mellon 0.0.25 → 0.0.26

package/dist/index.d.ts

@@ -1,54 +1,165 @@
-import { Detector } from './Detector';
-import { EnrollmentSession } from './EnrollmentSession';
-import { Storage } from './Storage';
-import { AudioUtils } from './AudioUtils';
-export type TriggerName = string;
-export interface Trigger {
-    name: TriggerName;
-    defaultRefPath?: string;
-}
-export interface Command {
-    name: string;
-    triggers: Trigger[];
-    onMatch?: (trigger: TriggerName, confidence: number) => any;
-}
-export interface DetectorConfig {
-    refsStorageKey?: string;
-    thresholdStorageKey?: string;
-    wasmPaths?: string;
-    modelPath?: string;
-    audioProcessorPath?: string;
-    ortCdnUrl?: string;
-    audioUtils?: AudioUtils;
-    /** Enable console logging. Pass `true` for info+warn+error, or a custom logger. Defaults to silent. */
-    log?: boolean | {
-        info?: (...a: unknown[]) => void;
-        warn?: (...a: unknown[]) => void;
-        error?: (...a: unknown[]) => void;
-    };
-}
-export interface EnrollmentSessionConfig {
-    wasmPaths?: string;
-    modelPath?: string;
-    ortCdnUrl?: string;
-    audioUtils?: AudioUtils;
-}
-export interface WordRef {
-    word_name: TriggerName;
-    model_type?: string;
-    embeddings: number[][];
-}
-/**
- * Called during {@link Detector.init} to report real download progress.
- * @param downloaded - total bytes received so far across all assets
- * @param total      - sum of known Content-Length values for all assets;
- *                     may still be 0 early on (before first header is received)
- */
-export type ProgressCallback = (downloaded: number, total: number) => void;
-export declare const DEFAULT_WASM_PATHS = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/";
-export declare const DEFAULT_ORT_CDN_URL = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs";
-export declare const DEFAULT_MODEL_PATH = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx";
-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 { Detector, EnrollmentSession, Storage, AudioUtils };
+/**
+ * Log Mel-filterbank feature extraction.
+ * Exact port of the AudioUtils used by EfficientWord-Net InBrowser.
+ *   - 64 Mel filters
+ *   - 512-point FFT
+ *   - 25 ms window / 10 ms step at 16 kHz
+ * Produces a [149 × 64] spectrogram from a 1.5 s (24 000-sample) input,
+ * matching the [1, 1, 149, 64] tensor expected by the ONNX Siamese model.
+ */
+export declare class AudioUtils {
+    private _nfft;
+    private _nfilt;
+    private _sampleRate;
+    private _melFilters;
+    private _fft;
+    constructor(sampleRate?: number, nfft?: number, nfilt?: number);
+    private _hzToMel;
+    private _melToHz;
+    private _createMelFilterbank;
+    /** Returns a flat Float32Array of shape [numFrames × nfilt]. */
+    logfbank(signal: Float32Array): Float32Array;
+    maxCosineSim(embedding: Float32Array, refs: number[][]): number;
+}
+
+export declare interface Command {
+    name: string;
+    triggers: Trigger[];
+    onMatch?: (trigger: TriggerName, confidence: number) => any;
+}
+
+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_MODEL_PATH = "https://huggingface.co/ComicScrip/mellon/resolve/main/model.onnx";
+
+export declare const DEFAULT_ORT_CDN_URL = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/ort.wasm.min.mjs";
+
+export declare const DEFAULT_REFS_STORAGE_KEY = "mellon-refs";
+
+export declare const DEFAULT_THRESHOLD_STORAGE_KEY = "mellon-threshold";
+
+export declare const DEFAULT_WASM_PATHS = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.24.3/dist/";
+
+export declare class Detector {
+    private _refsStorageKey;
+    private _thresholdStorageKey;
+    private _audioProcessorPath;
+    private _modelPath;
+    private _wasmPaths;
+    private _ortCdnUrl;
+    private _started;
+    private _inferring;
+    private _audioCtx;
+    private _stream;
+    private _refEmbeddings;
+    private _lastMatchAt;
+    private _lastInferenceAt;
+    private _audioUtils;
+    private _commands;
+    private _threshold;
+    private _initPromise;
+    private _log;
+    get threshold(): number;
+    set threshold(value: number);
+    get listening(): boolean;
+    constructor(commands: Command[], config?: DetectorConfig);
+    /**
+     * Streams `url`, calling `onProgress(downloaded, total)` after each chunk.
+     * Falls back to a single-shot fetch when the body stream is unavailable.
+     */
+    private _trackFetch;
+    private _init;
+    /**
+     * Loads the ONNX model and all reference embeddings.
+     * Must be called before {@link start}.
+     * Safe to call multiple times — the work is only done once.
+     *
+     * @param onProgress - optional callback invoked as each asset is loaded
+     */
+    init(onProgress?: ProgressCallback): Promise<void>;
+    /** Adds (or replaces) the reference embeddings for a word without restarting. */
+    addCustomWord(ref: WordRef): void;
+    start(): Promise<void>;
+    stop(): Promise<void>;
+    private _handleBuffer;
+}
+
+export declare interface DetectorConfig {
+    refsStorageKey?: string;
+    thresholdStorageKey?: string;
+    wasmPaths?: string;
+    modelPath?: string;
+    audioProcessorPath?: string;
+    ortCdnUrl?: string;
+    audioUtils?: AudioUtils;
+    /** Enable console logging. Pass `true` for info+warn+error, or a custom logger. Defaults to silent. */
+    log?: boolean | {
+        info?: (...a: unknown[]) => void;
+        warn?: (...a: unknown[]) => void;
+        error?: (...a: unknown[]) => void;
+    };
+}
+
+/**
+ * Records voice samples and generates an EfficientWord-Net reference file
+ * (compatible with the JSON format used by the ONNX Siamese model).
+ *
+ * Usage:
+ *   const session = new EnrollmentSession('suivant');
+ *   const count = await session.recordSample(); // repeat ≥ 3 times
+ *   const ref = await session.generateRef();    // needs ≥ 3 samples
+ *   Mellon.saveWord(ref);
+ *   mellon.addCustomWord(ref);
+ */
+export declare class EnrollmentSession {
+    private _config;
+    private _wordName;
+    private _samples;
+    private _audioUtils;
+    constructor(wordName: string, config?: EnrollmentSessionConfig);
+    /** Records 1.5 s of audio, stores the decoded PCM, returns new sample count. */
+    recordSample(): Promise<number>;
+    /** Returns the raw PCM Float32Array for the sample at the given index (16 kHz). */
+    getSample(index: number): Float32Array;
+    /** Removes the sample at the given index. Returns the new sample count. */
+    deleteSample(index: number): number;
+    /** Runs ONNX inference on every recorded sample to produce reference embeddings. */
+    generateRef(): Promise<WordRef>;
+}
+
+export declare interface EnrollmentSessionConfig {
+    wasmPaths?: string;
+    modelPath?: string;
+    ortCdnUrl?: string;
+    audioUtils?: AudioUtils;
+}
+
+/**
+ * Called during {@link Detector.init} to report real download progress.
+ * @param downloaded - total bytes received so far across all assets
+ * @param total      - sum of known Content-Length values for all assets;
+ *                     may still be 0 early on (before first header is received)
+ */
+export declare type ProgressCallback = (downloaded: number, total: number) => void;
+
+declare class Storage_2 {
+    static loadWords(storageKey?: string): WordRef[];
+    static saveWord(ref: WordRef, storageKey?: string): void;
+    static deleteWord(wordName: string, storageKey?: string): void;
+}
+export { Storage_2 as Storage }
+
+export declare interface Trigger {
+    name: TriggerName;
+    defaultRefPath?: string;
+}
+
+export declare type TriggerName = string;
+
+export declare interface WordRef {
+    word_name: TriggerName;
+    model_type?: string;
+    embeddings: number[][];
+}
+
+export { }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mellon",
-  "version": "0.0.25",
+  "version": "0.0.26",
   "description": "Offline, in-browser voice commands powered by EfficientWord-Net (ResNet-50 ArcFace).",
   "type": "module",
   "main": "./dist/mellon.cjs",
@@ -31,7 +31,7 @@
     "voice"
   ],
   "scripts": {
-    "build": "vite build && tsc -p tsconfig.json"
+    "build": "vite build"
   },
   "dependencies": {
     "fft.js": "^4.0.4"