chore: deploy Laya Tetris f4a43d5

deploy-sha.txt

@@ -1 +1 @@
-7a68b4693c1726fff95d3c23e5c1c09157dd9459
+f4a43d5b88afaecb9378ca78f0cd8f2f7916250e

flutter_bootstrap.js

@@ -42,6 +42,6 @@ _flutter.buildConfig = {"engineRevision":"5d531788691ec3404cac0cee66ead4007b1773
 
 _flutter.loader.load({
   serviceWorkerSettings: {
-    serviceWorkerVersion: "2668202796" /* Flutter's service worker is deprecated and will be removed in a future Flutter release. */
+    serviceWorkerVersion: "3715908120" /* Flutter's service worker is deprecated and will be removed in a future Flutter release. */
   }
 });

webgpu_bridge/llama_webgpu_bridge.js

@@ -5189,7 +5189,15 @@ var LlamaWebGpuBridge = class {
   _isOperationCallbackCurrent(operation, proxy) {
     return this._isOperationCurrent(operation) && proxy === this._workerProxy;
   }
-  _sendWorkerCancel(proxy) {
+  /**
+   * A bare `cancel()` still resolves with the text the worker generated before
+   * it read the cancel, so the pieces of that text keep streaming until the
+   * call returns. An aborted signal rejects instead, so its stream stops at once.
+   */
+  _streamsPastBareCancel(operation, proxy) {
+    return operation.state === "cancelling" && !operation.callerSignal?.aborted && proxy === this._workerProxy;
+  }
+  _sendWorkerCancel(proxy, operation = null) {
     if (!proxy) {
       return;
     }
@@ -5204,7 +5212,13 @@ var LlamaWebGpuBridge = class {
       return;
     }
     try {
-      Promise.resolve(proxy.call("cancel", [], null)).catch(() => {
+      Promise.resolve(proxy.call(
+        "cancel",
+        [],
+        null,
+        [],
+        operation ? { operationId: operation.id } : null
+      )).catch(() => {
       });
     } catch (_) {
     }
@@ -5225,7 +5239,7 @@ var LlamaWebGpuBridge = class {
         proxies2.add(operation.workerProxy);
       }
       for (const proxy of proxies2) {
-        this._sendWorkerCancel(proxy);
+        this._sendWorkerCancel(proxy, operation);
       }
       return;
     }
@@ -5244,7 +5258,7 @@ var LlamaWebGpuBridge = class {
       proxies.add(operation.workerProxy);
     }
     for (const proxy of proxies) {
-      this._sendWorkerCancel(proxy);
+      this._sendWorkerCancel(proxy, operation);
     }
     const runtimes = new Set(operation.runtimes || []);
     if (operation.runtime) {
@@ -5751,6 +5765,10 @@ var LlamaWebGpuBridge = class {
       return true;
     } catch (error) {
       this._throwIfDisposed();
+      if (this._isOperationCancellation(error)) {
+        this._workerModelMissing = true;
+        throw createAbortError("Bridge operation was cancelled.");
+      }
       this._emitBridgeWarn(
         `llamadart: multimodal worker setup failed once; restarting worker (${serializeWorkerError(error)}).`
       );
@@ -6227,18 +6245,22 @@ var LlamaWebGpuBridge = class {
       this._recordWorkerGeneration(operation, proxy);
     }
     const guardedEvent = typeof onEvent === "function" ? (event) => {
-      if (!operation || this._isOperationCallbackCurrent(operation, proxy) || event.event === "usage" && proxy === this._workerProxy) {
+      if (!operation || this._isOperationCallbackCurrent(operation, proxy) || event.event === "usage" && proxy === this._workerProxy || event.event === "token" && this._streamsPastBareCancel(operation, proxy)) {
         onEvent(event);
       }
     } : void 0;
     try {
-      const response = await proxy.call(
+      const pending = proxy.call(
         method,
         args,
         guardedEvent,
         transferList,
         operation ? { operationId: operation.id, workerGeneration: operation.workerGeneration } : void 0
       );
+      if (operation?.cancelRequested) {
+        this._sendWorkerCancel(proxy, operation);
+      }
+      const response = await pending;
       if (response?.state) {
         this._acceptWorkerState(operation, proxy, response.state);
       }
@@ -6580,7 +6602,7 @@ var LlamaWebGpuBridge = class {
           ),
           stallPromise
         ]);
-        if (this._hasMediaParts(options) && !isWarmup && !sawWorkerTokenEvent && String(workerResult || "").trim().length == 0) {
+        if (this._hasMediaParts(options) && !isWarmup && !sawWorkerTokenEvent && !this._activeOperation?.cancelRequested && String(workerResult || "").trim().length == 0) {
           this._emitBridgeWarn(
             "llamadart: multimodal worker produced empty response without token events."
           );
@@ -7805,8 +7827,9 @@ function installBridgeWorkerHost() {
       state
     });
   };
-  self.onmessage = async (event) => {
-    const message = event.data || {};
+  const callsInFlight = /* @__PURE__ */ new Map();
+  const isCallInFlight = (operationId) => operationId === void 0 ? callsInFlight.size > 0 : callsInFlight.has(operationId);
+  const handleMessage = async (message) => {
     const type = message.type;
     const id = message.id ?? 0;
     try {
@@ -7826,6 +7849,13 @@ function installBridgeWorkerHost() {
       }
       const method = String(message.method || "");
       const args = Array.isArray(message.args) ? message.args : [];
+      if (method === "cancel") {
+        if (isCallInFlight(message.operationId)) {
+          bridge.cancel();
+        }
+        self.postMessage({ type: "result", id, value: void 0 });
+        return;
+      }
       if (method === "loadModelFromUrl") {
         const url = args[0];
         const options = args[1] && typeof args[1] === "object" ? { ...args[1] } : {};
@@ -8097,6 +8127,25 @@ function installBridgeWorkerHost() {
       postError(id, error);
     }
   };
+  self.onmessage = async (event) => {
+    const message = event.data || {};
+    if (message.type !== "call" || message.method === "cancel") {
+      await handleMessage(message);
+      return;
+    }
+    const operationId = message.operationId;
+    callsInFlight.set(operationId, (callsInFlight.get(operationId) ?? 0) + 1);
+    try {
+      await handleMessage(message);
+    } finally {
+      const remaining = (callsInFlight.get(operationId) ?? 1) - 1;
+      if (remaining > 0) {
+        callsInFlight.set(operationId, remaining);
+      } else {
+        callsInFlight.delete(operationId);
+      }
+    }
+  };
 }
 function shouldAutoBootBridgeWorkerHost() {
   if (typeof WorkerGlobalScope === "undefined" || !(globalThis instanceof WorkerGlobalScope)) {

webgpu_bridge/manifest.json

@@ -5,8 +5,8 @@
       "size_bytes": 19006
     },
     "llama_webgpu_bridge.js": {
-      "sha256": "40e6ec9450b50039ac4482ee793d3f4f1a10539936064eac1cc1a2de04fca882",
-      "size_bytes": 297944
+      "sha256": "984e42a2e3d470ac74d922effc0402ba6e4d940027f2b8b088213b0ce0aaa10c",
+      "size_bytes": 299873
     },
     "llama_webgpu_bridge_worker.js": {
       "sha256": "47bbfa0fe897e708455b9497e9aa41d2e94489c329300c33b12936fb3169deca",
@@ -30,8 +30,8 @@
     }
   },
   "assets_repository": "leehack/llama-web-bridge-assets",
-  "bridge_assets_tag": "v0.1.70",
-  "bridge_commit": "3d78c7420b7ddeca6f5779315d9b7d98015c21b2",
+  "bridge_assets_tag": "v0.1.71",
+  "bridge_commit": "24b1b5f14ea6c20e001b18af42c4568b7dbba025",
   "bridge_repository": "leehack/llama-web-bridge",
   "capabilities": {
     "memory64": true,
@@ -66,8 +66,8 @@
       "size_bytes": 19006
     },
     "llama_webgpu_bridge.js": {
-      "sha256": "40e6ec9450b50039ac4482ee793d3f4f1a10539936064eac1cc1a2de04fca882",
-      "size_bytes": 297944
+      "sha256": "984e42a2e3d470ac74d922effc0402ba6e4d940027f2b8b088213b0ce0aaa10c",
+      "size_bytes": 299873
     },
     "llama_webgpu_bridge_worker.js": {
       "sha256": "47bbfa0fe897e708455b9497e9aa41d2e94489c329300c33b12936fb3169deca",
@@ -90,15 +90,15 @@
       "size_bytes": 10479436
     }
   },
-  "github_run_id": "39873093590",
-  "github_run_url": "https://github.com/leehack/llama-web-bridge/actions/runs/39873093590",
+  "github_run_id": "39963902003",
+  "github_run_url": "https://github.com/leehack/llama-web-bridge/actions/runs/39963902003",
   "llama_cpp_commit": "620d2225cc436ab0b001e3200e590e3b286ef9ab",
   "llama_cpp_tag": "v0.7.0",
   "native_commit": "298314bd0ff4650363e0e63b81399d5618869a63",
   "native_manifest_sha256": "92b9f1ea6d3ec3dd826122ffde3cf21b2b22036c368e826d86e927284c6908f4",
   "native_release_tag": "v0.7.0-2",
   "native_repository": "leehack/llamadart-native",
-  "orchestrator_correlation_id": "auto-stable-v0.7.0-2-92b9f1ea6d3ec3dd-build-3d78c7420b7ddeca",
+  "orchestrator_correlation_id": "auto-stable-v0.7.0-2-92b9f1ea6d3ec3dd-build-24b1b5f14ea6c20e",
   "qualification_gates": {
     "multimodal": "passed",
     "speech_to_text": "required-automated-qualification",
@@ -107,9 +107,9 @@
   },
   "release_channel": "stable",
   "release_rebuild": 0,
-  "release_tag": "v0.1.70",
+  "release_tag": "v0.1.71",
   "schema_version": 2,
-  "source_commit": "3d78c7420b7ddeca6f5779315d9b7d98015c21b2",
+  "source_commit": "24b1b5f14ea6c20e001b18af42c4568b7dbba025",
   "source_repository": "leehack/llama-web-bridge",
   "unproven_capabilities": {
     "hardware_gpu_acceleration": "unavailable-on-hosted-runners",

webgpu_bridge/sha256sums.txt

@@ -1,4 +1,4 @@
-40e6ec9450b50039ac4482ee793d3f4f1a10539936064eac1cc1a2de04fca882  llama_webgpu_bridge.js
+984e42a2e3d470ac74d922effc0402ba6e4d940027f2b8b088213b0ce0aaa10c  llama_webgpu_bridge.js
 05d04252ae79a75ce43ce912d92334c6438b84de83eaeafe2ef9a6d4a88ce036  llama_webgpu_core.js
 c760eadf57fe0d5bf3446d4102c2c5aad1612de7f055ea2bf491344a2b2db0e7  llama_webgpu_core.wasm
 47bbfa0fe897e708455b9497e9aa41d2e94489c329300c33b12936fb3169deca  llama_webgpu_bridge_worker.js