@tensamin/audio 0.1.0 → 0.1.2

package/dist/vad/vad-node.js

@@ -0,0 +1,92 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/vad/vad-node.ts
+var vad_node_exports = {};
+__export(vad_node_exports, {
+  EnergyVADPlugin: () => EnergyVADPlugin
+});
+module.exports = __toCommonJS(vad_node_exports);
+var energyVadWorkletCode = `
+class EnergyVadProcessor extends AudioWorkletProcessor {
+  constructor() {
+    super();
+    this.smoothing = 0.95;
+    this.energy = 0;
+    this.noiseFloor = 0.001;
+  }
+
+  process(inputs, outputs, parameters) {
+    const input = inputs[0];
+    if (!input || !input.length) return true;
+    const channel = input[0];
+    
+    // Calculate RMS
+    let sum = 0;
+    for (let i = 0; i < channel.length; i++) {
+      sum += channel[i] * channel[i];
+    }
+    const rms = Math.sqrt(sum / channel.length);
+
+    // Simple adaptive noise floor (very basic)
+    if (rms < this.noiseFloor) {
+      this.noiseFloor = this.noiseFloor * 0.99 + rms * 0.01;
+    } else {
+      this.noiseFloor = this.noiseFloor * 0.999 + rms * 0.001;
+    }
+
+    // Calculate "probability" based on SNR
+    // This is a heuristic mapping from energy to 0-1
+    const snr = rms / (this.noiseFloor + 1e-6);
+    const probability = Math.min(1, Math.max(0, (snr - 1.5) / 10)); // Arbitrary scaling
+
+    this.port.postMessage({ probability });
+
+    return true;
+  }
+}
+registerProcessor('energy-vad-processor', EnergyVadProcessor);
+`;
+var EnergyVADPlugin = class {
+  name = "energy-vad";
+  async createNode(context, config, onDecision) {
+    const blob = new Blob([energyVadWorkletCode], {
+      type: "application/javascript"
+    });
+    const url = URL.createObjectURL(blob);
+    try {
+      await context.audioWorklet.addModule(url);
+    } catch (e) {
+      console.warn("Failed to add Energy VAD worklet:", e);
+      throw e;
+    } finally {
+      URL.revokeObjectURL(url);
+    }
+    const node = new AudioWorkletNode(context, "energy-vad-processor");
+    node.port.onmessage = (event) => {
+      const { probability } = event.data;
+      onDecision(probability);
+    };
+    return node;
+  }
+};
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  EnergyVADPlugin
+});

package/dist/vad/vad-node.mjs

@@ -0,0 +1,6 @@
+import {
+  EnergyVADPlugin
+} from "../chunk-UMU2KIB6.mjs";
+export {
+  EnergyVADPlugin
+};

package/dist/vad/vad-state.d.mts

@@ -0,0 +1,15 @@
+import { AudioProcessingConfig, VADState } from '../types.mjs';
+import 'mitt';
+
+declare class VADStateMachine {
+    private config;
+    private currentState;
+    private lastSpeechTime;
+    private speechStartTime;
+    private frameDurationMs;
+    constructor(config: AudioProcessingConfig["vad"]);
+    updateConfig(config: Partial<AudioProcessingConfig["vad"]>): void;
+    processFrame(probability: number, timestamp: number): VADState;
+}
+
+export { VADStateMachine };

package/dist/vad/vad-state.d.ts

@@ -0,0 +1,15 @@
+import { AudioProcessingConfig, VADState } from '../types.js';
+import 'mitt';
+
+declare class VADStateMachine {
+    private config;
+    private currentState;
+    private lastSpeechTime;
+    private speechStartTime;
+    private frameDurationMs;
+    constructor(config: AudioProcessingConfig["vad"]);
+    updateConfig(config: Partial<AudioProcessingConfig["vad"]>): void;
+    processFrame(probability: number, timestamp: number): VADState;
+}
+
+export { VADStateMachine };

package/dist/vad/vad-state.js

@@ -0,0 +1,83 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/vad/vad-state.ts
+var vad_state_exports = {};
+__export(vad_state_exports, {
+  VADStateMachine: () => VADStateMachine
+});
+module.exports = __toCommonJS(vad_state_exports);
+var VADStateMachine = class {
+  config;
+  currentState = "silent";
+  lastSpeechTime = 0;
+  speechStartTime = 0;
+  frameDurationMs = 20;
+  // Assumed frame duration, updated by calls
+  constructor(config) {
+    this.config = {
+      enabled: config?.enabled ?? true,
+      pluginName: config?.pluginName ?? "energy-vad",
+      startThreshold: config?.startThreshold ?? 0.5,
+      stopThreshold: config?.stopThreshold ?? 0.4,
+      hangoverMs: config?.hangoverMs ?? 300,
+      preRollMs: config?.preRollMs ?? 200
+    };
+  }
+  updateConfig(config) {
+    this.config = { ...this.config, ...config };
+  }
+  processFrame(probability, timestamp) {
+    const { startThreshold, stopThreshold, hangoverMs } = this.config;
+    let newState = this.currentState;
+    if (this.currentState === "silent" || this.currentState === "speech_ending") {
+      if (probability >= startThreshold) {
+        newState = "speech_starting";
+        this.speechStartTime = timestamp;
+        this.lastSpeechTime = timestamp;
+      } else {
+        newState = "silent";
+      }
+    } else if (this.currentState === "speech_starting" || this.currentState === "speaking") {
+      if (probability >= stopThreshold) {
+        newState = "speaking";
+        this.lastSpeechTime = timestamp;
+      } else {
+        const timeSinceSpeech = timestamp - this.lastSpeechTime;
+        if (timeSinceSpeech < hangoverMs) {
+          newState = "speaking";
+        } else {
+          newState = "speech_ending";
+        }
+      }
+    }
+    if (newState === "speech_starting") newState = "speaking";
+    if (newState === "speech_ending") newState = "silent";
+    this.currentState = newState;
+    return {
+      isSpeaking: newState === "speaking",
+      probability,
+      state: newState
+    };
+  }
+};
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  VADStateMachine
+});

package/dist/vad/vad-state.mjs

@@ -0,0 +1,6 @@
+import {
+  VADStateMachine
+} from "../chunk-JJASCVEW.mjs";
+export {
+  VADStateMachine
+};

package/package.json

@@ -1,31 +1,25 @@
 {
   "name": "@tensamin/audio",
-  "version": "0.1.0",
+  "version": "0.1.2",
+  "main": "dist/index.js",
   "module": "dist/index.mjs",
   "types": "dist/index.d.ts",
-  "type": "module",
   "author": {
     "email": "aloisianer@proton.me",
     "name": "Alois"
   },
   "publishConfig": {
-    "registry": "https://registry.npmjs.org",
     "access": "public"
   },
   "repository": {
     "type": "git",
     "url": "https://github.com/Tensamin/Audio"
   },
-  "exports": {
-    ".": {
-      "import": "./dist/index.mjs",
-      "types": "./dist/index.d.ts"
-    }
-  },
+  "license": "MIT",
   "scripts": {
-    "build": "tsup src/index.ts --format esm --dts --clean",
-    "dev": "tsup src/index.ts --format esm --dts --watch",
-    "format": "bunx prettier --write ."
+    "build": "tsup src/ --format cjs,esm --dts --out-dir dist --clean",
+    "format": "bunx prettier --write .",
+    "lint": "tsc"
   },
   "dependencies": {
     "@sapphi-red/web-noise-suppressor": "^0.3.5",
@@ -35,10 +29,13 @@
     "livekit-client": "^2.0.0"
   },
   "devDependencies": {
+    "tsup": "^8.5.1",
     "@types/bun": "latest",
     "@types/web": "^0.0.298",
     "livekit-client": "^2.16.0",
-    "tsup": "^8.5.1",
     "typescript": "^5.9.3"
-  }
+  },
+  "files": [
+    "dist"
+  ]
 }

package/.github/workflows/publish.yml

@@ -1,23 +0,0 @@
-name: Publish
-
-on:
-  workflow_dispatch:
-
-permissions:
-  contents: write
-  id-token: write
-
-jobs:
-  publish:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-node@v4
-      
-      - name: Publish
-        run: |
-          npm ci
-          npm run build
-          npm pack
-          echo "//registry.npmjs.org/:_authToken=${{ secrets.NPM_TOKEN }}" > ~/.npmrc
-          npm publish --access public

package/src/context/audio-context.ts

@@ -1,69 +0,0 @@
-/**
- * Manages a shared AudioContext for the application.
- */
-
-let sharedContext: AudioContext | null = null;
-let activePipelines = 0;
-
-/**
- * Gets the shared AudioContext, creating it if necessary.
- * @param options Optional AudioContextOptions
- */
-export function getAudioContext(options?: AudioContextOptions): AudioContext {
-  if (typeof window === "undefined" || typeof AudioContext === "undefined") {
-    throw new Error(
-      "AudioContext is not supported in this environment (browser only).",
-    );
-  }
-
-  if (!sharedContext || sharedContext.state === "closed") {
-    sharedContext = new AudioContext(options);
-  }
-
-  return sharedContext;
-}
-
-/**
- * Registers a pipeline usage. Keeps track of active users.
- */
-export function registerPipeline(): void {
-  activePipelines++;
-}
-
-/**
- * Unregisters a pipeline usage.
- * Optionally closes the context if no pipelines are active (not implemented by default to avoid churn).
- */
-export function unregisterPipeline(): void {
-  activePipelines = Math.max(0, activePipelines - 1);
-}
-
-/**
- * Resumes the shared AudioContext.
- * Should be called in response to a user gesture.
- */
-export async function resumeAudioContext(): Promise<void> {
-  if (sharedContext && sharedContext.state === "suspended") {
-    await sharedContext.resume();
-  }
-}
-
-/**
- * Suspends the shared AudioContext.
- */
-export async function suspendAudioContext(): Promise<void> {
-  if (sharedContext && sharedContext.state === "running") {
-    await sharedContext.suspend();
-  }
-}
-
-/**
- * Closes the shared AudioContext and releases resources.
- */
-export async function closeAudioContext(): Promise<void> {
-  if (sharedContext && sharedContext.state !== "closed") {
-    await sharedContext.close();
-  }
-  sharedContext = null;
-  activePipelines = 0;
-}

package/src/extensibility/plugins.ts

@@ -1,45 +0,0 @@
-import type { NoiseSuppressionPlugin, VADPlugin } from "../types.js";
-import { RNNoisePlugin } from "../noise-suppression/rnnoise-node.js";
-import { EnergyVADPlugin } from "../vad/vad-node.js";
-
-const nsPlugins = new Map<string, NoiseSuppressionPlugin>();
-const vadPlugins = new Map<string, VADPlugin>();
-
-// Register defaults
-const defaultNs = new RNNoisePlugin();
-nsPlugins.set(defaultNs.name, defaultNs);
-
-const defaultVad = new EnergyVADPlugin();
-vadPlugins.set(defaultVad.name, defaultVad);
-
-export function registerNoiseSuppressionPlugin(plugin: NoiseSuppressionPlugin) {
-  nsPlugins.set(plugin.name, plugin);
-}
-
-export function registerVADPlugin(plugin: VADPlugin) {
-  vadPlugins.set(plugin.name, plugin);
-}
-
-export function getNoiseSuppressionPlugin(
-  name?: string,
-): NoiseSuppressionPlugin {
-  if (!name) return defaultNs;
-  const plugin = nsPlugins.get(name);
-  if (!plugin) {
-    console.warn(
-      `Noise suppression plugin '${name}' not found, falling back to default.`,
-    );
-    return defaultNs;
-  }
-  return plugin;
-}
-
-export function getVADPlugin(name?: string): VADPlugin {
-  if (!name) return defaultVad;
-  const plugin = vadPlugins.get(name);
-  if (!plugin) {
-    console.warn(`VAD plugin '${name}' not found, falling back to default.`);
-    return defaultVad;
-  }
-  return plugin;
-}

package/src/index.ts

@@ -1,8 +0,0 @@
-export * from "./types.js";
-export * from "./context/audio-context.js";
-export * from "./pipeline/audio-pipeline.js";
-export * from "./livekit/integration.js";
-export * from "./extensibility/plugins.js";
-export * from "./noise-suppression/rnnoise-node.js";
-export * from "./vad/vad-node.js";
-export * from "./vad/vad-state.js";

package/src/livekit/integration.ts

@@ -1,61 +0,0 @@
-import type { LocalAudioTrack } from "livekit-client";
-import { createAudioPipeline } from "../pipeline/audio-pipeline.js";
-import type { AudioPipelineHandle, AudioProcessingConfig } from "../types.js";
-
-/**
- * Attaches the audio processing pipeline to a LiveKit LocalAudioTrack.
- * This replaces the underlying MediaStreamTrack with the processed one.
- */
-export async function attachProcessingToTrack(
-  track: LocalAudioTrack,
-  config: AudioProcessingConfig = {},
-): Promise<AudioPipelineHandle> {
-  // 1. Get the original track
-  const originalTrack = track.mediaStreamTrack;
-
-  // 2. Create pipeline
-  const pipeline = await createAudioPipeline(originalTrack, config);
-
-  // 3. Replace the track in LiveKit
-  // Use replaceTrack which is the public API to swap the underlying MediaStreamTrack.
-  await track.replaceTrack(pipeline.processedTrack);
-
-  // 4. Handle intelligent muting if enabled
-  if (config.livekit?.manageTrackMute) {
-    let isVadMuted = false;
-
-    pipeline.events.on("vadChange", async (state) => {
-      if (state.isSpeaking) {
-        if (isVadMuted) {
-          // Only unmute if we were the ones who muted it
-          // And check if the track is not globally muted by user?
-          // This is tricky. If user muted manually, track.isMuted is true.
-          // We should probably check a separate flag or assume VAD overrides only when "active".
-          // For safety, we only unmute if we muted.
-          await track.unmute();
-          isVadMuted = false;
-        }
-      } else {
-        // Silence
-        if (!track.isMuted) {
-          await track.mute();
-          isVadMuted = true;
-        }
-      }
-    });
-  }
-
-  // 5. Handle cleanup
-  const originalDispose = pipeline.dispose;
-  pipeline.dispose = () => {
-    // Restore original track?
-    // Or just stop.
-    // If we dispose, we should probably try to restore the original track if it's still alive.
-    if (originalTrack.readyState === "live") {
-      track.replaceTrack(originalTrack).catch(console.error);
-    }
-    originalDispose();
-  };
-
-  return pipeline;
-}

package/src/noise-suppression/rnnoise-node.ts

@@ -1,62 +0,0 @@
-import {
-  RnnoiseWorkletNode,
-  loadRnnoise,
-} from "@sapphi-red/web-noise-suppressor";
-import type {
-  AudioProcessingConfig,
-  NoiseSuppressionPlugin,
-} from "../types.js";
-
-// Default URLs (can be overridden by config)
-// These defaults assume the assets are served from the same origin or a known CDN.
-// In a real package, we might want to bundle them or require the user to provide them.
-const DEFAULT_WASM_URL =
-  "https://unpkg.com/@sapphi-red/web-noise-suppressor@0.3.5/dist/rnnoise.wasm";
-const DEFAULT_SIMD_WASM_URL =
-  "https://unpkg.com/@sapphi-red/web-noise-suppressor@0.3.5/dist/rnnoise_simd.wasm";
-const DEFAULT_WORKLET_URL =
-  "https://unpkg.com/@sapphi-red/web-noise-suppressor@0.3.5/dist/noise-suppressor-worklet.min.js";
-
-export class RNNoisePlugin implements NoiseSuppressionPlugin {
-  name = "rnnoise-ns";
-  private wasmBuffer: ArrayBuffer | null = null;
-
-  async createNode(
-    context: AudioContext,
-    config: AudioProcessingConfig["noiseSuppression"],
-  ): Promise<AudioNode> {
-    if (!config?.enabled) {
-      // Return a passthrough gain node if disabled but requested (though pipeline usually handles this)
-      const pass = context.createGain();
-      return pass;
-    }
-
-    // 1. Load WASM if not loaded
-    // We use the library's loader which handles SIMD detection if we provide both URLs.
-    // But wait, loadRnnoise returns ArrayBuffer.
-    if (!this.wasmBuffer) {
-      this.wasmBuffer = await loadRnnoise({
-        url: config.wasmUrl || DEFAULT_WASM_URL,
-        simdUrl: DEFAULT_SIMD_WASM_URL, // We should probably allow config for this too, but for now default is fine.
-      });
-    }
-
-    // 2. Load Worklet
-    const workletUrl = config.workletUrl || DEFAULT_WORKLET_URL;
-
-    try {
-      await context.audioWorklet.addModule(workletUrl);
-    } catch (e) {
-      console.warn("Failed to add RNNoise worklet module:", e);
-      // Proceeding, assuming it might be already loaded.
-    }
-
-    // 3. Create Node
-    const node = new RnnoiseWorkletNode(context, {
-      wasmBinary: this.wasmBuffer,
-      maxChannels: 1, // Mono for now
-    });
-
-    return node;
-  }
-}