feat(qwen-live): protocol v7 playback receipts and interactive init — M5

.github/workflows/live-host-release.yml

@@ -346,6 +346,45 @@ jobs:
             gh release create "$FEED_TAG" "${stable_assets[@]}" --title 'Qwen Live Host latest' --notes 'Stable Qwen Live Host installer feed.' --latest=false
           fi
 
+      - name: 'Checkout source for qwen-live npm publish'
+        uses: 'actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10'
+
+      - name: 'Set up Node.js'
+        uses: 'actions/setup-node@49933f5360751b6f8e5e4b6c6f3a8c3c5c5c5c5c'
+        with:
+          node-version-file: '.nvmrc'
+          registry-url: 'https://registry.npmjs.org'
+
+      - name: 'Install dependencies'
+        run: 'npm install --ignore-scripts'
+
+      - name: 'Build qwen-live'
+        run: 'npm run build --workspace @qwen-code/qwen-live'
+
+      - name: 'Publish @qwen-code/qwen-live'
+        working-directory: 'packages/qwen-live'
+        env:
+          NODE_AUTH_TOKEN: '${{ secrets.NPM_TOKEN }}'
+          RELEASE_VERSION: '${{ needs.prepare.outputs.version }}'
+          NPM_TAG: "${{ inputs.prerelease == true && 'preview' || 'latest' }}"
+        run: |
+          set -euo pipefail
+          PACKAGE_NAME="$(node -p "require('./package.json').name")"
+          # Align the package version with the Host release version.
+          npm version "$RELEASE_VERSION" --no-git-tag-version --allow-same-version
+          if npm view "${PACKAGE_NAME}@${RELEASE_VERSION}" version >/dev/null 2>&1; then
+            echo "::notice::${PACKAGE_NAME}@${RELEASE_VERSION} already published; skipping"
+            exit 0
+          fi
+          # Check if the package exists at all — provenance requires an
+          # existing package, so the first publish omits it.
+          if npm view "${PACKAGE_NAME}" version >/dev/null 2>&1; then
+            npm publish --access public --provenance --tag "$NPM_TAG"
+          else
+            echo "::notice::First publish of ${PACKAGE_NAME}; omitting --provenance"
+            npm publish --access public --tag "$NPM_TAG"
+          fi
+
       - name: 'Publish release summary'
         env:
           RELEASE_URL: '${{ steps.release.outputs.url }}'

.gitignore

@@ -145,3 +145,4 @@ tmp/
 
 # Brand build workspaces (created by the desktop-brand-builder skill)
 brand-builds/
+workspace/

integration-tests/qwen-live-harness.ts

@@ -72,7 +72,7 @@ const SERVE_TOKEN = 'qwen-live-e2e-token';
 const LIVE_LISTENING_RE = /qwen-live listening on http:\/\/127\.0\.0\.1:(\d+)/;
 const DISPOSE_GRACE_MS = 10_000;
 const LIVE_HOST_BUNDLE_ID = 'com.alibaba.qwen-code.live-host';
-const LIVE_HOST_PROTOCOL_VERSION = 6;
+const LIVE_HOST_PROTOCOL_VERSION = 7;
 const LIVE_INPUT_AUDIO_EPOCH_BYTES = 8;
 
 // -- small async utilities ----------------------------------------------------

integration-tests/qwen-live-m1-call.test.ts

@@ -88,7 +88,7 @@ describeE2E('qwen-live M1 — end-to-end voice call', () => {
     expect(record['url']).toBe(stack.live.url);
     expect(typeof record['token']).toBe('string');
     expect(String(record['token']).length).toBeGreaterThan(0);
-    expect(record['protocolVersion']).toBe(6);
+    expect(record['protocolVersion']).toBe(7);
     expect(record['pid']).toBe(stack.live.proc.pid);
     expect(String(record['instanceNonce'])).toMatch(/^[A-Za-z0-9_-]{16,256}$/);
   });

packages/cli/src/serve/live/live-host-coordinator.ts

@@ -292,6 +292,18 @@ function parseHostMessage(text: string): LiveHostMessage | undefined {
       };
     }
   }
+  if (
+    value['type'] === 'host.playback_started' &&
+    typeof value['epoch'] === 'number'
+  ) {
+    return { type: 'host.playback_started', epoch: value['epoch'] };
+  }
+  if (
+    value['type'] === 'host.playback_completed' &&
+    typeof value['epoch'] === 'number'
+  ) {
+    return { type: 'host.playback_completed', epoch: value['epoch'] };
+  }
   return undefined;
 }
 
@@ -983,6 +995,15 @@ export class LiveHostCoordinator {
       this.handleShortcutResult(message);
       return;
     }
+    // v7 playback receipts: accepted but not forwarded to the built-in
+    // Live session coordinator (which uses byte estimation). The
+    // standalone qwen-live daemon wires these to its injector.
+    if (
+      message.type === 'host.playback_started' ||
+      message.type === 'host.playback_completed'
+    ) {
+      return;
+    }
     this.handleAction(message);
   }
 

packages/cli/src/serve/live/types.ts

@@ -4,7 +4,7 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 
-export const LIVE_HOST_PROTOCOL_VERSION = 6 as const;
+export const LIVE_HOST_PROTOCOL_VERSION = 7 as const;
 export const LIVE_HOST_BUNDLE_ID = 'com.alibaba.qwen-code.live-host' as const;
 export const LIVE_INPUT_AUDIO_EPOCH_BYTES = 8;
 
@@ -149,12 +149,24 @@ export type LiveHostScreenContextResult =
       error: string;
     };
 
+export interface LiveHostPlaybackStarted {
+  type: 'host.playback_started';
+  epoch: number;
+}
+
+export interface LiveHostPlaybackCompleted {
+  type: 'host.playback_completed';
+  epoch: number;
+}
+
 export type LiveHostMessage =
   | LiveHostHello
   | LiveHostAction
   | LiveHostPong
   | LiveHostShortcutResult
-  | LiveHostScreenContextResult;
+  | LiveHostScreenContextResult
+  | LiveHostPlaybackStarted
+  | LiveHostPlaybackCompleted;
 
 export type LiveDaemonMessage =
   | {

packages/live-host/src/main/__tests__/protocol.test.ts

@@ -142,7 +142,7 @@ describe('Live Host protocol', () => {
       /LIVE_HOST_BUNDLE_ID = '([^']+)'/u,
     )?.[1];
 
-    assert.equal(LIVE_PROTOCOL_VERSION, 6);
+    assert.equal(LIVE_PROTOCOL_VERSION, 7);
     assert.equal(daemonVersion, LIVE_PROTOCOL_VERSION);
     assert.equal(daemonBundleId, LIVE_HOST_BUNDLE_ID);
     assert.equal(Object.values(PROTOCOL_TYPE_PARITY).every(Boolean), true);

packages/live-host/src/main/daemon-connection.ts

@@ -458,6 +458,14 @@ export class LiveDaemonConnection {
     return true;
   }
 
+  sendPlaybackStarted(epoch: number): boolean {
+    return this.sendControl({ type: 'host.playback_started', epoch });
+  }
+
+  sendPlaybackCompleted(epoch: number): boolean {
+    return this.sendControl({ type: 'host.playback_completed', epoch });
+  }
+
   private async captureScreenContext(
     requestId: string,
     epoch: number,

packages/live-host/src/main/index.ts

@@ -575,6 +575,30 @@ function registerIpc(): void {
     });
     publishState();
   });
+  ipcMain.on('live:audio:playback-started', (event, epoch: unknown) => {
+    if (
+      !isTrustedSender(event) ||
+      typeof epoch !== 'number' ||
+      !Number.isSafeInteger(epoch) ||
+      epoch !== daemon.getEpoch()
+    ) {
+      return;
+    }
+    daemon.sendPlaybackStarted(epoch);
+  });
+
+  ipcMain.on('live:audio:playback-completed', (event, epoch: unknown) => {
+    if (
+      !isTrustedSender(event) ||
+      typeof epoch !== 'number' ||
+      !Number.isSafeInteger(epoch) ||
+      epoch !== daemon.getEpoch()
+    ) {
+      return;
+    }
+    daemon.sendPlaybackCompleted(epoch);
+  });
+
   ipcMain.handle('live:set-output-muted', (event, muted: unknown) => {
     if (!isTrustedSender(event) || typeof muted !== 'boolean') return;
     const inputMuted = live.inputMuted ?? false;
@@ -970,12 +994,13 @@ void app.whenReady().then(() => {
     },
     onOutputAudio: (audio) => {
       if (nativeServicesActive && !live.outputMuted) {
-        appendHostAudio(audio, daemon.getEpoch());
+        const epoch = daemon.getEpoch();
+        appendHostAudio(audio, epoch);
         writeLiveDiagnostic('output_frame_received', {
-          epoch: daemon.getEpoch(),
+          epoch,
           bytes: audio.byteLength,
         });
-        sendAudioCommand('live:audio:play', audio);
+        sendAudioCommand('live:audio:play', { audio, epoch });
       }
     },
     onClearOutput: () => {

packages/live-host/src/preload/audio-engine.ts

@@ -56,6 +56,8 @@ export class HostAudioEngine {
       event: string,
       details: AudioDiagnosticDetails,
     ) => void = () => {},
+    private readonly onPlaybackStarted: () => void = () => {},
+    private readonly onPlaybackCompleted: () => void = () => {},
   ) {}
 
   private readonly handleDeviceChange = (): void => {
@@ -255,10 +257,23 @@ export class HostAudioEngine {
         currentGeneration: this.outputGeneration,
         remainingSources: this.outputSources.size,
       });
+      // Only fire completion for a natural end (generation matches);
+      // clearOutput increments generation before stopping sources,
+      // so a stop-triggered onended sees a mismatch and stays silent.
+      if (
+        this.outputSources.size === 0 &&
+        generation === this.outputGeneration
+      ) {
+        this.onPlaybackCompleted();
+      }
     };
+    const wasEmpty = this.outputSources.size === 0;
     this.outputSources.add(source);
     source.start(schedule.startAt);
     this.outputCursor = schedule.endAt;
+    if (wasEmpty) {
+      this.onPlaybackStarted();
+    }
     this.onDiagnostic('output_frame_scheduled', {
       bytes: frame.byteLength,
       generation,

packages/live-host/src/preload/index.ts

@@ -14,6 +14,16 @@ const audio = new HostAudioEngine(
       ipcRenderer.send('live:audio:diagnostic', { event, details });
     }
   },
+  () => {
+    if (currentPlaybackEpoch !== undefined) {
+      ipcRenderer.send('live:audio:playback-started', currentPlaybackEpoch);
+    }
+  },
+  () => {
+    if (currentPlaybackEpoch !== undefined) {
+      ipcRenderer.send('live:audio:playback-completed', currentPlaybackEpoch);
+    }
+  },
 );
 
 const invoke = (channel: string, ...args: unknown[]): Promise<void> =>
@@ -85,14 +95,19 @@ ipcRenderer.on(
 ipcRenderer.on('live:audio:set-output-muted', (_event, muted: boolean) => {
   audio.setOutputMuted(muted);
 });
-ipcRenderer.on('live:audio:play', (_event, frame: Uint8Array) => {
-  void audio.play(frame).catch(() => {
-    audio.clearOutput();
-    ipcRenderer.send('live:audio:output-error', {
-      code: 'audio_output_unavailable',
+let currentPlaybackEpoch: number | undefined;
+ipcRenderer.on(
+  'live:audio:play',
+  (_event, payload: { audio: Uint8Array; epoch: number }) => {
+    currentPlaybackEpoch = payload.epoch;
+    void audio.play(payload.audio).catch(() => {
+      audio.clearOutput();
+      ipcRenderer.send('live:audio:output-error', {
+        code: 'audio_output_unavailable',
+      });
     });
-  });
-});
+  },
+);
 ipcRenderer.on('live:audio:clear', () => audio.clearOutput());
 
 let lastPointerInteractive = false;

packages/live-host/src/shared/protocol.ts

@@ -1,4 +1,4 @@
-export const LIVE_PROTOCOL_VERSION = 6;
+export const LIVE_PROTOCOL_VERSION = 7;
 export const LIVE_HOST_BUNDLE_ID = 'com.alibaba.qwen-code.live-host';
 export const MAX_CONTROL_FRAME_BYTES = 64 * 1024;
 export const MAX_INPUT_AUDIO_FRAME_BYTES = 64 * 1024;
@@ -112,7 +112,9 @@ export type HostControlMessage =
       requestId: string;
       success: false;
       error: string;
-    };
+    }
+  | { type: 'host.playback_started'; epoch: number }
+  | { type: 'host.playback_completed'; epoch: number };
 
 export type DaemonControlMessage =
   | {