@fonoster/autopilot 0.7.53 → 0.7.55

package/dist/Autopilot.d.ts

@@ -1,8 +1,8 @@
 import { AutopilotParams } from "./types";
 declare class Autopilot {
-    private params;
-    private actor;
-    private vadWorker;
+    private readonly params;
+    private readonly actor;
+    private readonly vadWorker;
     constructor(params: AutopilotParams);
     start(): void;
     stop(): void;

package/dist/envs.d.ts

@@ -7,3 +7,4 @@ export declare const NODE_ENV: string;
 export declare const SKIP_IDENTITY: boolean;
 export declare const UNSTRUCTURED_API_KEY: string | undefined;
 export declare const UNSTRUCTURED_API_URL: string;
+export declare const SILERO_VAD_MODEL_PATH: string;

package/dist/envs.js

@@ -3,7 +3,7 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.UNSTRUCTURED_API_URL = exports.UNSTRUCTURED_API_KEY = exports.SKIP_IDENTITY = exports.NODE_ENV = exports.KNOWLEDGE_BASE_ENABLED = exports.AWS_S3_SECRET_ACCESS_KEY = exports.AWS_S3_REGION = exports.AWS_S3_ENDPOINT = exports.AWS_S3_ACCESS_KEY_ID = void 0;
+exports.SILERO_VAD_MODEL_PATH = exports.UNSTRUCTURED_API_URL = exports.UNSTRUCTURED_API_KEY = exports.SKIP_IDENTITY = exports.NODE_ENV = exports.KNOWLEDGE_BASE_ENABLED = exports.AWS_S3_SECRET_ACCESS_KEY = exports.AWS_S3_REGION = exports.AWS_S3_ENDPOINT = exports.AWS_S3_ACCESS_KEY_ID = void 0;
 /*
  * Copyright (C) 2024 by Fonoster Inc (https://fonoster.com)
  * http://github.com/fonoster/fonoster
@@ -38,6 +38,7 @@ exports.NODE_ENV = e.NODE_ENV || "production";
 exports.SKIP_IDENTITY = e.SKIP_IDENTITY === "true";
 exports.UNSTRUCTURED_API_KEY = e.UNSTRUCTURED_API_KEY;
 exports.UNSTRUCTURED_API_URL = e.UNSTRUCTURED_API_URL ?? "https://api.unstructuredapp.io/general/v0/general";
+exports.SILERO_VAD_MODEL_PATH = e.SILERO_VAD_MODEL_PATH ?? (0, path_1.join)(__dirname, "..", "silero_vad.onnx");
 if (exports.KNOWLEDGE_BASE_ENABLED) {
     (0, common_1.assertEnvsAreSet)([
         "AWS_S3_ACCESS_KEY_ID",

package/dist/vad/SileroVad.d.ts

@@ -1,14 +1,14 @@
 import { Vad } from "./types";
 declare class SileroVad implements Vad {
     private vad;
-    private params;
+    private readonly params;
     constructor(params: {
-        pathToModel?: string;
+        pathToModel: string;
         activationThreshold: number;
         deactivationThreshold: number;
         debounceFrames: number;
     });
-    pathToModel?: string;
+    pathToModel: string;
     activationThreshold: number;
     deactivationThreshold: number;
     debounceFrames: number;

package/dist/vad/SileroVadModel.d.ts

@@ -1,15 +1,14 @@
 import { ONNXRuntimeAPI, SpeechProbabilities } from "./types";
 declare class SileroVadModel {
-    private ort;
-    private pathToModel;
-    _session: any;
-    _h: unknown;
-    _c: unknown;
-    _sr: unknown;
+    private readonly ort;
+    private readonly pathToModel;
+    private _session;
+    private _state;
+    private _sr;
     constructor(ort: ONNXRuntimeAPI, pathToModel: string);
-    static new: (ort: ONNXRuntimeAPI, pathToModel: string) => Promise<SileroVadModel>;
+    static readonly new: (ort: ONNXRuntimeAPI, pathToModel: string) => Promise<SileroVadModel>;
     init(): Promise<void>;
+    resetState: () => void;
     process(audioFrame: Float32Array): Promise<SpeechProbabilities>;
-    resetState(): void;
 }
 export { SileroVadModel };

package/dist/vad/SileroVadModel.js

@@ -21,16 +21,23 @@ exports.SileroVadModel = void 0;
  * limitations under the License.
  */
 const fs_1 = require("fs");
+function getNewState(ortInstance) {
+    const zeroes = Array(2 * 128).fill(0);
+    return new ortInstance.Tensor("float32", zeroes, [2, 1, 128]);
+}
 class SileroVadModel {
     constructor(ort, pathToModel) {
         this.ort = ort;
         this.pathToModel = pathToModel;
+        this.resetState = () => {
+            this._state = getNewState(this.ort);
+        };
     }
     async init() {
         const modelArrayBuffer = (0, fs_1.readFileSync)(this.pathToModel).buffer;
         this._session = await this.ort.InferenceSession.create(modelArrayBuffer);
         this._sr = new this.ort.Tensor("int64", [16000n]);
-        this.resetState();
+        this._state = getNewState(this.ort);
     }
     async process(audioFrame) {
         const t = new this.ort.Tensor("float32", audioFrame, [
@@ -39,22 +46,15 @@ class SileroVadModel {
         ]);
         const inputs = {
             input: t,
-            h: this._h,
-            c: this._c,
+            state: this._state,
             sr: this._sr
         };
         const out = await this._session.run(inputs);
-        this._h = out.hn;
-        this._c = out.cn;
-        const [isSpeech] = out.output.data;
+        this._state = out["stateN"];
+        const [isSpeech] = out["output"].data;
         const notSpeech = 1 - isSpeech;
         return { notSpeech, isSpeech };
     }
-    resetState() {
-        const zeroes = Array(2 * 64).fill(0);
-        this._h = new this.ort.Tensor("float32", zeroes, [2, 1, 64]);
-        this._c = new this.ort.Tensor("float32", zeroes, [2, 1, 64]);
-    }
 }
 exports.SileroVadModel = SileroVadModel;
 _a = SileroVadModel;

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.d.ts

@@ -1,5 +1,5 @@
 declare function makeVad(params: {
-    pathToModel?: string;
+    pathToModel: string;
     activationThreshold: number;
     deactivationThreshold: number;
     debounceFrames: number;

package/dist/vad/makeVad.js

@@ -53,17 +53,13 @@ exports.makeVad = makeVad;
  * See the License for the specific language governing permissions and
  * limitations under the License.
  */
-const path_1 = require("path");
-const logger_1 = require("@fonoster/logger");
 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 = 16000;
+const BUFFER_SIZE = 512;
 async function makeVad(params) {
     const { pathToModel, activationThreshold, deactivationThreshold, debounceFrames } = params;
-    const effectivePath = pathToModel || (0, path_1.join)(__dirname, "..", "..", "silero_vad.onnx");
-    const silero = await SileroVadModel_1.SileroVadModel.new(ort, effectivePath);
+    const silero = await SileroVadModel_1.SileroVadModel.new(ort, pathToModel);
     let audioBuffer = [];
     let isSpeechActive = false;
     let consecutiveSpeechFrames = 0;
@@ -77,7 +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

@@ -11,9 +11,9 @@ type ONNXRuntimeAPI = {
         create(modelArrayBuffer: ArrayBuffer): Promise<unknown>;
     };
     Tensor: {
-        new (type: "int64", dims: [16000n]): unknown;
-        new (type: "float32", data: number[], dims: [2, 1, 64]): unknown;
-        new (type: "float32", data: Float32Array, dims: [1, number]): unknown;
+        new (type: "int64", data: BigInt[]): unknown;
+        new (type: "int64", data: BigInt[], dims: [1]): unknown;
+        new (type: "float32", data: Float32Array | number[], dims: [2, 1, 128]): unknown;
         new (type: "float32", data: Float32Array, dims: [1, number]): unknown;
     };
 };

package/dist/vadWorker.js

@@ -20,7 +20,11 @@ Object.defineProperty(exports, "__esModule", { value: true });
  */
 const worker_threads_1 = require("worker_threads");
 const SileroVad_1 = require("./vad/SileroVad");
-const vad = new SileroVad_1.SileroVad(worker_threads_1.workerData);
+const envs_1 = require("./envs");
+const vad = new SileroVad_1.SileroVad({
+    ...worker_threads_1.workerData,
+    pathToModel: envs_1.SILERO_VAD_MODEL_PATH
+});
 vad.init().then(() => {
     worker_threads_1.parentPort?.on("message", (chunk) => {
         vad.processChunk(chunk, (voiceActivity) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fonoster/autopilot",
-  "version": "0.7.53",
+  "version": "0.7.55",
   "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": "^0.7.53",
-    "@fonoster/logger": "^0.7.53",
-    "@fonoster/types": "^0.7.53",
-    "@fonoster/voice": "^0.7.53",
+    "@fonoster/common": "^0.7.55",
+    "@fonoster/logger": "^0.7.55",
+    "@fonoster/types": "^0.7.55",
+    "@fonoster/voice": "^0.7.55",
     "@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": "62f407f61199e5e35a5a5af1a70dd055ce127547"
+  "gitHead": "9e6953d9c5fdd753d3742ef3da84b44d56c4c4ed"
 }