
       acknowledgeMark: vi.fn(),
       close: vi.fn(),
+      triggerGreeting: vi.fn(),
       isConnected: vi.fn(() => true),
     };
     const provider: RealtimeVoiceProviderPlugin = {
@@ -879,6 +893,7 @@ describe("google-meet plugin", () => {
     });
 
     callbacks?.onAudio(Buffer.from([1, 2, 3]));
+    callbacks?.onReady?.();
     callbacks?.onToolCall?.({
       itemId: "item-1",
       callId: "tool-call-1",
@@ -907,6 +922,9 @@ describe("google-meet plugin", () => {
         text: "Use the launch update.",
       });
     });
+    expect(bridge.triggerGreeting).toHaveBeenCalledWith("Say exactly: I'm here and listening.");
+    handle.speak("Say exactly: hello from the node.");
+    expect(bridge.triggerGreeting).toHaveBeenLastCalledWith("Say exactly: hello from the node.");
     expect(callbacks).toMatchObject({
       tools: [
         expect.objectContaining({

extensions/google-meet/index.ts

@@ -88,6 +88,10 @@ const googleMeetConfigSchema = {
     },
     "realtime.model": { label: "Realtime Model", advanced: true },
     "realtime.instructions": { label: "Realtime Instructions", advanced: true },
+    "realtime.introMessage": {
+      label: "Realtime Intro Message",
+      help: "Spoken once when the realtime bridge is ready. Set to an empty string to join silently.",
+    },
     "realtime.toolPolicy": {
       label: "Realtime Tool Policy",
       help: "Safe read-only tools are available by default; owner requests can unlock broader tools.",
@@ -111,7 +115,7 @@ const googleMeetConfigSchema = {
 
 const GoogleMeetToolSchema = Type.Object({
   action: Type.String({
-    enum: ["join", "status", "setup_status", "resolve_space", "preflight", "leave"],
+    enum: ["join", "status", "setup_status", "resolve_space", "preflight", "leave", "speak"],
     description: "Google Meet action to run",
   }),
   url: Type.Optional(Type.String({ description: "Explicit https://meet.google.com/... URL" })),
@@ -123,6 +127,7 @@ const GoogleMeetToolSchema = Type.Object({
   pin: Type.Optional(Type.String({ description: "Meet phone PIN for Twilio" })),
   dtmfSequence: Type.Optional(Type.String({ description: "Explicit DTMF sequence for Twilio" })),
   sessionId: Type.Optional(Type.String({ description: "Meet session ID" })),
+  message: Type.Optional(Type.String({ description: "Realtime instructions to speak now" })),
   meeting: Type.Optional(Type.String({ description: "Meet URL, meeting code, or spaces/{id}" })),
   accessToken: Type.Optional(Type.String({ description: "Access token override" })),
   refreshToken: Type.Optional(Type.String({ description: "Refresh token override" })),
@@ -265,6 +270,23 @@ export default definePluginEntry({
       },
     );
 
+    api.registerGatewayMethod(
+      "googlemeet.speak",
+      async ({ params, respond }: GatewayRequestHandlerOptions) => {
+        try {
+          const sessionId = normalizeOptionalString(params?.sessionId);
+          if (!sessionId) {
+            respond(false, { error: "sessionId required" });
+            return;
+          }
+          const rt = await ensureRuntime();
+          respond(true, rt.speak(sessionId, normalizeOptionalString(params?.message)));
+        } catch (err) {
+          sendError(respond, err);
+        }
+      },
+    );
+
     api.registerTool({
       name: "google_meet",
       label: "Google Meet",
@@ -318,6 +340,14 @@ export default definePluginEntry({
               }
               return json(await rt.leave(sessionId));
             }
+            case "speak": {
+              const rt = await ensureRuntime();
+              const sessionId = normalizeOptionalString(raw.sessionId);
+              if (!sessionId) {
+                throw new Error("sessionId required");
+              }
+              return json(rt.speak(sessionId, normalizeOptionalString(raw.message)));
+            }
             default:
               throw new Error("unknown google_meet action");
           }

extensions/google-meet/openclaw.plugin.json

@@ -108,6 +108,10 @@
       "label": "Realtime Instructions",
       "advanced": true
     },
+    "realtime.introMessage": {
+      "label": "Realtime Intro Message",
+      "help": "Spoken once when the realtime bridge is ready. Set to an empty string to join silently."
+    },
     "realtime.toolPolicy": {
       "label": "Realtime Tool Policy",
       "help": "Safe read-only tools are available by default; owner requests can unlock broader tools.",
@@ -312,6 +316,10 @@
             "type": "string",
             "default": "You are joining a private Google Meet as an OpenClaw agent. Keep spoken replies brief and natural. When a question needs deeper reasoning, current information, or tools, call openclaw_agent_consult before answering."
           },
+          "introMessage": {
+            "type": "string",
+            "default": "Say exactly: I'm here and listening."
+          },
           "toolPolicy": {
             "type": "string",
             "enum": ["safe-read-only", "owner", "none"],

extensions/google-meet/src/cli.ts

@@ -304,4 +304,20 @@ export function registerGoogleMeetCli(params: {
       }
       writeStdoutLine("left %s", sessionId);
     });
+
+  root
+    .command("speak")
+    .argument("<session-id>", "Meet session ID")
+    .argument("[message]", "Realtime instructions to speak now")
+    .action(async (sessionId: string, message?: string) => {
+      const rt = await params.ensureRuntime();
+      const result = rt.speak(sessionId, message);
+      if (!result.found) {
+        throw new Error("session not found");
+      }
+      if (!result.spoken) {
+        throw new Error("session has no active realtime audio bridge");
+      }
+      writeStdoutLine("speaking on %s", sessionId);
+    });
 }

extensions/google-meet/src/config.ts

@@ -48,6 +48,7 @@ export type GoogleMeetConfig = {
     provider?: string;
     model?: string;
     instructions?: string;
+    introMessage?: string;
     toolPolicy: GoogleMeetToolPolicy;
     providers: Record<string, Record<string, unknown>>;
   };
@@ -99,6 +100,7 @@ export const DEFAULT_GOOGLE_MEET_AUDIO_OUTPUT_COMMAND = [
 ] as const;
 
 export const DEFAULT_GOOGLE_MEET_REALTIME_INSTRUCTIONS = `You are joining a private Google Meet as an OpenClaw agent. Keep spoken replies brief and natural. When a question needs deeper reasoning, current information, or tools, call ${REALTIME_VOICE_AGENT_CONSULT_TOOL_NAME} before answering.`;
+export const DEFAULT_GOOGLE_MEET_REALTIME_INTRO_MESSAGE = "Say exactly: I'm here and listening.";
 
 export const DEFAULT_GOOGLE_MEET_CONFIG: GoogleMeetConfig = {
   enabled: true,
@@ -125,6 +127,7 @@ export const DEFAULT_GOOGLE_MEET_CONFIG: GoogleMeetConfig = {
   realtime: {
     provider: "openai",
     instructions: DEFAULT_GOOGLE_MEET_REALTIME_INSTRUCTIONS,
+    introMessage: DEFAULT_GOOGLE_MEET_REALTIME_INTRO_MESSAGE,
     toolPolicy: "safe-read-only",
     providers: {},
   },
@@ -339,6 +342,9 @@ export function resolveGoogleMeetConfigWithEnv(
       instructions:
         normalizeOptionalString(realtime.instructions) ??
         DEFAULT_GOOGLE_MEET_CONFIG.realtime.instructions,
+      introMessage:
+        normalizeOptionalString(realtime.introMessage) ??
+        DEFAULT_GOOGLE_MEET_CONFIG.realtime.introMessage,
       toolPolicy: resolveToolPolicy(
         realtime.toolPolicy,
         DEFAULT_GOOGLE_MEET_CONFIG.realtime.toolPolicy,

extensions/google-meet/src/realtime-node.ts

@@ -19,6 +19,7 @@ export type ChromeNodeRealtimeAudioBridgeHandle = {
   providerId: string;
   nodeId: string;
   bridgeId: string;
+  speak: (instructions?: string) => void;
   stop: () => Promise<void>;
 };
 
@@ -81,6 +82,8 @@ export async function startNodeRealtimeAudioBridge(params: {
     provider: resolved.provider,
     providerConfig: resolved.providerConfig,
     instructions: params.config.realtime.instructions,
+    initialGreetingInstructions: params.config.realtime.introMessage,
+    triggerGreetingOnReady: Boolean(params.config.realtime.introMessage),
     markStrategy: "ack-immediately",
     tools: resolveGoogleMeetRealtimeTools(params.config.realtime.toolPolicy),
     audioSink: {
@@ -188,6 +191,9 @@ export async function startNodeRealtimeAudioBridge(params: {
     providerId: resolved.provider.id,
     nodeId: params.nodeId,
     bridgeId: params.bridgeId,
+    speak: (instructions) => {
+      bridge?.triggerGreeting(instructions);
+    },
     stop,
   };
 }

extensions/google-meet/src/realtime.ts

@@ -41,6 +41,7 @@ export type ChromeRealtimeAudioBridgeHandle = {
   providerId: string;
   inputCommand: string[];
   outputCommand: string[];
+  speak: (instructions?: string) => void;
   stop: () => Promise<void>;
 };
 
@@ -148,6 +149,8 @@ export async function startCommandRealtimeAudioBridge(params: {
     provider: resolved.provider,
     providerConfig: resolved.providerConfig,
     instructions: params.config.realtime.instructions,
+    initialGreetingInstructions: params.config.realtime.introMessage,
+    triggerGreetingOnReady: Boolean(params.config.realtime.introMessage),
     markStrategy: "ack-immediately",
     tools: resolveGoogleMeetRealtimeTools(params.config.realtime.toolPolicy),
     audioSink: {
@@ -210,6 +213,9 @@ export async function startCommandRealtimeAudioBridge(params: {
     providerId: resolved.provider.id,
     inputCommand: params.inputCommand,
     outputCommand: params.outputCommand,
+    speak: (instructions) => {
+      bridge?.triggerGreeting(instructions);
+    },
     stop,
   };
 }

extensions/google-meet/src/runtime.ts

@@ -49,6 +49,7 @@ function resolveMode(input: GoogleMeetMode | undefined, config: GoogleMeetConfig
 export class GoogleMeetRuntime {
   readonly #sessions = new Map<string, GoogleMeetSession>();
   readonly #sessionStops = new Map<string, () => Promise<void>>();
+  readonly #sessionSpeakers = new Map<string, (instructions?: string) => void>();
 
   constructor(
     private readonly params: {
@@ -151,6 +152,7 @@ export class GoogleMeetRuntime {
           result.audioBridge?.type === "node-command-pair"
         ) {
           this.#sessionStops.set(session.id, result.audioBridge.stop);
+          this.#sessionSpeakers.set(session.id, result.audioBridge.speak);
         }
         session.notes.push(
           result.audioBridge
@@ -215,10 +217,28 @@ export class GoogleMeetRuntime {
     const stop = this.#sessionStops.get(sessionId);
     if (stop) {
       this.#sessionStops.delete(sessionId);
+      this.#sessionSpeakers.delete(sessionId);
       await stop();
     }
     session.state = "ended";
     session.updatedAt = nowIso();
     return { found: true, session };
   }
+
+  speak(
+    sessionId: string,
+    instructions?: string,
+  ): { found: boolean; spoken: boolean; session?: GoogleMeetSession } {
+    const session = this.#sessions.get(sessionId);
+    if (!session) {
+      return { found: false, spoken: false };
+    }
+    const speak = this.#sessionSpeakers.get(sessionId);
+    if (!speak || session.state !== "active") {
+      return { found: true, spoken: false, session };
+    }
+    speak(instructions || this.params.config.realtime.introMessage);
+    session.updatedAt = nowIso();
+    return { found: true, spoken: true, session };
+  }
 }

extensions/google-meet/src/transports/chrome.ts

@@ -201,10 +201,14 @@ function parseNodeStartResult(raw: unknown): {
   bridgeId?: string;
   audioBridge?: { type?: string };
 } {
-  if (!raw || typeof raw !== "object") {
+  const value =
+    raw && typeof raw === "object" && "payload" in raw
+      ? (raw as { payload?: unknown }).payload
+      : raw;
+  if (!value || typeof value !== "object") {
     throw new Error("Google Meet node returned an invalid start result.");
   }
-  return raw as {
+  return value as {
     launched?: boolean;
     bridgeId?: string;
     audioBridge?: { type?: string };