@fonoster/autopilot 0.7.54 → 0.7.56

package/dist/vad/SileroVadModel.js

@@ -35,7 +35,8 @@ class SileroVadModel {
     }
     async init() {
         const modelArrayBuffer = (0, fs_1.readFileSync)(this.pathToModel).buffer;
-        this._session = await this.ort.InferenceSession.create(modelArrayBuffer);
+        const sessionOption = { interOpNumThreads: 1, intraOpNumThreads: 1 };
+        this._session = await this.ort.InferenceSession.create(modelArrayBuffer, sessionOption);
         this._sr = new this.ort.Tensor("int64", [16000n]);
         this._state = getNewState(this.ort);
     }

package/dist/vad/chunkToFloat32Array.js

@@ -24,16 +24,18 @@ exports.chunkToFloat32Array = chunkToFloat32Array;
 //
 // Q. Would it be the same if we just created a new Uint8Array from the chunk?
 function chunkToFloat32Array(chunk) {
-    // Check if byteOffset is not aligned
-    const alignedByteOffset = chunk.byteOffset % Int16Array.BYTES_PER_ELEMENT === 0;
     let int16Array;
+    const alignedByteOffset = chunk.byteOffset % Int16Array.BYTES_PER_ELEMENT === 0;
     if (alignedByteOffset) {
         int16Array = new Int16Array(chunk.buffer, chunk.byteOffset, chunk.byteLength / Int16Array.BYTES_PER_ELEMENT);
     }
     else {
-        // Create a new aligned Uint8Array and then an Int16Array from it
         const alignedChunk = new Uint8Array(chunk);
         int16Array = new Int16Array(alignedChunk.buffer, alignedChunk.byteOffset, alignedChunk.byteLength / Int16Array.BYTES_PER_ELEMENT);
     }
-    return new Float32Array(Array.from(int16Array, (sample) => sample / 32768.0));
+    const floatArray = new Float32Array(int16Array.length);
+    for (let i = 0; i < int16Array.length; i++) {
+        floatArray[i] = int16Array[i] / 32768.0;
+    }
+    return floatArray;
 }

package/dist/vad/makeVad.js

@@ -34,11 +34,28 @@ var __importStar = (this && this.__importStar) || (function () {
 })();
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.makeVad = makeVad;
-const logger_1 = require("@fonoster/logger");
+/* eslint-disable no-loops/no-loops */
+/*
+ * Copyright (C) 2024 by Fonoster Inc (https://fonoster.com)
+ * http://github.com/fonoster/fonoster
+ *
+ * This file is part of Fonoster
+ *
+ * Licensed under the MIT License (the "License");
+ * you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ *    https://opensource.org/licenses/MIT
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
 const ort = __importStar(require("onnxruntime-node"));
 const chunkToFloat32Array_1 = require("./chunkToFloat32Array");
 const SileroVadModel_1 = require("./SileroVadModel");
-const logger = (0, logger_1.getLogger)({ service: "autopilot", filePath: __filename });
 const BUFFER_SIZE = 512;
 async function makeVad(params) {
     const { pathToModel, activationThreshold, deactivationThreshold, debounceFrames } = params;
@@ -56,7 +73,6 @@ async function makeVad(params) {
             const audioFrame = buffer.slice(0, BUFFER_SIZE);
             const remainingBuffer = buffer.slice(BUFFER_SIZE);
             const result = await silero.process(new Float32Array(audioFrame));
-            logger.silly("last vad result", { ...result });
             if (result.isSpeech > activationThreshold) {
                 consecutiveNonSpeechFrames = 0; // Reset non-speech counter
                 consecutiveSpeechFrames++;

package/dist/vad/types.d.ts

@@ -8,7 +8,10 @@ type SpeechProbabilities = {
 };
 type ONNXRuntimeAPI = {
     InferenceSession: {
-        create(modelArrayBuffer: ArrayBuffer): Promise<unknown>;
+        create(modelArrayBuffer: ArrayBuffer, sessionOption: {
+            interOpNumThreads: number;
+            intraOpNumThreads: number;
+        }): Promise<unknown>;
     };
     Tensor: {
         new (type: "int64", data: BigInt[]): unknown;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fonoster/autopilot",
-  "version": "0.7.54",
+  "version": "0.7.56",
   "description": "Voice AI for the Fonoster platform",
   "author": "Pedro Sanders <psanders@fonoster.com>",
   "homepage": "https://github.com/fonoster/fonoster#readme",
@@ -36,10 +36,10 @@
   },
   "dependencies": {
     "@aws-sdk/client-s3": "^3.712.0",
-    "@fonoster/common": "^0.7.54",
-    "@fonoster/logger": "^0.7.54",
-    "@fonoster/types": "^0.7.54",
-    "@fonoster/voice": "^0.7.54",
+    "@fonoster/common": "^0.7.56",
+    "@fonoster/logger": "^0.7.56",
+    "@fonoster/types": "^0.7.56",
+    "@fonoster/voice": "^0.7.56",
     "@langchain/community": "^0.3.19",
     "@langchain/core": "^0.3.23",
     "@langchain/groq": "^0.1.2",
@@ -57,5 +57,5 @@
   "devDependencies": {
     "typescript": "^5.5.4"
   },
-  "gitHead": "f762eb57b3e07b679bbb2194d48abaeaf75e88ef"
+  "gitHead": "f918235ae68a49970b45022637347f673ca63941"
 }