speechflow 1.4.5 → 1.5.1

package/speechflow-cli/package.json

@@ -1,116 +1,125 @@
 {
-    "private":                              "true",
-    "name":                                 "speechflow",
-    "version":                              "0.0.0",
-    "homepage":                             "https://github.com/rse/speechflow",
-    "description":                          "Speech Processing Flow Graph",
-    "license":                              "GPL-3.0-only",
+    "private":                                  true,
+    "name":                                     "speechflow",
+    "version":                                  "0.0.0",
+    "homepage":                                 "https://github.com/rse/speechflow",
+    "description":                              "Speech Processing Flow Graph",
+    "license":                                  "GPL-3.0-only",
     "author": {
-        "name":                             "Dr. Ralf S. Engelschall",
-        "email":                            "rse@engelschall.com",
-        "url":                              "http://engelschall.com"
+        "name":                                 "Dr. Ralf S. Engelschall",
+        "email":                                "rse@engelschall.com",
+        "url":                                  "http://engelschall.com"
     },
     "repository": {
-        "type":                             "git",
-        "url":                              "git+https://github.com/rse/speechflow.git"
+        "type":                                 "git",
+        "url":                                  "git+https://github.com/rse/speechflow.git"
     },
     "dependencies": {
-        "cli-io":                           "0.9.13",
-        "yargs":                            "18.0.0",
-        "flowlink":                         "1.0.0",
-        "js-yaml":                          "4.1.0",
-        "@gpeng/naudiodon":                 "2.4.1",
-        "@deepgram/sdk":                    "4.11.2",
-        "deepl-node":                       "1.19.0",
-        "@elevenlabs/elevenlabs-js":        "2.8.0",
-        "stream-transform":                 "3.4.0",
-        "get-stream":                       "9.0.1",
-        "@dotenvx/dotenvx":                 "1.48.4",
-        "speex-resampler":                  "3.0.1",
-        "object-path":                      "0.11.8",
-        "ws":                               "8.18.3",
-        "bufferutil":                       "4.0.9",
-        "utf-8-validate":                   "6.0.5",
-        "@hapi/hapi":                       "21.4.2",
-        "@hapi/boom":                       "10.0.1",
-        "@hapi/inert":                      "7.1.0",
-        "hapi-plugin-header":               "1.1.8",
-        "hapi-plugin-websocket":            "2.4.11",
-        "@opensumi/reconnecting-websocket": "4.4.0",
-        "ollama":                           "0.5.16",
-        "openai":                           "5.12.0",
-        "@rse/ffmpeg":                      "1.4.2",
-        "ffmpeg-stream":                    "1.0.1",
-        "installed-packages":               "1.0.13",
-        "syspath":                          "1.0.8",
-        "wav":                              "1.0.2",
-        "mqtt":                             "5.14.0",
-        "cbor2":                            "2.0.1",
-        "arktype":                          "2.1.20",
-        "pure-uuid":                        "1.8.1",
-        "wavefile":                         "11.0.0",
-        "audio-inspect":                    "0.0.4",
-        "@huggingface/transformers":        "3.7.1",
-        "kokoro-js":                        "1.2.1",
-        "@ericedouard/vad-node-realtime":   "0.2.0",
-        "luxon":                            "3.7.1",
-        "node-interval-tree":               "2.1.2",
-        "wrap-text":                        "1.0.10",
-        "cli-table3":                       "0.6.5",
-        "@rse/stx":                         "1.0.7"
+        "cli-io":                               "0.9.13",
+        "yargs":                                "18.0.0",
+        "flowlink":                             "1.2.1",
+        "js-yaml":                              "4.1.0",
+        "@gpeng/naudiodon":                     "2.4.1",
+        "@deepgram/sdk":                        "4.11.2",
+        "deepl-node":                           "1.19.0",
+        "@elevenlabs/elevenlabs-js":            "2.13.0",
+        "stream-transform":                     "3.4.0",
+        "get-stream":                           "9.0.1",
+        "@dotenvx/dotenvx":                     "1.49.0",
+        "speex-resampler":                      "3.0.1",
+        "@sapphi-red/speex-preprocess-wasm":    "0.4.0",
+        "@shiguredo/rnnoise-wasm":              "2025.1.5",
+        "@aws-sdk/client-transcribe-streaming": "3.879.0",
+        "@aws-sdk/client-translate":            "3.879.0",
+        "@aws-sdk/client-polly":                "3.879.0",
+        "@google-cloud/translate":              "9.2.0",
+        "node-web-audio-api":                   "1.0.4",
+        "object-path":                          "0.11.8",
+        "ws":                                   "8.18.3",
+        "bufferutil":                           "4.0.9",
+        "utf-8-validate":                       "6.0.5",
+        "@hapi/hapi":                           "21.4.3",
+        "@hapi/boom":                           "10.0.1",
+        "@hapi/inert":                          "7.1.0",
+        "hapi-plugin-header":                   "1.1.8",
+        "hapi-plugin-websocket":                "2.4.11",
+        "@opensumi/reconnecting-websocket":     "4.4.0",
+        "ollama":                               "0.5.17",
+        "openai":                               "5.18.0",
+        "@rse/ffmpeg":                          "1.4.2",
+        "ffmpeg-stream":                        "1.0.1",
+        "installed-packages":                   "1.0.13",
+        "syspath":                              "1.0.8",
+        "wav":                                  "1.0.2",
+        "mqtt":                                 "5.14.0",
+        "cbor2":                                "2.0.1",
+        "arktype":                              "2.1.20",
+        "pure-uuid":                            "1.8.1",
+        "wavefile":                             "11.0.0",
+        "audio-inspect":                        "0.0.4",
+        "@huggingface/transformers":            "3.7.2",
+        "kokoro-js":                            "1.2.1",
+        "@ericedouard/vad-node-realtime":       "0.2.0",
+        "osc-js":                               "2.4.1",
+        "luxon":                                "3.7.1",
+        "node-interval-tree":                   "2.1.2",
+        "wrap-text":                            "1.0.10",
+        "cli-table3":                           "0.6.5",
+        "@rse/stx":                             "1.1.0"
     },
     "devDependencies": {
-        "eslint":                           "9.32.0",
-        "@eslint/js":                       "9.32.0",
-        "neostandard":                      "0.12.2",
-        "eslint-plugin-promise":            "7.2.1",
-        "eslint-plugin-import":             "2.32.0",
-        "eslint-plugin-node":               "11.1.0",
-        "@typescript-eslint/eslint-plugin": "8.39.0",
-        "@typescript-eslint/parser":        "8.39.0",
-        "oxlint":                           "1.10.0",
-        "eslint-plugin-oxlint":             "1.10.0",
-        "@biomejs/biome":                   "2.0.6",
-        "eslint-config-biome":              "2.1.3",
+        "eslint":                               "9.34.0",
+        "@eslint/js":                           "9.34.0",
+        "neostandard":                          "0.12.2",
+        "eslint-plugin-promise":                "7.2.1",
+        "eslint-plugin-import":                 "2.32.0",
+        "eslint-plugin-node":                   "11.1.0",
+        "typescript-eslint":                    "8.42.0",
+        "@typescript-eslint/eslint-plugin":     "8.42.0",
+        "@typescript-eslint/parser":            "8.42.0",
+        "oxlint":                               "1.14.0",
+        "eslint-plugin-oxlint":                 "1.14.0",
+        "@biomejs/biome":                       "2.0.6",
+        "eslint-config-biome":                  "2.1.3",
 
-        "@types/node":                      "24.2.0",
-        "@types/yargs":                     "17.0.33",
-        "@types/js-yaml":                   "4.0.9",
-        "@types/object-path":               "0.11.4",
-        "@types/ws":                        "8.18.1",
-        "@types/resolve":                   "1.20.6",
-        "@types/wav":                       "1.0.4",
-        "@types/luxon":                     "3.7.1",
-        "@types/wrap-text":                 "1.0.2",
+        "@types/node":                          "24.3.0",
+        "@types/yargs":                         "17.0.33",
+        "@types/js-yaml":                       "4.0.9",
+        "@types/object-path":                   "0.11.4",
+        "@types/ws":                            "8.18.1",
+        "@types/resolve":                       "1.20.6",
+        "@types/wav":                           "1.0.4",
+        "@types/luxon":                         "3.7.1",
+        "@types/wrap-text":                     "1.0.2",
 
-        "patch-package":                    "8.0.0",
-        "stmux":                            "1.8.11",
-        "nodemon":                          "3.1.10",
-        "shx":                              "0.4.0",
-        "@yao-pkg/pkg":                     "6.6.0",
-        "typescript":                       "5.9.2",
-        "delay-cli":                        "2.0.0",
-        "cross-env":                        "10.0.0"
+        "patch-package":                        "8.0.0",
+        "stmux":                                "1.8.11",
+        "nodemon":                              "3.1.10",
+        "shx":                                  "0.4.0",
+        "@yao-pkg/pkg":                         "6.6.0",
+        "typescript":                           "5.9.2",
+        "delay-cli":                            "2.0.0",
+        "cross-env":                            "10.0.0"
     },
     "overrides": {
-        "@huggingface/transformers":        { "onnxruntime-node": "1.23.0-dev.20250703-7fc6235861" }
+        "@huggingface/transformers":            { "onnxruntime-node": "1.23.0-dev.20250703-7fc6235861" }
     },
-    "upd":                                  [ "!@biomejs/biome" ],
+    "upd":                                      [ "!@biomejs/biome" ],
     "engines": {
-        "node":                             ">=22.0.0"
+        "node":                                 ">=22.0.0"
     },
-    "bin":                                  { "speechflow": "dst/speechflow.js" },
-    "types":                                "./dst/speechflow-node.d.ts",
-    "module":                               "./dst/speechflow-node.js",
-    "main":                                 "./dst/speechflow-node.js",
+    "bin":                                      { "speechflow": "dst/speechflow.js" },
+    "types":                                    "./dst/speechflow-node.d.ts",
+    "module":                                   "./dst/speechflow-node.js",
+    "main":                                     "./dst/speechflow-node.js",
     "exports": {
         ".": {
-            "import":                       { "types": "./dst/speechflow-node.d.ts", "default": "./dst/speechflow-node.js" },
-            "require":                      { "types": "./dst/speechflow-node.d.ts", "default": "./dst/speechflow-node.js" }
+            "import":                           { "types": "./dst/speechflow-node.d.ts", "default": "./dst/speechflow-node.js" },
+            "require":                          { "types": "./dst/speechflow-node.d.ts", "default": "./dst/speechflow-node.js" }
         }
     },
     "scripts": {
-        "postinstall":                      "npm start patch-apply",
-        "start":                            "stx -v4 -c etc/stx.conf"
+        "postinstall":                          "npm start patch-apply",
+        "start":                                "stx -v4 -c etc/stx.conf"
     }
 }

package/speechflow-cli/src/lib.d.ts

@@ -9,3 +9,27 @@ declare module "node:stream" {
     export function compose (...streams: Stream[]): Duplex
 }
 
+/*  type definitions for AudioWorkletProcessor  */
+declare interface AudioWorkletProcessor {
+    readonly port: MessagePort
+    process(
+        inputs:  Float32Array[][],
+        outputs: Float32Array[][],
+        parameters: Record<string, Float32Array>
+    ): boolean
+}
+declare const AudioWorkletProcessor: {
+    prototype: AudioWorkletProcessor
+    new(): AudioWorkletProcessor
+}
+declare interface AudioParamDescriptor {
+    name:            string
+    defaultValue?:   number
+    minValue?:       number
+    maxValue?:       number
+    automationRate?: "a-rate" | "k-rate"
+}
+declare function registerProcessor(
+    name: string,
+    processorCtor: new (...args: any[]) => AudioWorkletProcessor
+): void

package/speechflow-cli/src/speechflow-node-a2a-compressor-wt.ts

@@ -0,0 +1,151 @@
+/*
+**  SpeechFlow - Speech Processing Flow Graph
+**  Copyright (c) 2024-2025 Dr. Ralf S. Engelschall <rse@engelschall.com>
+**  Licensed under GPL 3.0 <https://spdx.org/licenses/GPL-3.0-only>
+*/
+
+import * as utils from "./speechflow-utils"
+
+/*  downward compressor with soft knee  */
+class CompressorProcessor extends AudioWorkletProcessor {
+    /*  internal state  */
+    private env: number[] = []
+    private sampleRate: number
+    public reduction = 0
+
+    /*  eslint no-undef: off */
+    static get parameterDescriptors(): AudioParamDescriptor[] {
+        return [
+            { name: "threshold",  defaultValue: -23,   minValue: -100,   maxValue: 0,   automationRate: "k-rate" }, // dBFS
+            { name: "ratio",      defaultValue: 4.0,   minValue: 1.0,    maxValue: 20,  automationRate: "k-rate" }, // compression ratio
+            { name: "attack",     defaultValue: 0.010, minValue: 0.0,    maxValue: 1,   automationRate: "k-rate" }, // seconds
+            { name: "release",    defaultValue: 0.050, minValue: 0.0,    maxValue: 1,   automationRate: "k-rate" }, // seconds
+            { name: "knee",       defaultValue: 6.0,   minValue: 0.0,    maxValue: 40,  automationRate: "k-rate" }, // dB
+            { name: "makeup",     defaultValue: 0.0,   minValue: -24,    maxValue: 24,  automationRate: "k-rate" }  // dB
+        ]
+    }
+
+    /*  class constructor for custom option processing  */
+    constructor (options: any) {
+        super()
+        const { sampleRate } = options.processorOptions
+        this.sampleRate = sampleRate as number
+    }
+
+    /*  determine gain difference  */
+    private gainDBFor (levelDB: number, thresholdDB: number, ratio: number, kneeDB: number): number {
+        /*  short-circuit for unreasonable ratio  */
+        if (ratio <= 1.0)
+            return 0
+
+        /*  determine thresholds  */
+        const halfKnee  = kneeDB * 0.5
+        const belowThr  = levelDB < thresholdDB
+        const aboveKnee = levelDB >= (thresholdDB + halfKnee)
+
+        /*  short-circuit for no compression (below threshold)  */
+        if (belowThr)
+            return 0
+
+        /*  apply soft-knee  */
+        if (kneeDB > 0 && !aboveKnee) {
+            const x = (levelDB - thresholdDB) / kneeDB
+            const idealGainDB = (thresholdDB + (levelDB - thresholdDB) / ratio) - levelDB
+            return idealGainDB * x * x
+        }
+
+        /*  determine target level  */
+        const targetOut = thresholdDB + (levelDB - thresholdDB) / ratio
+
+        /*  return gain difference  */
+        return targetOut - levelDB
+    }
+
+    /*  update envelope (smoothed amplitude contour) for single channel  */
+    private updateEnvelopeForChannel (
+        chan:           number,
+        samples:        Float32Array,
+        attack:         number,
+        release:        number
+    ): void {
+        /*  fetch old envelope value  */
+        if (this.env[chan] === undefined)
+            this.env[chan] = 1e-12
+        let env = this.env[chan]
+
+        /*  calculate attack/release alpha values  */
+        const alphaA = Math.exp(-1 / (attack  * this.sampleRate))
+        const alphaR = Math.exp(-1 / (release * this.sampleRate))
+
+        /*  iterate over all samples and calculate RMS  */
+        for (const s of samples) {
+            const x = Math.abs(s)
+            const det = x * x
+            if (det > env)
+                env = alphaA * env + (1 - alphaA) * det
+            else
+                env = alphaR * env + (1 - alphaR) * det
+        }
+        this.env[chan] = Math.sqrt(Math.max(env, 1e-12))
+    }
+
+    /*  process a single sample frame  */
+    process(
+        inputs:     Float32Array[][],
+        outputs:    Float32Array[][],
+        parameters: Record<string, Float32Array>
+    ): boolean {
+        /*  sanity check  */
+        const input  = inputs[0]
+        const output = outputs[0]
+        if (!input || input.length === 0 || !output)
+            return true
+
+        /*  determine number of channels  */
+        const nCh = input.length
+
+        /*  initially just copy input to output (pass-through)  */
+        for (let c = 0; c < output.length; c++) {
+            if (!output[c] || !input[c])
+                continue
+            output[c].set(input[c])
+        }
+
+        /*  fetch parameters  */
+        const thresholdDB = parameters["threshold"][0]
+        const ratio       = parameters["ratio"][0]
+        const kneeDB      = parameters["knee"][0]
+        const attackS     = Math.max(parameters["attack"][0],  1 / this.sampleRate)
+        const releaseS    = Math.max(parameters["release"][0], 1 / this.sampleRate)
+        const makeupDB    = parameters["makeup"][0]
+
+        /*  update envelope per channel  */
+        for (let ch = 0; ch < nCh; ch++)
+            this.updateEnvelopeForChannel(ch, input[ch], attackS, releaseS)
+
+        /*  determine linear value from decibel makeup value */
+        const makeUpLin = utils.dB2lin(makeupDB)
+
+        /*  iterate over all channels  */
+        this.reduction = 0
+        for (let ch = 0; ch < nCh; ch++) {
+            const levelDB = utils.lin2dB(this.env[ch])
+            const gainDB  = this.gainDBFor(levelDB, thresholdDB, ratio, kneeDB)
+            const gainLin = utils.dB2lin(gainDB) * makeUpLin
+
+            /*  on first channel, calculate reduction  */
+            if (ch === 0)
+                this.reduction = Math.min(0, gainDB)
+
+            /*  apply gain change to channel  */
+            const inp = input[ch]
+            const out = output[ch]
+            for (let i = 0; i < inp.length; i++)
+                out[i] = inp[i] * gainLin
+        }
+        return true
+    }
+}
+
+/*  register the new audio nodes  */
+registerProcessor("compressor", CompressorProcessor)