mellon 0.0.18 → 0.0.19

package/README.md

@@ -17,7 +17,7 @@ Offline, fully in-browser **hotword / wake-word detection** powered by [Efficien
 2. [Quick start](#quick-start)
 3. [Enrolling words](#enrolling-custom-words)
 4. [API reference](#api-reference)
-   - [Mellon](#mellon)
+   - [Detector](#detector)
    - [EnrollmentSession](#enrollmentsession)
 5. [Science behind the lib](#science-behind-the-lib)
 ---
@@ -31,9 +31,9 @@ npm install mellon
 ## Quick start
 
 ```js
-import { Mellon } from 'mellon'
+import { Detector } from 'mellon'
 
-const hotWordDetection = new Mellon([
+const hotWordDetection = new Detector([
   {
     name: 'openDoors',
     triggers: [{ name: 'mellon', defaultRefPath: '/mellon-assets/mellon_ref.json' }],
@@ -71,9 +71,9 @@ await hotWordDetection.start() // opens the mic and listens for all registered t
 ## Enrolling custom words
 
 ```js
-import { Mellon, EnrollmentSession } from 'mellon'
+import { Detector, EnrollmentSession, Storage } from 'mellon'
 
-const hotwordDetection = new Mellon([{
+const hotwordDetection = new Detector([{
   name: 'startEngine',
   triggers: [{ name: 'start' }],
   onMatch: (triggerNameMatched, confidence) => { console.log('starting engine...') }
@@ -99,19 +99,19 @@ hotwordDetection.addCustomWord(ref)
 await hotwordDetection.start()
 
 // 4b. Persist for future sessions
-Mellon.saveWord(ref)
+Storage.saveWord(ref)
 ```
 
 ---
 
 ## API reference
 
-### `Mellon` 
+### `Detector` 
 
 The easiest way to use the library. Wraps mic access, AudioWorklet wiring, and detector management into a single class.
 
 ```ts
-class Mellon {
+class Detector {
   constructor(commands: Command[], config?: MellonConfig)
   readonly threshold:  number   // read/write; persisted in localStorage
   readonly listening:  boolean
@@ -121,7 +121,19 @@ class Mellon {
   stop(): Promise<void>
   addCustomWord(ref: WordRef): void
 
-  // Storage helpers — static, work without a Mellon instance
+  // Storage helpers — static, work without a Detector instance
+  static loadWords(storageKey?: string): WordRef[]
+  static saveWord(ref: WordRef, storageKey?: string): void
+  static deleteWord(wordName: string, storageKey?: string): void
+}
+```
+
+### `Storage`
+
+Static helpers for persisting enrolled word references in `localStorage`.
+
+```ts
+class Storage {
   static loadWords(storageKey?: string): WordRef[]
   static saveWord(ref: WordRef, storageKey?: string): void
   static deleteWord(wordName: string, storageKey?: string): void

package/dist/index.d.ts

@@ -1,5 +1,7 @@
-import { Mellon } from './Mellon';
+import { Detector } from './Mellon';
 import { EnrollmentSession } from './EnrollmentSession';
+import { Storage } from './Storage';
+import { AudioUtils } from './AudioUtils';
 export type TriggerName = string;
 export interface Trigger {
     name: TriggerName;
@@ -17,11 +19,13 @@ export interface MellonConfig {
     modelPath?: string;
     audioProcessorPath?: string;
     ortCdnUrl?: string;
+    audioUtils?: AudioUtils;
 }
 export interface EnrollmentSessionConfig {
     wasmPaths?: string;
     modelPath?: string;
     ortCdnUrl?: string;
+    audioUtils?: AudioUtils;
 }
 export interface WordRef {
     word_name: TriggerName;
@@ -34,4 +38,4 @@ export declare const DEFAULT_MODEL_PATH = "https://huggingface.co/ComicScrip/mel
 export declare const DEFAULT_AUDIO_PROCESSOR_PATH = "https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js";
 export declare const DEFAULT_REFS_STORAGE_KEY = "mellon-refs";
 export declare const DEFAULT_THRESHOLD_STORAGE_KEY = "mellon-threshold";
-export { Mellon, EnrollmentSession };
+export { Detector, EnrollmentSession, Storage, AudioUtils };

package/dist/mellon.cjs

@@ -1 +1 @@
-"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function Dt(m){return m&&m.__esModule&&Object.prototype.hasOwnProperty.call(m,"default")?m.default:m}var lt,ut;function Ut(){if(ut)return lt;ut=1;function m(o){if(this.size=o|0,this.size<=1||(this.size&this.size-1)!==0)throw new Error("FFT size must be a power of two and bigger than 1");this._csize=o<<1;for(var t=new Array(this.size*2),s=0;s<t.length;s+=2){const i=Math.PI*s/this.size;t[s]=Math.cos(i),t[s+1]=-Math.sin(i)}this.table=t;for(var n=0,r=1;this.size>r;r<<=1)n++;this._width=n%2===0?n-1:n,this._bitrev=new Array(1<<this._width);for(var a=0;a<this._bitrev.length;a++){this._bitrev[a]=0;for(var e=0;e<this._width;e+=2){var h=this._width-e-2;this._bitrev[a]|=(a>>>e&3)<<h}}this._out=null,this._data=null,this._inv=0}return lt=m,m.prototype.fromComplexArray=function(t,s){for(var n=s||new Array(t.length>>>1),r=0;r<t.length;r+=2)n[r>>>1]=t[r];return n},m.prototype.createComplexArray=function(){const t=new Array(this._csize);for(var s=0;s<t.length;s++)t[s]=0;return t},m.prototype.toComplexArray=function(t,s){for(var n=s||this.createComplexArray(),r=0;r<n.length;r+=2)n[r]=t[r>>>1],n[r+1]=0;return n},m.prototype.completeSpectrum=function(t){for(var s=this._csize,n=s>>>1,r=2;r<n;r+=2)t[s-r]=t[r],t[s-r+1]=-t[r+1]},m.prototype.transform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._transform4(),this._out=null,this._data=null},m.prototype.realTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._realTransform4(),this._out=null,this._data=null},m.prototype.inverseTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=1,this._transform4();for(var n=0;n<t.length;n++)t[n]/=this.size;this._out=null,this._data=null},m.prototype._transform4=function(){var t=this._out,s=this._csize,n=this._width,r=1<<n,a=s/r<<1,e,h,i=this._bitrev;if(a===4)for(e=0,h=0;e<s;e+=a,h++){const v=i[h];this._singleTransform2(e,v,r)}else for(e=0,h=0;e<s;e+=a,h++){const v=i[h];this._singleTransform4(e,v,r)}var l=this._inv?-1:1,c=this.table;for(r>>=2;r>=2;r>>=2){a=s/r<<1;var d=a>>>2;for(e=0;e<s;e+=a)for(var _=e+d,u=e,f=0;u<_;u+=2,f+=r){const v=u,p=v+d,g=p+d,w=g+d,T=t[v],F=t[v+1],A=t[p],y=t[p+1],b=t[g],S=t[g+1],P=t[w],C=t[w+1],E=T,R=F,M=c[f],D=l*c[f+1],U=A*M-y*D,z=A*D+y*M,j=c[2*f],L=l*c[2*f+1],B=b*j-S*L,K=b*L+S*j,N=c[3*f],O=l*c[3*f+1],k=P*N-C*O,G=P*O+C*N,$=E+B,x=R+K,I=E-B,Y=R-K,J=U+k,H=z+G,W=l*(U-k),Q=l*(z-G),tt=$+J,rt=x+H,et=$-J,ot=x-H,nt=I+Q,at=Y-W,it=I-Q,ct=Y+W;t[v]=tt,t[v+1]=rt,t[p]=nt,t[p+1]=at,t[g]=et,t[g+1]=ot,t[w]=it,t[w+1]=ct}}},m.prototype._singleTransform2=function(t,s,n){const r=this._out,a=this._data,e=a[s],h=a[s+1],i=a[s+n],l=a[s+n+1],c=e+i,d=h+l,_=e-i,u=h-l;r[t]=c,r[t+1]=d,r[t+2]=_,r[t+3]=u},m.prototype._singleTransform4=function(t,s,n){const r=this._out,a=this._data,e=this._inv?-1:1,h=n*2,i=n*3,l=a[s],c=a[s+1],d=a[s+n],_=a[s+n+1],u=a[s+h],f=a[s+h+1],v=a[s+i],p=a[s+i+1],g=l+u,w=c+f,T=l-u,F=c-f,A=d+v,y=_+p,b=e*(d-v),S=e*(_-p),P=g+A,C=w+y,E=T+S,R=F-b,M=g-A,D=w-y,U=T-S,z=F+b;r[t]=P,r[t+1]=C,r[t+2]=E,r[t+3]=R,r[t+4]=M,r[t+5]=D,r[t+6]=U,r[t+7]=z},m.prototype._realTransform4=function(){var t=this._out,s=this._csize,n=this._width,r=1<<n,a=s/r<<1,e,h,i=this._bitrev;if(a===4)for(e=0,h=0;e<s;e+=a,h++){const ht=i[h];this._singleRealTransform2(e,ht>>>1,r>>>1)}else for(e=0,h=0;e<s;e+=a,h++){const ht=i[h];this._singleRealTransform4(e,ht>>>1,r>>>1)}var l=this._inv?-1:1,c=this.table;for(r>>=2;r>=2;r>>=2){a=s/r<<1;var d=a>>>1,_=d>>>1,u=_>>>1;for(e=0;e<s;e+=a)for(var f=0,v=0;f<=u;f+=2,v+=r){var p=e+f,g=p+_,w=g+_,T=w+_,F=t[p],A=t[p+1],y=t[g],b=t[g+1],S=t[w],P=t[w+1],C=t[T],E=t[T+1],R=F,M=A,D=c[v],U=l*c[v+1],z=y*D-b*U,j=y*U+b*D,L=c[2*v],B=l*c[2*v+1],K=S*L-P*B,N=S*B+P*L,O=c[3*v],k=l*c[3*v+1],G=C*O-E*k,$=C*k+E*O,x=R+K,I=M+N,Y=R-K,J=M-N,H=z+G,W=j+$,Q=l*(z-G),tt=l*(j-$),rt=x+H,et=I+W,ot=Y+tt,nt=J-Q;if(t[p]=rt,t[p+1]=et,t[g]=ot,t[g+1]=nt,f===0){var at=x-H,it=I-W;t[w]=at,t[w+1]=it;continue}if(f!==u){var ct=Y,yt=-J,Tt=x,At=-I,Ft=-l*tt,bt=-l*Q,St=-l*W,Pt=-l*H,Ct=ct+Ft,Et=yt+bt,Rt=Tt+Pt,Mt=At-St,ft=e+_-f,vt=e+d-f;t[ft]=Ct,t[ft+1]=Et,t[vt]=Rt,t[vt+1]=Mt}}}},m.prototype._singleRealTransform2=function(t,s,n){const r=this._out,a=this._data,e=a[s],h=a[s+n],i=e+h,l=e-h;r[t]=i,r[t+1]=0,r[t+2]=l,r[t+3]=0},m.prototype._singleRealTransform4=function(t,s,n){const r=this._out,a=this._data,e=this._inv?-1:1,h=n*2,i=n*3,l=a[s],c=a[s+n],d=a[s+h],_=a[s+i],u=l+d,f=l-d,v=c+_,p=e*(c-_),g=u+v,w=f,T=-p,F=u-v,A=f,y=p;r[t]=g,r[t+1]=0,r[t+2]=w,r[t+3]=T,r[t+4]=F,r[t+5]=0,r[t+6]=A,r[t+7]=y},lt}var zt=Ut();const xt=Dt(zt);class pt{constructor(o=16e3,t=512,s=64){this.sampleRate=o,this.nfft=t,this.nfilt=s,this.fft=new xt(t),this.melFilters=this.createMelFilterbank()}hzToMel(o){return 2595*Math.log10(1+o/700)}melToHz(o){return 700*(10**(o/2595)-1)}createMelFilterbank(){const t=this.sampleRate/2,s=this.hzToMel(0),n=this.hzToMel(t),r=new Float32Array(this.nfilt+2);for(let i=0;i<this.nfilt+2;i++)r[i]=s+i*(n-s)/(this.nfilt+1);const e=r.map(i=>this.melToHz(i)).map(i=>Math.floor((this.nfft+1)*i/this.sampleRate)),h=[];for(let i=0;i<this.nfilt;i++){const l=new Float32Array(Math.floor(this.nfft/2)+1);for(let c=e[i];c<e[i+1];c++)l[c]=(c-e[i])/(e[i+1]-e[i]);for(let c=e[i+1];c<e[i+2];c++)l[c]=(e[i+2]-c)/(e[i+2]-e[i+1]);h.push(l)}return h}logfbank(o){const t=Math.floor(.025*this.sampleRate),s=Math.floor(.01*this.sampleRate),n=1+Math.ceil((o.length-t)/s),r=new Float32Array(n*this.nfilt),a=new Float32Array(this.nfft),e=this.fft.createComplexArray();for(let h=0;h<n;h++){const i=h*s;a.fill(0);for(let d=0;d<t&&i+d<o.length;d++)a[d]=o[i+d];const l=this.fft.toComplexArray(a,null);this.fft.transform(e,l);const c=new Float32Array(Math.floor(this.nfft/2)+1);for(let d=0;d<c.length;d++){const _=e[2*d],u=e[2*d+1];c[d]=1/this.nfft*(_*_+u*u),c[d]===0&&(c[d]=1e-30)}for(let d=0;d<this.nfilt;d++){let _=0;const u=this.melFilters[d];for(let f=0;f<c.length;f++)_+=c[f]*u[f];_===0&&(_=1e-30),r[h*this.nfilt+d]=Math.log(_)}}return r}}async function mt(m=Z,o=q){const t=await import(o);return t.env.wasm.wasmPaths=m,t.env.wasm.numThreads=1,t}let dt=null;async function _t(m=st,o=Z,t=q){return dt||(dt=mt(o,t).then(s=>s.InferenceSession.create(m,{executionProviders:["wasm"],graphOptimizationLevel:"all"}))),dt}const It=new pt;class X{constructor(o,t){this._started=!1,this._inferring=!1,this._audioCtx=null,this._stream=null,this._refEmbeddings=new Map,this._lastMatchAt=0,this._lastInferenceAt=0;const{refsStorageKey:s=V,thresholdStorageKey:n=wt,wasmPaths:r=Z,modelPath:a=st,audioProcessorPath:e=gt,ortCdnUrl:h=q}=t||{};this._commands=o,this._refsStorageKey=s,this._thresholdStorageKey=n,this._audioProcessorPath=e,this._wasmPaths=r,this._modelPath=a,this._ortCdnUrl=h;try{const i=localStorage.getItem(this._thresholdStorageKey);this._threshold=i!==null?Math.max(0,Math.min(1,Number(i))):.65}catch{this._threshold=.65}this._initPromise=this._init()}get threshold(){return this._threshold}set threshold(o){this._threshold=Math.max(0,Math.min(1,o));try{localStorage.setItem(this._thresholdStorageKey,String(this._threshold))}catch{}}get listening(){return this._started}async _init(){await _t(this._modelPath,this._wasmPaths,this._ortCdnUrl);const o=new Set;for(const t of this._commands)for(const s of t.triggers)if(!o.has(s.name)&&(o.add(s.name),s.defaultRefPath)){const n=await fetch(s.defaultRefPath);if(n.ok){const r=await n.json();this.addCustomWord(r)}}for(const t of X.loadWords(this._refsStorageKey))this._refEmbeddings.set(t.word_name,t.embeddings);console.info("[Mellon] init complete, loaded refs:",[...this._refEmbeddings.keys()])}async init(){await this._initPromise}addCustomWord(o){if(!(Array.isArray(o.embeddings)&&o.embeddings.length>0))throw new Error("invalid ref file for : "+o.word_name);this._refEmbeddings.set(o.word_name,o.embeddings)}async start(){if(this._started)return;await this._initPromise;let o;try{o=await navigator.mediaDevices.getUserMedia({audio:{noiseSuppression:!1,echoCancellation:!1,autoGainControl:!1,channelCount:1}})}catch{o=await navigator.mediaDevices.getUserMedia({audio:!0})}this._stream=o;const t=new AudioContext({sampleRate:16e3});this._audioCtx=t,await t.audioWorklet.addModule(this._audioProcessorPath);const s=t.createMediaStreamSource(o),n=new AudioWorkletNode(t,"audio-processor");n.port.onmessage=r=>{this._handleBuffer(r.data)},s.connect(n),n.connect(t.destination),this._started=!0}async stop(){if(this._started=!1,this._audioCtx&&(await this._audioCtx.close(),this._audioCtx=null),this._stream){for(const o of this._stream.getTracks())o.stop();this._stream=null}}async _handleBuffer(o){if(this._inferring)return;const t=Date.now();if(!(t-this._lastInferenceAt<300)){this._lastInferenceAt=t,this._inferring=!0;try{const[s,n]=await Promise.all([mt(this._wasmPaths,this._ortCdnUrl),_t(this._modelPath,this._wasmPaths,this._ortCdnUrl)]),r=It.logfbank(o),a=new s.Tensor("float32",r,[1,1,149,64]),e=await n.run({input:a}),h=e[Object.keys(e)[0]].data;let i=!1;for(const l of this._commands){if(i)break;for(const c of l.triggers){const d=this._refEmbeddings.get(c.name);if(!d)continue;const _=this._maxCosineSim(h,d);if(_>=this._threshold&&t-this._lastMatchAt>2e3){this._lastMatchAt=t,console.info(`[Mellon] match: "${c}" sim=${_.toFixed(3)}`),typeof l.onMatch=="function"&&l.onMatch(c.name,_),i=!0;break}}}}catch(s){console.error("[Mellon] inference error:",s)}finally{this._inferring=!1}}}_maxCosineSim(o,t){let s=0;for(const n of t){let r=0;for(let e=0;e<n.length;e++)r+=o[e]*n[e];const a=(r+1)/2;a>s&&(s=a)}return s}static loadWords(o=V){try{const t=localStorage.getItem(o);return t?JSON.parse(t):[]}catch{return[]}}static saveWord(o,t=V){const s=X.loadWords(t).filter(n=>n.word_name!==o.word_name);localStorage.setItem(t,JSON.stringify([...s,o]))}static deleteWord(o,t=V){try{const s=X.loadWords(t).filter(n=>n.word_name!==o);localStorage.setItem(t,JSON.stringify(s))}catch{}}}const Ht=new pt;class Wt{constructor(o,t){this._config={},this._samples=[],this._wordName=o,this._config.modelPath=(t==null?void 0:t.modelPath)||st,this._config.wasmPaths=(t==null?void 0:t.wasmPaths)||Z,this._config.ortCdnUrl=(t==null?void 0:t.ortCdnUrl)||q}async recordSample(){const o=await navigator.mediaDevices.getUserMedia({audio:!0}),t=new AudioContext({sampleRate:16e3}),s=await new Promise((a,e)=>{const h=new MediaRecorder(o),i=[];h.ondataavailable=l=>{l.data.size>0&&i.push(l.data)},h.onstop=async()=>{var l;for(const c of o.getTracks())c.stop();try{const d=await new Blob(i,{type:((l=i[0])==null?void 0:l.type)||"audio/webm"}).arrayBuffer(),_=await t.decodeAudioData(d);await t.close(),a(_.getChannelData(0).slice())}catch(c){e(c)}},h.start(),setTimeout(()=>{try{h.stop()}catch{}},1500)}),n=24e3,r=new Float32Array(n);return r.set(s.slice(0,n)),this._samples.push(r),this._samples.length}deleteSample(o){if(o<0||o>=this._samples.length)throw new RangeError(`index ${o} out of bounds (${this._samples.length} samples)`);return this._samples.splice(o,1),this._samples.length}async generateRef(){const[o,t]=await Promise.all([mt(this._config.wasmPaths,this._config.ortCdnUrl),_t(this._config.modelPath,this._config.wasmPaths,this._config.ortCdnUrl)]),s=[];for(const n of this._samples){const r=Ht.logfbank(n),a=new o.Tensor("float32",r,[1,1,149,64]),e=await t.run({input:a}),h=Array.from(e[Object.keys(e)[0]].data);s.push(h)}return{word_name:this._wordName,model_type:"resnet_50_arc",embeddings:s}}}const Z="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/",q="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs",st="https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx",gt="https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js",V="mellon-refs",wt="mellon-threshold";exports.DEFAULT_AUDIO_PROCESSOR_PATH=gt;exports.DEFAULT_MODEL_PATH=st;exports.DEFAULT_ORT_CDN_URL=q;exports.DEFAULT_REFS_STORAGE_KEY=V;exports.DEFAULT_THRESHOLD_STORAGE_KEY=wt;exports.DEFAULT_WASM_PATHS=Z;exports.EnrollmentSession=Wt;exports.Mellon=X;
+"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});function Dt(m){return m&&m.__esModule&&Object.prototype.hasOwnProperty.call(m,"default")?m.default:m}var ht,pt;function Ut(){if(pt)return ht;pt=1;function m(o){if(this.size=o|0,this.size<=1||(this.size&this.size-1)!==0)throw new Error("FFT size must be a power of two and bigger than 1");this._csize=o<<1;for(var t=new Array(this.size*2),s=0;s<t.length;s+=2){const i=Math.PI*s/this.size;t[s]=Math.cos(i),t[s+1]=-Math.sin(i)}this.table=t;for(var n=0,r=1;this.size>r;r<<=1)n++;this._width=n%2===0?n-1:n,this._bitrev=new Array(1<<this._width);for(var a=0;a<this._bitrev.length;a++){this._bitrev[a]=0;for(var e=0;e<this._width;e+=2){var l=this._width-e-2;this._bitrev[a]|=(a>>>e&3)<<l}}this._out=null,this._data=null,this._inv=0}return ht=m,m.prototype.fromComplexArray=function(t,s){for(var n=s||new Array(t.length>>>1),r=0;r<t.length;r+=2)n[r>>>1]=t[r];return n},m.prototype.createComplexArray=function(){const t=new Array(this._csize);for(var s=0;s<t.length;s++)t[s]=0;return t},m.prototype.toComplexArray=function(t,s){for(var n=s||this.createComplexArray(),r=0;r<n.length;r+=2)n[r]=t[r>>>1],n[r+1]=0;return n},m.prototype.completeSpectrum=function(t){for(var s=this._csize,n=s>>>1,r=2;r<n;r+=2)t[s-r]=t[r],t[s-r+1]=-t[r+1]},m.prototype.transform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._transform4(),this._out=null,this._data=null},m.prototype.realTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=0,this._realTransform4(),this._out=null,this._data=null},m.prototype.inverseTransform=function(t,s){if(t===s)throw new Error("Input and output buffers must be different");this._out=t,this._data=s,this._inv=1,this._transform4();for(var n=0;n<t.length;n++)t[n]/=this.size;this._out=null,this._data=null},m.prototype._transform4=function(){var t=this._out,s=this._csize,n=this._width,r=1<<n,a=s/r<<1,e,l,i=this._bitrev;if(a===4)for(e=0,l=0;e<s;e+=a,l++){const u=i[l];this._singleTransform2(e,u,r)}else for(e=0,l=0;e<s;e+=a,l++){const u=i[l];this._singleTransform4(e,u,r)}var h=this._inv?-1:1,c=this.table;for(r>>=2;r>=2;r>>=2){a=s/r<<1;var _=a>>>2;for(e=0;e<s;e+=a)for(var d=e+_,v=e,f=0;v<d;v+=2,f+=r){const u=v,p=u+_,w=p+_,g=w+_,T=t[u],F=t[u+1],A=t[p],y=t[p+1],b=t[w],S=t[w+1],P=t[g],C=t[g+1],E=T,M=F,R=c[f],D=h*c[f+1],U=A*R-y*D,z=A*D+y*R,j=c[2*f],L=h*c[2*f+1],B=b*j-S*L,K=b*L+S*j,N=c[3*f],O=h*c[3*f+1],k=P*N-C*O,G=P*O+C*N,Y=E+B,x=M+K,I=E-B,$=M-K,J=U+k,H=z+G,W=h*(U-k),Q=h*(z-G),tt=Y+J,rt=x+H,et=Y-J,ot=x-H,nt=I+Q,at=$-W,it=I-Q,ct=$+W;t[u]=tt,t[u+1]=rt,t[p]=nt,t[p+1]=at,t[w]=et,t[w+1]=ot,t[g]=it,t[g+1]=ct}}},m.prototype._singleTransform2=function(t,s,n){const r=this._out,a=this._data,e=a[s],l=a[s+1],i=a[s+n],h=a[s+n+1],c=e+i,_=l+h,d=e-i,v=l-h;r[t]=c,r[t+1]=_,r[t+2]=d,r[t+3]=v},m.prototype._singleTransform4=function(t,s,n){const r=this._out,a=this._data,e=this._inv?-1:1,l=n*2,i=n*3,h=a[s],c=a[s+1],_=a[s+n],d=a[s+n+1],v=a[s+l],f=a[s+l+1],u=a[s+i],p=a[s+i+1],w=h+v,g=c+f,T=h-v,F=c-f,A=_+u,y=d+p,b=e*(_-u),S=e*(d-p),P=w+A,C=g+y,E=T+S,M=F-b,R=w-A,D=g-y,U=T-S,z=F+b;r[t]=P,r[t+1]=C,r[t+2]=E,r[t+3]=M,r[t+4]=R,r[t+5]=D,r[t+6]=U,r[t+7]=z},m.prototype._realTransform4=function(){var t=this._out,s=this._csize,n=this._width,r=1<<n,a=s/r<<1,e,l,i=this._bitrev;if(a===4)for(e=0,l=0;e<s;e+=a,l++){const lt=i[l];this._singleRealTransform2(e,lt>>>1,r>>>1)}else for(e=0,l=0;e<s;e+=a,l++){const lt=i[l];this._singleRealTransform4(e,lt>>>1,r>>>1)}var h=this._inv?-1:1,c=this.table;for(r>>=2;r>=2;r>>=2){a=s/r<<1;var _=a>>>1,d=_>>>1,v=d>>>1;for(e=0;e<s;e+=a)for(var f=0,u=0;f<=v;f+=2,u+=r){var p=e+f,w=p+d,g=w+d,T=g+d,F=t[p],A=t[p+1],y=t[w],b=t[w+1],S=t[g],P=t[g+1],C=t[T],E=t[T+1],M=F,R=A,D=c[u],U=h*c[u+1],z=y*D-b*U,j=y*U+b*D,L=c[2*u],B=h*c[2*u+1],K=S*L-P*B,N=S*B+P*L,O=c[3*u],k=h*c[3*u+1],G=C*O-E*k,Y=C*k+E*O,x=M+K,I=R+N,$=M-K,J=R-N,H=z+G,W=j+Y,Q=h*(z-G),tt=h*(j-Y),rt=x+H,et=I+W,ot=$+tt,nt=J-Q;if(t[p]=rt,t[p+1]=et,t[w]=ot,t[w+1]=nt,f===0){var at=x-H,it=I-W;t[g]=at,t[g+1]=it;continue}if(f!==v){var ct=$,yt=-J,Tt=x,At=-I,Ft=-h*tt,bt=-h*Q,St=-h*W,Pt=-h*H,Ct=ct+Ft,Et=yt+bt,Mt=Tt+Pt,Rt=At-St,ut=e+d-f,vt=e+_-f;t[ut]=Ct,t[ut+1]=Et,t[vt]=Mt,t[vt+1]=Rt}}}},m.prototype._singleRealTransform2=function(t,s,n){const r=this._out,a=this._data,e=a[s],l=a[s+n],i=e+l,h=e-l;r[t]=i,r[t+1]=0,r[t+2]=h,r[t+3]=0},m.prototype._singleRealTransform4=function(t,s,n){const r=this._out,a=this._data,e=this._inv?-1:1,l=n*2,i=n*3,h=a[s],c=a[s+n],_=a[s+l],d=a[s+i],v=h+_,f=h-_,u=c+d,p=e*(c-d),w=v+u,g=f,T=-p,F=v-u,A=f,y=p;r[t]=w,r[t+1]=0,r[t+2]=g,r[t+3]=T,r[t+4]=F,r[t+5]=0,r[t+6]=A,r[t+7]=y},ht}var zt=Ut();const xt=Dt(zt);class mt{constructor(o=16e3,t=512,s=64){this._sampleRate=o,this._nfft=t,this._nfilt=s,this._fft=new xt(t),this._melFilters=this._createMelFilterbank()}_hzToMel(o){return 2595*Math.log10(1+o/700)}_melToHz(o){return 700*(10**(o/2595)-1)}_createMelFilterbank(){const t=this._sampleRate/2,s=this._hzToMel(0),n=this._hzToMel(t),r=new Float32Array(this._nfilt+2);for(let i=0;i<this._nfilt+2;i++)r[i]=s+i*(n-s)/(this._nfilt+1);const e=r.map(i=>this._melToHz(i)).map(i=>Math.floor((this._nfft+1)*i/this._sampleRate)),l=[];for(let i=0;i<this._nfilt;i++){const h=new Float32Array(Math.floor(this._nfft/2)+1);for(let c=e[i];c<e[i+1];c++)h[c]=(c-e[i])/(e[i+1]-e[i]);for(let c=e[i+1];c<e[i+2];c++)h[c]=(e[i+2]-c)/(e[i+2]-e[i+1]);l.push(h)}return l}logfbank(o){const t=Math.floor(.025*this._sampleRate),s=Math.floor(.01*this._sampleRate),n=1+Math.ceil((o.length-t)/s),r=new Float32Array(n*this._nfilt),a=new Float32Array(this._nfft),e=this._fft.createComplexArray();for(let l=0;l<n;l++){const i=l*s;a.fill(0);for(let _=0;_<t&&i+_<o.length;_++)a[_]=o[i+_];const h=this._fft.toComplexArray(a,null);this._fft.transform(e,h);const c=new Float32Array(Math.floor(this._nfft/2)+1);for(let _=0;_<c.length;_++){const d=e[2*_],v=e[2*_+1];c[_]=1/this._nfft*(d*d+v*v),c[_]===0&&(c[_]=1e-30)}for(let _=0;_<this._nfilt;_++){let d=0;const v=this._melFilters[_];for(let f=0;f<c.length;f++)d+=c[f]*v[f];d===0&&(d=1e-30),r[l*this._nfilt+_]=Math.log(d)}}return r}maxCosineSim(o,t){let s=0;for(const n of t){let r=0;for(let e=0;e<n.length;e++)r+=o[e]*n[e];const a=(r+1)/2;a>s&&(s=a)}return s}}async function ft(m=Z,o=q){const t=await import(o);return t.env.wasm.wasmPaths=m,t.env.wasm.numThreads=1,t}let _t=null;async function dt(m=st,o=Z,t=q){return _t||(_t=ft(o,t).then(s=>s.InferenceSession.create(m,{executionProviders:["wasm"],graphOptimizationLevel:"all"}))),_t}class X{static loadWords(o=V){try{const t=localStorage.getItem(o);return t?JSON.parse(t):[]}catch{return[]}}static saveWord(o,t=V){const s=X.loadWords(t).filter(n=>n.word_name!==o.word_name);localStorage.setItem(t,JSON.stringify([...s,o]))}static deleteWord(o,t=V){try{const s=X.loadWords(t).filter(n=>n.word_name!==o);localStorage.setItem(t,JSON.stringify(s))}catch{}}}class It{constructor(o,t){this._started=!1,this._inferring=!1,this._audioCtx=null,this._stream=null,this._refEmbeddings=new Map,this._lastMatchAt=0,this._lastInferenceAt=0;const{refsStorageKey:s=V,thresholdStorageKey:n=gt,wasmPaths:r=Z,modelPath:a=st,audioProcessorPath:e=wt,ortCdnUrl:l=q,audioUtils:i=new mt}=t||{};this._audioUtils=i,this._commands=o,this._refsStorageKey=s,this._thresholdStorageKey=n,this._audioProcessorPath=e,this._wasmPaths=r,this._modelPath=a,this._ortCdnUrl=l;try{const h=localStorage.getItem(this._thresholdStorageKey);this._threshold=h!==null?Math.max(0,Math.min(1,Number(h))):.65}catch{this._threshold=.65}this._initPromise=this._init()}get threshold(){return this._threshold}set threshold(o){this._threshold=Math.max(0,Math.min(1,o));try{localStorage.setItem(this._thresholdStorageKey,String(this._threshold))}catch{}}get listening(){return this._started}async _init(){await dt(this._modelPath,this._wasmPaths,this._ortCdnUrl);const o=new Set;for(const t of this._commands)for(const s of t.triggers)if(!o.has(s.name)&&(o.add(s.name),s.defaultRefPath)){const n=await fetch(s.defaultRefPath);if(n.ok){const r=await n.json();this.addCustomWord(r)}}for(const t of X.loadWords(this._refsStorageKey))this._refEmbeddings.set(t.word_name,t.embeddings);console.info("[Mellon] init complete, loaded refs:",[...this._refEmbeddings.keys()])}async init(){await this._initPromise}addCustomWord(o){if(!(Array.isArray(o.embeddings)&&o.embeddings.length>0))throw new Error("invalid ref file for : "+o.word_name);this._refEmbeddings.set(o.word_name,o.embeddings)}async start(){if(this._started)return;await this._initPromise;let o;try{o=await navigator.mediaDevices.getUserMedia({audio:{noiseSuppression:!1,echoCancellation:!1,autoGainControl:!1,channelCount:1}})}catch{o=await navigator.mediaDevices.getUserMedia({audio:!0})}this._stream=o;const t=new AudioContext({sampleRate:16e3});this._audioCtx=t,await t.audioWorklet.addModule(this._audioProcessorPath);const s=t.createMediaStreamSource(o),n=new AudioWorkletNode(t,"audio-processor");n.port.onmessage=r=>{this._handleBuffer(r.data)},s.connect(n),n.connect(t.destination),this._started=!0}async stop(){if(this._started=!1,this._audioCtx&&(await this._audioCtx.close(),this._audioCtx=null),this._stream){for(const o of this._stream.getTracks())o.stop();this._stream=null}}async _handleBuffer(o){if(this._inferring)return;const t=Date.now();if(!(t-this._lastInferenceAt<300)){this._lastInferenceAt=t,this._inferring=!0;try{const[s,n]=await Promise.all([ft(this._wasmPaths,this._ortCdnUrl),dt(this._modelPath,this._wasmPaths,this._ortCdnUrl)]),r=this._audioUtils.logfbank(o),a=new s.Tensor("float32",r,[1,1,149,64]),e=await n.run({input:a}),l=e[Object.keys(e)[0]].data;let i=!1;for(const h of this._commands){if(i)break;for(const c of h.triggers){const _=this._refEmbeddings.get(c.name);if(!_)continue;const d=this._audioUtils.maxCosineSim(l,_);if(d>=this._threshold&&t-this._lastMatchAt>2e3){this._lastMatchAt=t,console.info(`[Mellon] match: "${c}" sim=${d.toFixed(3)}`),typeof h.onMatch=="function"&&h.onMatch(c.name,d),i=!0;break}}}}catch(s){console.error("[Mellon] inference error:",s)}finally{this._inferring=!1}}}}class Ht{constructor(o,t){this._config={},this._samples=[],this._wordName=o,this._config.modelPath=(t==null?void 0:t.modelPath)||st,this._config.wasmPaths=(t==null?void 0:t.wasmPaths)||Z,this._config.ortCdnUrl=(t==null?void 0:t.ortCdnUrl)||q,this._audioUtils=(t==null?void 0:t.audioUtils)??new mt}async recordSample(){const o=await navigator.mediaDevices.getUserMedia({audio:!0}),t=new AudioContext({sampleRate:16e3}),s=await new Promise((a,e)=>{const l=new MediaRecorder(o),i=[];l.ondataavailable=h=>{h.data.size>0&&i.push(h.data)},l.onstop=async()=>{var h;for(const c of o.getTracks())c.stop();try{const _=await new Blob(i,{type:((h=i[0])==null?void 0:h.type)||"audio/webm"}).arrayBuffer(),d=await t.decodeAudioData(_);await t.close(),a(d.getChannelData(0).slice())}catch(c){e(c)}},l.start(),setTimeout(()=>{try{l.stop()}catch{}},1500)}),n=24e3,r=new Float32Array(n);return r.set(s.slice(0,n)),this._samples.push(r),this._samples.length}deleteSample(o){if(o<0||o>=this._samples.length)throw new RangeError(`index ${o} out of bounds (${this._samples.length} samples)`);return this._samples.splice(o,1),this._samples.length}async generateRef(){const[o,t]=await Promise.all([ft(this._config.wasmPaths,this._config.ortCdnUrl),dt(this._config.modelPath,this._config.wasmPaths,this._config.ortCdnUrl)]),s=[];for(const n of this._samples){const r=this._audioUtils.logfbank(n),a=new o.Tensor("float32",r,[1,1,149,64]),e=await t.run({input:a}),l=Array.from(e[Object.keys(e)[0]].data);s.push(l)}return{word_name:this._wordName,model_type:"resnet_50_arc",embeddings:s}}}const Z="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/",q="https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs",st="https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx",wt="https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js",V="mellon-refs",gt="mellon-threshold";exports.AudioUtils=mt;exports.DEFAULT_AUDIO_PROCESSOR_PATH=wt;exports.DEFAULT_MODEL_PATH=st;exports.DEFAULT_ORT_CDN_URL=q;exports.DEFAULT_REFS_STORAGE_KEY=V;exports.DEFAULT_THRESHOLD_STORAGE_KEY=gt;exports.DEFAULT_WASM_PATHS=Z;exports.Detector=It;exports.EnrollmentSession=Ht;exports.Storage=X;

package/dist/mellon.mjs

@@ -1,11 +1,11 @@
-function Rt(_) {
-  return _ && _.__esModule && Object.prototype.hasOwnProperty.call(_, "default") ? _.default : _;
+function Rt(m) {
+  return m && m.__esModule && Object.prototype.hasOwnProperty.call(m, "default") ? m.default : m;
 }
-var ht, ut;
+var ht, vt;
 function Et() {
-  if (ut) return ht;
-  ut = 1;
-  function _(o) {
+  if (vt) return ht;
+  vt = 1;
+  function m(o) {
     if (this.size = o | 0, this.size <= 1 || (this.size & this.size - 1) !== 0)
       throw new Error("FFT size must be a power of two and bigger than 1");
     this._csize = o << 1;
@@ -26,66 +26,66 @@ function Et() {
     }
     this._out = null, this._data = null, this._inv = 0;
   }
-  return ht = _, _.prototype.fromComplexArray = function(t, s) {
+  return ht = m, m.prototype.fromComplexArray = function(t, s) {
     for (var n = s || new Array(t.length >>> 1), r = 0; r < t.length; r += 2)
       n[r >>> 1] = t[r];
     return n;
-  }, _.prototype.createComplexArray = function() {
+  }, m.prototype.createComplexArray = function() {
     const t = new Array(this._csize);
     for (var s = 0; s < t.length; s++)
       t[s] = 0;
     return t;
-  }, _.prototype.toComplexArray = function(t, s) {
+  }, m.prototype.toComplexArray = function(t, s) {
     for (var n = s || this.createComplexArray(), r = 0; r < n.length; r += 2)
       n[r] = t[r >>> 1], n[r + 1] = 0;
     return n;
-  }, _.prototype.completeSpectrum = function(t) {
+  }, m.prototype.completeSpectrum = function(t) {
     for (var s = this._csize, n = s >>> 1, r = 2; r < n; r += 2)
       t[s - r] = t[r], t[s - r + 1] = -t[r + 1];
-  }, _.prototype.transform = function(t, s) {
+  }, m.prototype.transform = function(t, s) {
     if (t === s)
       throw new Error("Input and output buffers must be different");
     this._out = t, this._data = s, this._inv = 0, this._transform4(), this._out = null, this._data = null;
-  }, _.prototype.realTransform = function(t, s) {
+  }, m.prototype.realTransform = function(t, s) {
     if (t === s)
       throw new Error("Input and output buffers must be different");
     this._out = t, this._data = s, this._inv = 0, this._realTransform4(), this._out = null, this._data = null;
-  }, _.prototype.inverseTransform = function(t, s) {
+  }, m.prototype.inverseTransform = function(t, s) {
     if (t === s)
       throw new Error("Input and output buffers must be different");
     this._out = t, this._data = s, this._inv = 1, this._transform4();
     for (var n = 0; n < t.length; n++)
       t[n] /= this.size;
     this._out = null, this._data = null;
-  }, _.prototype._transform4 = function() {
+  }, m.prototype._transform4 = function() {
     var t = this._out, s = this._csize, n = this._width, r = 1 << n, a = s / r << 1, e, h, i = this._bitrev;
     if (a === 4)
       for (e = 0, h = 0; e < s; e += a, h++) {
-        const v = i[h];
-        this._singleTransform2(e, v, r);
+        const u = i[h];
+        this._singleTransform2(e, u, r);
       }
     else
       for (e = 0, h = 0; e < s; e += a, h++) {
-        const v = i[h];
-        this._singleTransform4(e, v, r);
+        const u = i[h];
+        this._singleTransform4(e, u, r);
       }
     var l = this._inv ? -1 : 1, c = this.table;
     for (r >>= 2; r >= 2; r >>= 2) {
       a = s / r << 1;
-      var d = a >>> 2;
+      var _ = a >>> 2;
       for (e = 0; e < s; e += a)
-        for (var m = e + d, u = e, f = 0; u < m; u += 2, f += r) {
-          const v = u, p = v + d, w = p + d, g = w + d, b = t[v], T = t[v + 1], F = t[p], y = t[p + 1], A = t[w], C = t[w + 1], P = t[g], M = t[g + 1], S = b, R = T, E = c[f], D = l * c[f + 1], z = F * E - y * D, x = F * D + y * E, B = c[2 * f], N = l * c[2 * f + 1], H = A * B - C * N, K = A * N + C * B, k = c[3 * f], $ = l * c[3 * f + 1], G = P * k - M * $, J = P * $ + M * k, L = S + H, U = R + K, I = S - H, Y = R - K, Q = z + G, W = x + J, j = l * (z - G), V = l * (x - J), X = L + Q, st = U + W, rt = L - Q, et = U - W, ot = I + V, nt = Y - j, at = I - V, it = Y + j;
-          t[v] = X, t[v + 1] = st, t[p] = ot, t[p + 1] = nt, t[w] = rt, t[w + 1] = et, t[g] = at, t[g + 1] = it;
+        for (var d = e + _, v = e, f = 0; v < d; v += 2, f += r) {
+          const u = v, p = u + _, w = p + _, g = w + _, b = t[u], T = t[u + 1], F = t[p], y = t[p + 1], A = t[w], C = t[w + 1], P = t[g], M = t[g + 1], S = b, R = T, E = c[f], U = l * c[f + 1], D = F * E - y * U, z = F * U + y * E, B = c[2 * f], N = l * c[2 * f + 1], H = A * B - C * N, K = A * N + C * B, k = c[3 * f], $ = l * c[3 * f + 1], G = P * k - M * $, J = P * $ + M * k, L = S + H, x = R + K, I = S - H, Y = R - K, Q = D + G, W = z + J, j = l * (D - G), V = l * (z - J), X = L + Q, st = x + W, rt = L - Q, et = x - W, ot = I + V, nt = Y - j, at = I - V, it = Y + j;
+          t[u] = X, t[u + 1] = st, t[p] = ot, t[p + 1] = nt, t[w] = rt, t[w + 1] = et, t[g] = at, t[g + 1] = it;
         }
     }
-  }, _.prototype._singleTransform2 = function(t, s, n) {
-    const r = this._out, a = this._data, e = a[s], h = a[s + 1], i = a[s + n], l = a[s + n + 1], c = e + i, d = h + l, m = e - i, u = h - l;
-    r[t] = c, r[t + 1] = d, r[t + 2] = m, r[t + 3] = u;
-  }, _.prototype._singleTransform4 = function(t, s, n) {
-    const r = this._out, a = this._data, e = this._inv ? -1 : 1, h = n * 2, i = n * 3, l = a[s], c = a[s + 1], d = a[s + n], m = a[s + n + 1], u = a[s + h], f = a[s + h + 1], v = a[s + i], p = a[s + i + 1], w = l + u, g = c + f, b = l - u, T = c - f, F = d + v, y = m + p, A = e * (d - v), C = e * (m - p), P = w + F, M = g + y, S = b + C, R = T - A, E = w - F, D = g - y, z = b - C, x = T + A;
-    r[t] = P, r[t + 1] = M, r[t + 2] = S, r[t + 3] = R, r[t + 4] = E, r[t + 5] = D, r[t + 6] = z, r[t + 7] = x;
-  }, _.prototype._realTransform4 = function() {
+  }, m.prototype._singleTransform2 = function(t, s, n) {
+    const r = this._out, a = this._data, e = a[s], h = a[s + 1], i = a[s + n], l = a[s + n + 1], c = e + i, _ = h + l, d = e - i, v = h - l;
+    r[t] = c, r[t + 1] = _, r[t + 2] = d, r[t + 3] = v;
+  }, m.prototype._singleTransform4 = function(t, s, n) {
+    const r = this._out, a = this._data, e = this._inv ? -1 : 1, h = n * 2, i = n * 3, l = a[s], c = a[s + 1], _ = a[s + n], d = a[s + n + 1], v = a[s + h], f = a[s + h + 1], u = a[s + i], p = a[s + i + 1], w = l + v, g = c + f, b = l - v, T = c - f, F = _ + u, y = d + p, A = e * (_ - u), C = e * (d - p), P = w + F, M = g + y, S = b + C, R = T - A, E = w - F, U = g - y, D = b - C, z = T + A;
+    r[t] = P, r[t + 1] = M, r[t + 2] = S, r[t + 3] = R, r[t + 4] = E, r[t + 5] = U, r[t + 6] = D, r[t + 7] = z;
+  }, m.prototype._realTransform4 = function() {
     var t = this._out, s = this._csize, n = this._width, r = 1 << n, a = s / r << 1, e, h, i = this._bitrev;
     if (a === 4)
       for (e = 0, h = 0; e < s; e += a, h++) {
@@ -100,48 +100,48 @@ function Et() {
     var l = this._inv ? -1 : 1, c = this.table;
     for (r >>= 2; r >= 2; r >>= 2) {
       a = s / r << 1;
-      var d = a >>> 1, m = d >>> 1, u = m >>> 1;
+      var _ = a >>> 1, d = _ >>> 1, v = d >>> 1;
       for (e = 0; e < s; e += a)
-        for (var f = 0, v = 0; f <= u; f += 2, v += r) {
-          var p = e + f, w = p + m, g = w + m, b = g + m, T = t[p], F = t[p + 1], y = t[w], A = t[w + 1], C = t[g], P = t[g + 1], M = t[b], S = t[b + 1], R = T, E = F, D = c[v], z = l * c[v + 1], x = y * D - A * z, B = y * z + A * D, N = c[2 * v], H = l * c[2 * v + 1], K = C * N - P * H, k = C * H + P * N, $ = c[3 * v], G = l * c[3 * v + 1], J = M * $ - S * G, L = M * G + S * $, U = R + K, I = E + k, Y = R - K, Q = E - k, W = x + J, j = B + L, V = l * (x - J), X = l * (B - L), st = U + W, rt = I + j, et = Y + X, ot = Q - V;
+        for (var f = 0, u = 0; f <= v; f += 2, u += r) {
+          var p = e + f, w = p + d, g = w + d, b = g + d, T = t[p], F = t[p + 1], y = t[w], A = t[w + 1], C = t[g], P = t[g + 1], M = t[b], S = t[b + 1], R = T, E = F, U = c[u], D = l * c[u + 1], z = y * U - A * D, B = y * D + A * U, N = c[2 * u], H = l * c[2 * u + 1], K = C * N - P * H, k = C * H + P * N, $ = c[3 * u], G = l * c[3 * u + 1], J = M * $ - S * G, L = M * G + S * $, x = R + K, I = E + k, Y = R - K, Q = E - k, W = z + J, j = B + L, V = l * (z - J), X = l * (B - L), st = x + W, rt = I + j, et = Y + X, ot = Q - V;
           if (t[p] = st, t[p + 1] = rt, t[w] = et, t[w + 1] = ot, f === 0) {
-            var nt = U - W, at = I - j;
+            var nt = x - W, at = I - j;
             t[g] = nt, t[g + 1] = at;
             continue;
           }
-          if (f !== u) {
-            var it = Y, wt = -Q, gt = U, yt = -I, bt = -l * X, Ft = -l * V, Tt = -l * j, At = -l * W, Ct = it + bt, Pt = wt + Ft, Mt = gt + At, St = yt - Tt, ft = e + m - f, vt = e + d - f;
-            t[ft] = Ct, t[ft + 1] = Pt, t[vt] = Mt, t[vt + 1] = St;
+          if (f !== v) {
+            var it = Y, wt = -Q, gt = x, yt = -I, bt = -l * X, Ft = -l * V, Tt = -l * j, At = -l * W, Ct = it + bt, Pt = wt + Ft, Mt = gt + At, St = yt - Tt, ft = e + d - f, ut = e + _ - f;
+            t[ft] = Ct, t[ft + 1] = Pt, t[ut] = Mt, t[ut + 1] = St;
           }
         }
     }
-  }, _.prototype._singleRealTransform2 = function(t, s, n) {
+  }, m.prototype._singleRealTransform2 = function(t, s, n) {
     const r = this._out, a = this._data, e = a[s], h = a[s + n], i = e + h, l = e - h;
     r[t] = i, r[t + 1] = 0, r[t + 2] = l, r[t + 3] = 0;
-  }, _.prototype._singleRealTransform4 = function(t, s, n) {
-    const r = this._out, a = this._data, e = this._inv ? -1 : 1, h = n * 2, i = n * 3, l = a[s], c = a[s + n], d = a[s + h], m = a[s + i], u = l + d, f = l - d, v = c + m, p = e * (c - m), w = u + v, g = f, b = -p, T = u - v, F = f, y = p;
+  }, m.prototype._singleRealTransform4 = function(t, s, n) {
+    const r = this._out, a = this._data, e = this._inv ? -1 : 1, h = n * 2, i = n * 3, l = a[s], c = a[s + n], _ = a[s + h], d = a[s + i], v = l + _, f = l - _, u = c + d, p = e * (c - d), w = v + u, g = f, b = -p, T = v - u, F = f, y = p;
     r[t] = w, r[t + 1] = 0, r[t + 2] = g, r[t + 3] = b, r[t + 4] = T, r[t + 5] = 0, r[t + 6] = F, r[t + 7] = y;
   }, ht;
 }
-var Dt = Et();
-const zt = /* @__PURE__ */ Rt(Dt);
+var Ut = Et();
+const Dt = /* @__PURE__ */ Rt(Ut);
 class pt {
   constructor(o = 16e3, t = 512, s = 64) {
-    this.sampleRate = o, this.nfft = t, this.nfilt = s, this.fft = new zt(t), this.melFilters = this.createMelFilterbank();
+    this._sampleRate = o, this._nfft = t, this._nfilt = s, this._fft = new Dt(t), this._melFilters = this._createMelFilterbank();
   }
-  hzToMel(o) {
+  _hzToMel(o) {
     return 2595 * Math.log10(1 + o / 700);
   }
-  melToHz(o) {
+  _melToHz(o) {
     return 700 * (10 ** (o / 2595) - 1);
   }
-  createMelFilterbank() {
-    const t = this.sampleRate / 2, s = this.hzToMel(0), n = this.hzToMel(t), r = new Float32Array(this.nfilt + 2);
-    for (let i = 0; i < this.nfilt + 2; i++)
-      r[i] = s + i * (n - s) / (this.nfilt + 1);
-    const e = r.map((i) => this.melToHz(i)).map((i) => Math.floor((this.nfft + 1) * i / this.sampleRate)), h = [];
-    for (let i = 0; i < this.nfilt; i++) {
-      const l = new Float32Array(Math.floor(this.nfft / 2) + 1);
+  _createMelFilterbank() {
+    const t = this._sampleRate / 2, s = this._hzToMel(0), n = this._hzToMel(t), r = new Float32Array(this._nfilt + 2);
+    for (let i = 0; i < this._nfilt + 2; i++)
+      r[i] = s + i * (n - s) / (this._nfilt + 1);
+    const e = r.map((i) => this._melToHz(i)).map((i) => Math.floor((this._nfft + 1) * i / this._sampleRate)), h = [];
+    for (let i = 0; i < this._nfilt; i++) {
+      const l = new Float32Array(Math.floor(this._nfft / 2) + 1);
       for (let c = e[i]; c < e[i + 1]; c++)
         l[c] = (c - e[i]) / (e[i + 1] - e[i]);
       for (let c = e[i + 1]; c < e[i + 2]; c++)
@@ -152,62 +152,93 @@ class pt {
   }
   /** Returns a flat Float32Array of shape [numFrames × nfilt]. */
   logfbank(o) {
-    const t = Math.floor(0.025 * this.sampleRate), s = Math.floor(0.01 * this.sampleRate), n = 1 + Math.ceil((o.length - t) / s), r = new Float32Array(n * this.nfilt), a = new Float32Array(this.nfft), e = this.fft.createComplexArray();
+    const t = Math.floor(0.025 * this._sampleRate), s = Math.floor(0.01 * this._sampleRate), n = 1 + Math.ceil((o.length - t) / s), r = new Float32Array(n * this._nfilt), a = new Float32Array(this._nfft), e = this._fft.createComplexArray();
     for (let h = 0; h < n; h++) {
       const i = h * s;
       a.fill(0);
-      for (let d = 0; d < t && i + d < o.length; d++)
-        a[d] = o[i + d];
-      const l = this.fft.toComplexArray(a, null);
-      this.fft.transform(e, l);
-      const c = new Float32Array(Math.floor(this.nfft / 2) + 1);
-      for (let d = 0; d < c.length; d++) {
-        const m = e[2 * d], u = e[2 * d + 1];
-        c[d] = 1 / this.nfft * (m * m + u * u), c[d] === 0 && (c[d] = 1e-30);
+      for (let _ = 0; _ < t && i + _ < o.length; _++)
+        a[_] = o[i + _];
+      const l = this._fft.toComplexArray(a, null);
+      this._fft.transform(e, l);
+      const c = new Float32Array(Math.floor(this._nfft / 2) + 1);
+      for (let _ = 0; _ < c.length; _++) {
+        const d = e[2 * _], v = e[2 * _ + 1];
+        c[_] = 1 / this._nfft * (d * d + v * v), c[_] === 0 && (c[_] = 1e-30);
       }
-      for (let d = 0; d < this.nfilt; d++) {
-        let m = 0;
-        const u = this.melFilters[d];
+      for (let _ = 0; _ < this._nfilt; _++) {
+        let d = 0;
+        const v = this._melFilters[_];
         for (let f = 0; f < c.length; f++)
-          m += c[f] * u[f];
-        m === 0 && (m = 1e-30), r[h * this.nfilt + d] = Math.log(m);
+          d += c[f] * v[f];
+        d === 0 && (d = 1e-30), r[h * this._nfilt + _] = Math.log(d);
       }
     }
     return r;
   }
+  maxCosineSim(o, t) {
+    let s = 0;
+    for (const n of t) {
+      let r = 0;
+      for (let e = 0; e < n.length; e++) r += o[e] * n[e];
+      const a = (r + 1) / 2;
+      a > s && (s = a);
+    }
+    return s;
+  }
 }
-async function mt(_ = q, o = tt) {
+async function dt(m = q, o = tt) {
   const t = await import(
     /* @vite-ignore */
     o
   );
-  return t.env.wasm.wasmPaths = _, t.env.wasm.numThreads = 1, t;
+  return t.env.wasm.wasmPaths = m, t.env.wasm.numThreads = 1, t;
 }
 let lt = null;
-async function dt(_ = _t, o = q, t = tt) {
-  return lt || (lt = mt(o, t).then(
-    (s) => s.InferenceSession.create(_, {
+async function _t(m = mt, o = q, t = tt) {
+  return lt || (lt = dt(o, t).then(
+    (s) => s.InferenceSession.create(m, {
       executionProviders: ["wasm"],
       graphOptimizationLevel: "all"
     })
   )), lt;
 }
-const xt = new pt();
 class O {
+  static loadWords(o = Z) {
+    try {
+      const t = localStorage.getItem(o);
+      return t ? JSON.parse(t) : [];
+    } catch {
+      return [];
+    }
+  }
+  static saveWord(o, t = Z) {
+    const s = O.loadWords(t).filter((n) => n.word_name !== o.word_name);
+    localStorage.setItem(t, JSON.stringify([...s, o]));
+  }
+  static deleteWord(o, t = Z) {
+    try {
+      const s = O.loadWords(t).filter((n) => n.word_name !== o);
+      localStorage.setItem(t, JSON.stringify(s));
+    } catch {
+    }
+  }
+}
+class It {
   constructor(o, t) {
     this._started = !1, this._inferring = !1, this._audioCtx = null, this._stream = null, this._refEmbeddings = /* @__PURE__ */ new Map(), this._lastMatchAt = 0, this._lastInferenceAt = 0;
     const {
       refsStorageKey: s = Z,
-      thresholdStorageKey: n = Wt,
+      thresholdStorageKey: n = xt,
       wasmPaths: r = q,
-      modelPath: a = _t,
-      audioProcessorPath: e = It,
-      ortCdnUrl: h = tt
+      modelPath: a = mt,
+      audioProcessorPath: e = zt,
+      ortCdnUrl: h = tt,
+      audioUtils: i = new pt()
     } = t || {};
-    this._commands = o, this._refsStorageKey = s, this._thresholdStorageKey = n, this._audioProcessorPath = e, this._wasmPaths = r, this._modelPath = a, this._ortCdnUrl = h;
+    this._audioUtils = i, this._commands = o, this._refsStorageKey = s, this._thresholdStorageKey = n, this._audioProcessorPath = e, this._wasmPaths = r, this._modelPath = a, this._ortCdnUrl = h;
     try {
-      const i = localStorage.getItem(this._thresholdStorageKey);
-      this._threshold = i !== null ? Math.max(0, Math.min(1, Number(i))) : 0.65;
+      const l = localStorage.getItem(this._thresholdStorageKey);
+      this._threshold = l !== null ? Math.max(0, Math.min(1, Number(l))) : 0.65;
     } catch {
       this._threshold = 0.65;
     }
@@ -227,7 +258,7 @@ class O {
     return this._started;
   }
   async _init() {
-    await dt(this._modelPath, this._wasmPaths, this._ortCdnUrl);
+    await _t(this._modelPath, this._wasmPaths, this._ortCdnUrl);
     const o = /* @__PURE__ */ new Set();
     for (const t of this._commands)
       for (const s of t.triggers)
@@ -288,16 +319,16 @@ class O {
     if (!(t - this._lastInferenceAt < 300)) {
       this._lastInferenceAt = t, this._inferring = !0;
       try {
-        const [s, n] = await Promise.all([mt(this._wasmPaths, this._ortCdnUrl), dt(this._modelPath, this._wasmPaths, this._ortCdnUrl)]), r = xt.logfbank(o), a = new s.Tensor("float32", r, [1, 1, 149, 64]), e = await n.run({ input: a }), h = e[Object.keys(e)[0]].data;
+        const [s, n] = await Promise.all([dt(this._wasmPaths, this._ortCdnUrl), _t(this._modelPath, this._wasmPaths, this._ortCdnUrl)]), r = this._audioUtils.logfbank(o), a = new s.Tensor("float32", r, [1, 1, 149, 64]), e = await n.run({ input: a }), h = e[Object.keys(e)[0]].data;
         let i = !1;
         for (const l of this._commands) {
           if (i) break;
           for (const c of l.triggers) {
-            const d = this._refEmbeddings.get(c.name);
-            if (!d) continue;
-            const m = this._maxCosineSim(h, d);
-            if (m >= this._threshold && t - this._lastMatchAt > 2e3) {
-              this._lastMatchAt = t, console.info(`[Mellon] match: "${c}" sim=${m.toFixed(3)}`), typeof l.onMatch == "function" && l.onMatch(c.name, m), i = !0;
+            const _ = this._refEmbeddings.get(c.name);
+            if (!_) continue;
+            const d = this._audioUtils.maxCosineSim(h, _);
+            if (d >= this._threshold && t - this._lastMatchAt > 2e3) {
+              this._lastMatchAt = t, console.info(`[Mellon] match: "${c}" sim=${d.toFixed(3)}`), typeof l.onMatch == "function" && l.onMatch(c.name, d), i = !0;
               break;
             }
           }
@@ -309,40 +340,10 @@ class O {
       }
     }
   }
-  _maxCosineSim(o, t) {
-    let s = 0;
-    for (const n of t) {
-      let r = 0;
-      for (let e = 0; e < n.length; e++) r += o[e] * n[e];
-      const a = (r + 1) / 2;
-      a > s && (s = a);
-    }
-    return s;
-  }
-  static loadWords(o = Z) {
-    try {
-      const t = localStorage.getItem(o);
-      return t ? JSON.parse(t) : [];
-    } catch {
-      return [];
-    }
-  }
-  static saveWord(o, t = Z) {
-    const s = O.loadWords(t).filter((n) => n.word_name !== o.word_name);
-    localStorage.setItem(t, JSON.stringify([...s, o]));
-  }
-  static deleteWord(o, t = Z) {
-    try {
-      const s = O.loadWords(t).filter((n) => n.word_name !== o);
-      localStorage.setItem(t, JSON.stringify(s));
-    } catch {
-    }
-  }
 }
-const Ut = new pt();
-class jt {
+class Wt {
   constructor(o, t) {
-    this._config = {}, this._samples = [], this._wordName = o, this._config.modelPath = (t == null ? void 0 : t.modelPath) || _t, this._config.wasmPaths = (t == null ? void 0 : t.wasmPaths) || q, this._config.ortCdnUrl = (t == null ? void 0 : t.ortCdnUrl) || tt;
+    this._config = {}, this._samples = [], this._wordName = o, this._config.modelPath = (t == null ? void 0 : t.modelPath) || mt, this._config.wasmPaths = (t == null ? void 0 : t.wasmPaths) || q, this._config.ortCdnUrl = (t == null ? void 0 : t.ortCdnUrl) || tt, this._audioUtils = (t == null ? void 0 : t.audioUtils) ?? new pt();
   }
   /** Records 1.5 s of audio, stores the decoded PCM, returns new sample count. */
   async recordSample() {
@@ -354,8 +355,8 @@ class jt {
         var l;
         for (const c of o.getTracks()) c.stop();
         try {
-          const d = await new Blob(i, { type: ((l = i[0]) == null ? void 0 : l.type) || "audio/webm" }).arrayBuffer(), m = await t.decodeAudioData(d);
-          await t.close(), a(m.getChannelData(0).slice());
+          const _ = await new Blob(i, { type: ((l = i[0]) == null ? void 0 : l.type) || "audio/webm" }).arrayBuffer(), d = await t.decodeAudioData(_);
+          await t.close(), a(d.getChannelData(0).slice());
         } catch (c) {
           e(c);
         }
@@ -376,22 +377,24 @@ class jt {
   }
   /** Runs ONNX inference on every recorded sample to produce reference embeddings. */
   async generateRef() {
-    const [o, t] = await Promise.all([mt(this._config.wasmPaths, this._config.ortCdnUrl), dt(this._config.modelPath, this._config.wasmPaths, this._config.ortCdnUrl)]), s = [];
+    const [o, t] = await Promise.all([dt(this._config.wasmPaths, this._config.ortCdnUrl), _t(this._config.modelPath, this._config.wasmPaths, this._config.ortCdnUrl)]), s = [];
     for (const n of this._samples) {
-      const r = Ut.logfbank(n), a = new o.Tensor("float32", r, [1, 1, 149, 64]), e = await t.run({ input: a }), h = Array.from(e[Object.keys(e)[0]].data);
+      const r = this._audioUtils.logfbank(n), a = new o.Tensor("float32", r, [1, 1, 149, 64]), e = await t.run({ input: a }), h = Array.from(e[Object.keys(e)[0]].data);
       s.push(h);
     }
     return { word_name: this._wordName, model_type: "resnet_50_arc", embeddings: s };
   }
 }
-const q = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/", tt = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs", _t = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx", It = "https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js", Z = "mellon-refs", Wt = "mellon-threshold";
+const q = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/", tt = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs", mt = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx", zt = "https://cdn.jsdelivr.net/npm/mellon@0.0.14/dist/assets/audio-processor.js", Z = "mellon-refs", xt = "mellon-threshold";
 export {
-  It as DEFAULT_AUDIO_PROCESSOR_PATH,
-  _t as DEFAULT_MODEL_PATH,
+  pt as AudioUtils,
+  zt as DEFAULT_AUDIO_PROCESSOR_PATH,
+  mt as DEFAULT_MODEL_PATH,
   tt as DEFAULT_ORT_CDN_URL,
   Z as DEFAULT_REFS_STORAGE_KEY,
-  Wt as DEFAULT_THRESHOLD_STORAGE_KEY,
+  xt as DEFAULT_THRESHOLD_STORAGE_KEY,
   q as DEFAULT_WASM_PATHS,
-  jt as EnrollmentSession,
-  O as Mellon
+  It as Detector,
+  Wt as EnrollmentSession,
+  O as Storage
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mellon",
-  "version": "0.0.18",
+  "version": "0.0.19",
   "description": "Offline, in-browser voice commands powered by EfficientWord-Net (ResNet-50 ArcFace).",
   "type": "module",
   "main": "./dist/mellon.cjs",