osense-vad 1.0.0

package/dist/real-time-vad.js

@@ -0,0 +1,171 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.RealTimeVAD = exports.DEFAULT_MODEL = void 0;
+exports.getDefaultRealTimeVADOptions = getDefaultRealTimeVADOptions;
+const frame_processor_1 = require("./common/frame-processor");
+const messages_1 = require("./common/messages");
+const models_1 = require("./common/models");
+const resampler_1 = require("./common/resampler");
+exports.DEFAULT_MODEL = "v6";
+/**
+ * Build default options based on chosen model
+ */
+function getDefaultRealTimeVADOptions(model = exports.DEFAULT_MODEL) {
+    const frameOpts = model === "v6"
+        ? frame_processor_1.defaultV6FrameProcessorOptions
+        : model === "v5"
+            ? frame_processor_1.defaultV5FrameProcessorOptions
+            : frame_processor_1.defaultLegacyFrameProcessorOptions;
+    return {
+        ...frameOpts,
+        sampleRate: 16000,
+        onFrameProcessed: () => { },
+        onVADMisfire: () => {
+            /* no-op */
+        },
+        onSpeechStart: () => {
+            /* no-op */
+        },
+        onSpeechRealStart: () => {
+            /* no-op */
+        },
+        onSpeechEnd: () => {
+            /* no-op */
+        },
+        ortConfig: undefined,
+        model,
+    };
+}
+/**
+ * RealTimeVAD processes raw audio buffers, frames, and emits events
+ */
+class RealTimeVAD {
+    /**
+     * Construct a new instance with provided options and loaded model
+     */
+    constructor(options, modelInstance) {
+        this.options = options;
+        this.buffer = new Float32Array(0);
+        this.active = false;
+        this.resampler = null;
+        /** Handle events emitted by the frame processor */
+        this.handleEvent = (ev) => {
+            switch (ev.msg) {
+                case messages_1.Message.FrameProcessed:
+                    this.options.onFrameProcessed(ev.probs, ev.frame);
+                    break;
+                case messages_1.Message.SpeechStart:
+                    this.options.onSpeechStart();
+                    break;
+                case messages_1.Message.SpeechRealStart:
+                    this.options.onSpeechRealStart();
+                    break;
+                case messages_1.Message.VADMisfire:
+                    this.options.onVADMisfire();
+                    break;
+                case messages_1.Message.SpeechEnd:
+                    this.options.onSpeechEnd(ev.audio);
+                    break;
+            }
+        };
+        this.modelInstance = modelInstance;
+        this.frameSize = options.frameSamples;
+        this.frameProcessor = new frame_processor_1.FrameProcessor(modelInstance.process, modelInstance.reset_state, {
+            frameSamples: options.frameSamples,
+            positiveSpeechThreshold: options.positiveSpeechThreshold,
+            negativeSpeechThreshold: options.negativeSpeechThreshold,
+            redemptionFrames: options.redemptionFrames,
+            preSpeechPadFrames: options.preSpeechPadFrames,
+            minSpeechFrames: options.minSpeechFrames,
+            submitUserSpeechOnPause: options.submitUserSpeechOnPause,
+        });
+        if (options.sampleRate > 16000) {
+            this.resampler = new resampler_1.Resampler({
+                nativeSampleRate: options.sampleRate,
+                targetSampleRate: 16000,
+                targetFrameSize: this.frameSize,
+            });
+        }
+    }
+    /**
+     * Create and initialize a RealTimeVAD instance
+     */
+    static async new(ort, modelFetcher, opts = {}) {
+        const fullOpts = {
+            ...getDefaultRealTimeVADOptions(opts.model),
+            ...opts,
+        };
+        (0, frame_processor_1.validateOptions)(fullOpts);
+        if (fullOpts.ortConfig)
+            fullOpts.ortConfig(ort);
+        const modelVersion = fullOpts.model || exports.DEFAULT_MODEL;
+        const factory = modelVersion === "v6"
+            ? models_1.SileroV6.new
+            : modelVersion === "v5"
+                ? models_1.SileroV5.new
+                : models_1.SileroLegacy.new;
+        const modelInstance = await factory(ort, modelFetcher);
+        return new RealTimeVAD(fullOpts, modelInstance);
+    }
+    /** Start processing incoming frames */
+    start() {
+        this.active = true;
+        this.frameProcessor.resume();
+    }
+    /** Pause processing; may emit end-segment on pause */
+    pause() {
+        this.active = false;
+        this.frameProcessor.pause(this.handleEvent);
+    }
+    /** Feed raw audio (any sample rate) into the VAD */
+    async processAudio(audioData) {
+        if (!this.active)
+            return;
+        let data = audioData;
+        if (this.resampler) {
+            const chunks = [];
+            for await (const frame of this.resampler.stream(audioData)) {
+                chunks.push(frame);
+            }
+            data = new Float32Array(chunks.reduce((sum, c) => sum + c.length, 0));
+            let off = 0;
+            for (const c of chunks) {
+                data.set(c, off);
+                off += c.length;
+            }
+        }
+        // append to internal buffer
+        const tmp = new Float32Array(this.buffer.length + data.length);
+        tmp.set(this.buffer);
+        tmp.set(data, this.buffer.length);
+        this.buffer = tmp;
+        // process complete frames
+        while (this.buffer.length >= this.frameSize) {
+            const frame = this.buffer.subarray(0, this.frameSize);
+            this.buffer = this.buffer.subarray(this.frameSize);
+            await this.frameProcessor.process(frame, this.handleEvent);
+        }
+    }
+    /** Flush any remaining audio and end segment */
+    async flush() {
+        if (this.buffer.length > 0 && this.buffer.length < this.frameSize) {
+            const pad = new Float32Array(this.frameSize);
+            pad.set(this.buffer);
+            await this.frameProcessor.process(pad, this.handleEvent);
+        }
+        this.frameProcessor.endSegment(this.handleEvent);
+        this.buffer = new Float32Array(0);
+    }
+    /** Reset internal state */
+    reset() {
+        this.buffer = new Float32Array(0);
+        this.modelInstance.reset_state();
+    }
+    /** Clean up resources */
+    destroy() {
+        this.pause();
+        this.reset();
+    }
+}
+exports.RealTimeVAD = RealTimeVAD;
+//# sourceMappingURL=real-time-vad.js.map

package/dist/real-time-vad.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"real-time-vad.js","sourceRoot":"","sources":["../src/real-time-vad.ts"],"names":[],"mappings":";;;AAqDA,oEA4BC;AAhFD,8DAQkC;AAClC,gDAA4C;AAC5C,4CAMyB;AACzB,kDAA+C;AAGlC,QAAA,aAAa,GAAiB,IAAI,CAAC;AA6BhD;;GAEG;AACH,SAAgB,4BAA4B,CAC1C,QAAsB,qBAAa;IAEnC,MAAM,SAAS,GACb,KAAK,KAAK,IAAI;QACZ,CAAC,CAAC,gDAA8B;QAChC,CAAC,CAAC,KAAK,KAAK,IAAI;YACd,CAAC,CAAC,gDAA8B;YAChC,CAAC,CAAC,oDAAkC,CAAC;IAC3C,OAAO;QACL,GAAG,SAAS;QACZ,UAAU,EAAE,KAAK;QACjB,gBAAgB,EAAE,GAAG,EAAE,GAAE,CAAC;QAC1B,YAAY,EAAE,GAAG,EAAE;YACjB,WAAW;QACb,CAAC;QACD,aAAa,EAAE,GAAG,EAAE;YAClB,WAAW;QACb,CAAC;QACD,iBAAiB,EAAE,GAAG,EAAE;YACtB,WAAW;QACb,CAAC;QACD,WAAW,EAAE,GAAG,EAAE;YAChB,WAAW;QACb,CAAC;QACD,SAAS,EAAE,SAAS;QACpB,KAAK;KACgB,CAAC;AAC1B,CAAC;AAED;;GAEG;AACH,MAAa,WAAW;IAQtB;;OAEG;IACH,YACU,OAA2B,EACnC,aAAkB;QADV,YAAO,GAAP,OAAO,CAAoB;QAT7B,WAAM,GAAG,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC;QAE7B,WAAM,GAAG,KAAK,CAAC;QACf,cAAS,GAAqB,IAAI,CAAC;QA4H3C,mDAAmD;QAC3C,gBAAW,GAAG,CAAC,EAAuB,EAAQ,EAAE;YACtD,QAAQ,EAAE,CAAC,GAAG,EAAE,CAAC;gBACf,KAAK,kBAAO,CAAC,cAAc;oBACzB,IAAI,CAAC,OAAO,CAAC,gBAAgB,CAAC,EAAE,CAAC,KAAM,EAAE,EAAE,CAAC,KAAM,CAAC,CAAC;oBACpD,MAAM;gBACR,KAAK,kBAAO,CAAC,WAAW;oBACtB,IAAI,CAAC,OAAO,CAAC,aAAa,EAAE,CAAC;oBAC7B,MAAM;gBACR,KAAK,kBAAO,CAAC,eAAe;oBAC1B,IAAI,CAAC,OAAO,CAAC,iBAAiB,EAAE,CAAC;oBACjC,MAAM;gBACR,KAAK,kBAAO,CAAC,UAAU;oBACrB,IAAI,CAAC,OAAO,CAAC,YAAY,EAAE,CAAC;oBAC5B,MAAM;gBACR,KAAK,kBAAO,CAAC,SAAS;oBACpB,IAAI,CAAC,OAAO,CAAC,WAAW,CAAC,EAAE,CAAC,KAAM,CAAC,CAAC;oBACpC,MAAM;YACV,CAAC;QACH,CAAC,CAAC;QAtIA,IAAI,CAAC,aAAa,GAAG,aAAa,CAAC;QACnC,IAAI,CAAC,SAAS,GAAG,OAAO,CAAC,YAAY,CAAC;QAEtC,IAAI,CAAC,cAAc,GAAG,IAAI,gCAAc,CACtC,aAAa,CAAC,OAAO,EACrB,aAAa,CAAC,WAAW,EACzB;YACE,YAAY,EAAE,OAAO,CAAC,YAAY;YAClC,uBAAuB,EAAE,OAAO,CAAC,uBAAuB;YACxD,uBAAuB,EAAE,OAAO,CAAC,uBAAuB;YACxD,gBAAgB,EAAE,OAAO,CAAC,gBAAgB;YAC1C,kBAAkB,EAAE,OAAO,CAAC,kBAAkB;YAC9C,eAAe,EAAE,OAAO,CAAC,eAAe;YACxC,uBAAuB,EAAE,OAAO,CAAC,uBAAuB;SACzD,CACF,CAAC;QAEF,IAAI,OAAO,CAAC,UAAU,GAAG,KAAK,EAAE,CAAC;YAC/B,IAAI,CAAC,SAAS,GAAG,IAAI,qBAAS,CAAC;gBAC7B,gBAAgB,EAAE,OAAO,CAAC,UAAU;gBACpC,gBAAgB,EAAE,KAAK;gBACvB,eAAe,EAAE,IAAI,CAAC,SAAS;aAChC,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED;;OAEG;IACH,MAAM,CAAC,KAAK,CAAC,GAAG,CACd,GAAQ,EACR,YAAwC,EACxC,OAAoC,EAAE;QAEtC,MAAM,QAAQ,GAAuB;YACnC,GAAG,4BAA4B,CAAC,IAAI,CAAC,KAAK,CAAC;YAC3C,GAAG,IAAI;SACR,CAAC;QACF,IAAA,iCAAe,EAAC,QAAQ,CAAC,CAAC;QAE1B,IAAI,QAAQ,CAAC,SAAS;YAAE,QAAQ,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC;QAEhD,MAAM,YAAY,GAAiB,QAAQ,CAAC,KAAK,IAAI,qBAAa,CAAC;QACnE,MAAM,OAAO,GACX,YAAY,KAAK,IAAI;YACnB,CAAC,CAAC,iBAAQ,CAAC,GAAG;YACd,CAAC,CAAC,YAAY,KAAK,IAAI;gBACrB,CAAC,CAAC,iBAAQ,CAAC,GAAG;gBACd,CAAC,CAAC,qBAAY,CAAC,GAAG,CAAC;QACzB,MAAM,aAAa,GAAG,MAAM,OAAO,CAAC,GAAG,EAAE,YAAY,CAAC,CAAC;QAEvD,OAAO,IAAI,WAAW,CAAC,QAAQ,EAAE,aAAa,CAAC,CAAC;IAClD,CAAC;IAED,uCAAuC;IACvC,KAAK;QACH,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC;QACnB,IAAI,CAAC,cAAc,CAAC,MAAM,EAAE,CAAC;IAC/B,CAAC;IAED,sDAAsD;IACtD,KAAK;QACH,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC;QACpB,IAAI,CAAC,cAAc,CAAC,KAAK,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAC9C,CAAC;IAED,oDAAoD;IACpD,KAAK,CAAC,YAAY,CAAC,SAAuB;QACxC,IAAI,CAAC,IAAI,CAAC,MAAM;YAAE,OAAO;QAEzB,IAAI,IAAI,GAAG,SAAS,CAAC;QACrB,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,MAAM,MAAM,GAAmB,EAAE,CAAC;YAClC,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,SAAS,CAAC,EAAE,CAAC;gBAC3D,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACrB,CAAC;YACD,IAAI,GAAG,IAAI,YAAY,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,CAAC;YACtE,IAAI,GAAG,GAAG,CAAC,CAAC;YACZ,KAAK,MAAM,CAAC,IAAI,MAAM,EAAE,CAAC;gBACvB,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;gBACjB,GAAG,IAAI,CAAC,CAAC,MAAM,CAAC;YAClB,CAAC;QACH,CAAC;QAED,4BAA4B;QAC5B,MAAM,GAAG,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC;QAC/D,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QACrB,GAAG,CAAC,GAAG,CAAC,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;QAClC,IAAI,CAAC,MAAM,GAAG,GAAG,CAAC;QAElB,0BAA0B;QAC1B,OAAO,IAAI,CAAC,MAAM,CAAC,MAAM,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YAC5C,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,EAAE,IAAI,CAAC,SAAS,CAAC,CAAC;YACtD,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;YACnD,MAAM,IAAI,CAAC,cAAc,CAAC,OAAO,CAAC,KAAK,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;QAC7D,CAAC;IACH,CAAC;IAED,gDAAgD;IAChD,KAAK,CAAC,KAAK;QACT,IAAI,IAAI,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,MAAM,GAAG,IAAI,CAAC,SAAS,EAAE,CAAC;YAClE,MAAM,GAAG,GAAG,IAAI,YAAY,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;YAC7C,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;YACrB,MAAM,IAAI,CAAC,cAAc,CAAC,OAAO,CAAC,GAAG,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;QAC3D,CAAC;QACD,IAAI,CAAC,cAAc,CAAC,UAAU,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;QACjD,IAAI,CAAC,MAAM,GAAG,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC;IACpC,CAAC;IAED,2BAA2B;IAC3B,KAAK;QACH,IAAI,CAAC,MAAM,GAAG,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC;QAClC,IAAI,CAAC,aAAa,CAAC,WAAW,EAAE,CAAC;IACnC,CAAC;IAuBD,yBAAyB;IACzB,OAAO;QACL,IAAI,CAAC,KAAK,EAAE,CAAC;QACb,IAAI,CAAC,KAAK,EAAE,CAAC;IACf,CAAC;CACF;AA5JD,kCA4JC"}

package/package.json

@@ -0,0 +1,66 @@
+{
+  "name": "osense-vad",
+  "version": "1.0.0",
+  "description": "A Node.js library for Voice Activity Detection using Silero VAD",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "scripts": {
+    "build": "./scripts/build.sh",
+		"test": "mocha test/*.spec.js",
+		"clean": "rm -rf dist"
+  },
+  "keywords": [
+    "vad",
+    "voice-activity-detection",
+    "silero",
+    "audio",
+    "speech",
+    "onnx"
+  ],
+  "author": {
+    "name": "Agent Voice Response",
+    "email": "info@agentvoiceresponse.com",
+    "url": "https://github.com/agentvoiceresponse"
+  },
+  "license": "MIT",
+  "dependencies": {
+    "onnxruntime-node": "^1.22.0-rev"
+  },
+  "devDependencies": {
+    "@eslint/eslintrc": "^3.3.1",
+    "@eslint/js": "^9.28.0",
+    "@types/mocha": "^10.0.1",
+    "@types/node": "^22.15.30",
+    "@typescript-eslint/eslint-plugin": "^8.34.0",
+    "@typescript-eslint/parser": "^8.34.0",
+    "chai": "^4.3.7",
+    "eslint": "^9.28.0",
+    "globals": "^16.2.0",
+    "mocha": "^10.2.0",
+    "sinon": "^15.0.1",
+    "typescript": "^5.8.3",
+    "wav-decoder": "^1.3.0"
+  },
+  "files": [
+    "dist/**/*",
+    "silero_vad_v6.onnx",
+    "silero_vad_v5.onnx",
+    "silero_vad_legacy.onnx",
+    "README.md",
+    "LICENSE"
+  ],
+  "engines": {
+    "node": ">=16.0.0"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/agentvoiceresponse/avr-vad.git"
+  },
+  "bugs": {
+    "url": "https://github.com/agentvoiceresponse/avr-vad/issues"
+  },
+  "homepage": "https://github.com/agentvoiceresponse/avr-vad#readme",
+  "publishConfig": {
+    "access": "public"
+  }
+}