speechflow 2.4.1 → 2.4.2

package/speechflow-cli/etc/oxlint.jsonc

@@ -150,6 +150,10 @@
         "import-style":                       "off",
         "no-implicit-globals":                "off",
         "no-anonymous-default-export":        "off",
+        "prefer-named-capture-group":         "off",
+        "prefer-number-coercion":             "off",
+        "method-signature-style":             "off",
+        "no-sync":                            "off",
 
         "complexity":                         [ "error", { "max": 40 } ]
     }

package/speechflow-cli/etc/stx.conf

@@ -6,15 +6,15 @@
 
 #   [speechflow-cli] make patches
 patch-make
-    npm shrinkwrap && \
+    npm install --package-lock-only --save && \
+    shx echo "allow-remote=all" >.npmrc && \
     patch-package --patch-dir package.d \
         "@ericedouard/vad-node-realtime" \
-        "@shiguredo" \
-        "sherpa-onnx" \
+        "@shiguredo/rnnoise-wasm" \
         "kitten-tts-js" \
         "speex-resampler" \
         "@typescript-eslint/typescript-estree" && \
-    shx rm -f npm-shrinkwrap.json
+    shx rm -f package-lock.json .npmrc
 
 #   [speechflow-cli] (internal) apply patches
 patch-apply

package/speechflow-cli/package.json

@@ -18,56 +18,56 @@
         "cli-io":                               "0.9.13",
         "yargs":                                "18.0.0",
         "flowlink":                             "1.2.3",
-        "js-yaml":                              "4.1.1",
+        "js-yaml":                              "5.2.1",
         "@gpeng/naudiodon":                     "2.4.1",
-        "execa":                                "9.6.1",
+        "execa":                                "10.0.0",
         "shell-parser":                         "1.0.0",
-        "@deepgram/sdk":                        "4.11.3",
+        "@deepgram/sdk":                        "5.6.0",
         "deepl-node":                           "1.27.0",
-        "@elevenlabs/elevenlabs-js":            "2.49.1",
+        "@elevenlabs/elevenlabs-js":            "2.58.0",
         "get-stream":                           "9.0.1",
-        "@dotenvx/dotenvx":                     "1.69.2",
+        "@dotenvx/dotenvx":                     "2.14.0",
         "speex-resampler":                      "3.0.1",
         "@sapphi-red/speex-preprocess-wasm":    "0.4.0",
         "@shiguredo/rnnoise-wasm":              "2025.1.5",
-        "sherpa-onnx":                          "1.12.25",
-        "axios":                                "1.16.1",
-        "@aws-sdk/client-transcribe-streaming": "3.1057.0",
-        "@aws-sdk/client-translate":            "3.1057.0",
-        "@aws-sdk/client-polly":                "3.1057.0",
-        "@google-cloud/translate":              "9.4.1",
-        "@google-cloud/speech":                 "7.3.1",
+        "sherpa-onnx":                          "1.13.4",
+        "axios":                                "1.18.1",
+        "@aws-sdk/client-transcribe-streaming": "3.1090.0",
+        "@aws-sdk/client-translate":            "3.1090.0",
+        "@aws-sdk/client-polly":                "3.1090.0",
+        "@google-cloud/translate":              "9.4.2",
+        "@google-cloud/speech":                 "7.5.0",
         "@google-cloud/text-to-speech":         "6.4.1",
         "node-web-audio-api":                   "2.0.0",
         "object-path":                          "0.11.8",
         "glob":                                 "13.0.6",
-        "ws":                                   "8.21.0",
+        "ws":                                   "8.21.1",
         "bufferutil":                           "4.1.0",
         "utf-8-validate":                       "6.0.6",
         "@hapi/hapi":                           "21.4.9",
         "@hapi/boom":                           "10.0.1",
-        "@hapi/inert":                          "7.1.1",
+        "@hapi/inert":                          "7.1.2",
         "hapi-plugin-header":                   "1.1.8",
         "hapi-plugin-websocket":                "2.4.11",
         "@opensumi/reconnecting-websocket":     "4.4.0",
         "ollama":                               "0.6.3",
-        "openai":                               "6.39.1",
-        "@anthropic-ai/sdk":                    "0.100.1",
-        "@google/genai":                        "1.52.0",
+        "openai":                               "6.48.0",
+        "@anthropic-ai/sdk":                    "0.112.3",
+        "@google/genai":                        "2.12.0",
         "@rse/ffmpeg":                          "1.4.2",
         "ffmpeg-stream":                        "1.0.1",
         "installed-packages":                   "1.0.13",
         "syspath":                              "1.0.8",
         "wav":                                  "1.0.2",
-        "mqtt":                                 "5.15.1",
+        "mqtt":                                 "5.15.2",
         "vban":                                 "1.5.6",
         "cbor2":                                "2.3.0",
-        "arktype":                              "2.2.0",
+        "arktype":                              "2.2.3",
         "pure-uuid":                            "2.0.0",
         "wavefile":                             "11.0.0",
-        "audio-inspect":                        "0.0.4",
-        "@huggingface/transformers":            "3.8.1",
-        "@huggingface/hub":                     "2.13.0",
+        "audio-inspect":                        "0.0.8",
+        "@huggingface/transformers":            "4.2.0",
+        "@huggingface/hub":                     "2.13.3",
         "kokoro-js":                            "1.2.1",
         "kitten-tts-js":                        "0.1.2",
         "@ericedouard/vad-node-realtime":       "0.2.0",
@@ -79,33 +79,33 @@
         "@2toad/profanity":                     "3.3.0",
         "cli-table3":                           "0.6.5",
         "mkdirp":                               "3.0.1",
-        "@soundtouchjs/audio-worklet":          "1.0.10",
+        "@soundtouchjs/audio-worklet":          "2.1.0",
         "werift":                               "0.23.0",
         "@discordjs/opus":                      "0.10.0",
-        "@rse/stx":                             "1.1.5"
+        "@rse/stx":                             "1.1.6"
     },
     "devDependencies": {
-        "eslint":                               "9.39.4",
-        "@eslint/js":                           "9.39.4",
+        "eslint":                               "9.39.5",
+        "@eslint/js":                           "9.39.5",
         "neostandard":                          "0.13.0",
         "eslint-plugin-promise":                "7.3.0",
         "eslint-plugin-import":                 "2.32.0",
         "eslint-plugin-node":                   "11.1.0",
-        "typescript-eslint":                    "8.60.0",
-        "@typescript-eslint/eslint-plugin":     "8.60.0",
-        "@typescript-eslint/parser":            "8.60.0",
-        "oxlint":                               "1.67.0",
-        "oxlint-plugin-complexity":             "2.1.3",
-        "eslint-plugin-oxlint":                 "1.67.0",
+        "typescript-eslint":                    "8.64.0",
+        "@typescript-eslint/eslint-plugin":     "8.64.0",
+        "@typescript-eslint/parser":            "8.64.0",
+        "oxlint":                               "1.73.0",
+        "oxlint-plugin-complexity":             "2.1.5",
+        "eslint-plugin-oxlint":                 "1.73.0",
 
-        "@types/node":                          "24.10.1",
+        "@types/node":                          "26.1.1",
         "@types/yargs":                         "17.0.35",
         "@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/luxon":                         "3.7.2",
         "@types/wrap-text":                     "1.0.2",
 
         "check-dependencies":                   "2.0.0",
@@ -113,14 +113,12 @@
         "stmux":                                "1.8.11",
         "nodemon":                              "3.1.14",
         "shx":                                  "0.4.0",
-        "@yao-pkg/pkg":                         "6.20.0",
+        "@yao-pkg/pkg":                         "6.21.0",
         "typescript":                           "6.0.3",
         "delay-cli":                            "3.0.0",
         "cross-env":                            "10.1.0"
     },
     "overrides": {
-        "@huggingface/transformers":            { "onnxruntime-node": "1.22.0", "onnxruntime-web": "1.22.0" },
-        "@ericedouard/vad-node-realtime":       { "onnxruntime-node": "1.22.0", "onnxruntime-web": "1.22.0" },
         "typescript-eslint":                    { "typescript": ">=5.0.0" },
         "@typescript-eslint/parser":            { "typescript": ">=5.0.0" },
         "@typescript-eslint/eslint-plugin":     { "typescript": ">=5.0.0" },
@@ -130,16 +128,10 @@
         "neostandard":                          { "eslint": ">=9.0.0" }
     },
     "upd": [
-        "!onnxruntime-node",
-        "!@types/node",
-        "!audio-inspect",
-        "!@deepgram/sdk",
-        "!sherpa-onnx",
+        "!oxlint",
         "!eslint",
         "!@eslint/js",
-        "!@google/genai",
-        "!@huggingface/transformers",
-        "!@soundtouchjs/audio-worklet"
+        "!typescript"
     ],
     "engines": {
         "node":                                 ">=22.0.0"
@@ -154,6 +146,21 @@
             "require":                          { "types": "./dst/speechflow-node.d.ts", "default": "./dst/speechflow-node.js" }
         }
     },
+    "allowScripts": {
+        "@discordjs/opus":                      true,
+        "@google/genai":                        true,
+        "@gpeng/naudiodon":                     true,
+        "@rse/ffmpeg":                          true,
+        "bufferutil":                           true,
+        "es5-ext":                              true,
+        "esbuild":                              true,
+        "fsevents":                             true,
+        "node-pty":                             true,
+        "onnxruntime-node":                     true,
+        "protobufjs":                           true,
+        "sharp":                                true,
+        "utf-8-validate":                       true
+    },
     "scripts": {
         "postinstall":                          "npm start patch-apply",
         "start":                                "stx -v4 -c etc/stx.conf"

package/speechflow-cli/src/speechflow-main-cli.ts

@@ -11,7 +11,7 @@ import path              from "node:path"
 import CLIio             from "cli-io"
 import yargs             from "yargs"
 import { hideBin }       from "yargs/helpers"
-import jsYAML            from "js-yaml"
+import * as jsYAML       from "js-yaml"
 import dotenvx           from "@dotenvx/dotenvx"
 import syspath           from "syspath"
 import chalk             from "chalk"

package/speechflow-cli/src/speechflow-node-a2a-ffmpeg.ts

@@ -93,10 +93,12 @@ export default class SpeechFlowNodeA2AFFMPEG extends SpeechFlowNode {
         util.run("starting FFmpeg process", () => this.ffmpeg!.run())
 
         /*  establish a duplex stream and connect it to FFmpeg  */
+        /*  NOTICE: the object form of Stream.Duplex.from() accepts Node streams at
+            run-time, but @types/node types it for Web streams only, so we have to cast  */
         const ffmpegStream = Stream.Duplex.from({
             writable: streamInput,
             readable: streamOutput
-        })
+        } as unknown as Parameters<typeof Stream.Duplex.from>[0])
 
         /*  wrap streams with conversions for chunk vs plain audio  */
         const wrapper1 = util.createTransformStreamForWritableSide("audio", 1)

package/speechflow-cli/src/speechflow-node-a2a-gender.ts

@@ -11,7 +11,7 @@ import Stream  from "node:stream"
 /*  external dependencies  */
 import * as Transformers     from "@huggingface/transformers"
 import { WaveFile }          from "wavefile"
-import { getRMS, AudioData } from "audio-inspect"
+import { analyze, AudioData } from "audio-inspect"
 import { Duration }          from "luxon"
 
 /*  internal dependencies  */
@@ -145,7 +145,8 @@ export default class SpeechFlowNodeA2AGender extends SpeechFlowNode {
                 duration:         data.length / sampleRateTarget,
                 length:           data.length
             } satisfies AudioData
-            const rms = getRMS(audioData, { asDB: true })
+            const analysis = await analyze(audioData, { features: { rms: { asDB: true } } })
+            const rms = analysis.results.rms ?? -60
             if (rms < this.params.volumeThreshold)
                 return genderLast
 

package/speechflow-cli/src/speechflow-node-a2a-meter.ts

@@ -8,7 +8,7 @@
 import Stream from "node:stream"
 
 /*  external dependencies  */
-import { getLUFS, getRMS, AudioData } from "audio-inspect"
+import { analyze, AudioData } from "audio-inspect"
 
 /*  internal dependencies  */
 import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"
@@ -69,8 +69,8 @@ export default class SpeechFlowNodeA2AMeter extends SpeechFlowNode {
         const samplesPerChunk = Math.floor(this.config.audioSampleRate * chunkDuration)
         this.chunkBuffer = new Float32Array(0)
 
-        /*  setup chunking interval  */
-        this.calcInterval = setInterval(() => {
+        /*  calculate the loudness metrics of the accumulated chunk data  */
+        const calcMeter = async () => {
             /*  short-circuit during destruction  */
             if (this.closing)
                 return
@@ -99,14 +99,19 @@ export default class SpeechFlowNodeA2AMeter extends SpeechFlowNode {
                 duration:         sampleWindowDuration,
                 length:           sampleWindow.length
             } satisfies AudioData
-            const lufs = getLUFS(audioDataLUFS, {
-                channelMode: this.config.audioChannels === 1 ? "mono" : "stereo",
-                calculateShortTerm:     false,
-                calculateMomentary:     true,
-                calculateLoudnessRange: false,
-                calculateTruePeak:      false
+            const analysisLUFS = await analyze(audioDataLUFS, {
+                features: {
+                    lufs: {
+                        channelMode: this.config.audioChannels === 1 ? "mono" : "stereo",
+                        calculateShortTerm:     false,
+                        calculateMomentary:     true,
+                        calculateLoudnessRange: false,
+                        calculateTruePeak:      false
+                    }
+                }
             })
-            lufsm = lufs.momentary ? Math.max(-60, lufs.momentary[0]) : -60
+            const lufs = analysisLUFS.results.lufs
+            lufsm = lufs?.momentary !== undefined ? Math.max(-60, lufs.momentary) : -60
 
             /*  calculate the RMS metric  */
             const totalSamples   = chunkData.length / this.config.audioChannels
@@ -118,9 +123,14 @@ export default class SpeechFlowNodeA2AMeter extends SpeechFlowNode {
                 duration,
                 length:           chunkData.length
             } satisfies AudioData
-            rms = Math.max(-60, getRMS(audioDataRMS, {
-                asDB: true
-            }))
+            const analysisRMS = await analyze(audioDataRMS, {
+                features: {
+                    rms: {
+                        asDB: true
+                    }
+                }
+            })
+            rms = Math.max(-60, analysisRMS.results.rms ?? -60)
 
             /*  automatically clear measurement (in case no new measurements happen)  */
             if (this.silenceTimer !== null)
@@ -129,6 +139,13 @@ export default class SpeechFlowNodeA2AMeter extends SpeechFlowNode {
                 lufsm = -60
                 rms   = -60
             }, 500)
+        }
+
+        /*  setup chunking interval  */
+        this.calcInterval = setInterval(() => {
+            calcMeter().catch((err: unknown) => {
+                this.log("warning", `meter calculation failed: ${err instanceof Error ? err.message : String(err)}`)
+            })
         }, chunkDuration * 1000)
 
         /*  setup loudness emitting interval  */

package/speechflow-cli/src/speechflow-node-a2a-pitch.ts

@@ -18,7 +18,6 @@ import * as util                           from "./speechflow-util"
 /*  parameter configuration  */
 type AudioPitchShifterConfig = {
     rate?:      number
-    tempo?:     number
     pitch?:     number
     semitones?: number
 }
@@ -38,7 +37,6 @@ class AudioPitchShifter extends util.WebAudio {
         super(sampleRate, channels)
         this.config = {
             rate:      config.rate      ?? 1.0,
-            tempo:     config.tempo     ?? 1.0,
             pitch:     config.pitch     ?? 1.0,
             semitones: config.semitones ?? 0.0
         }
@@ -50,10 +48,11 @@ class AudioPitchShifter extends util.WebAudio {
 
         /*  add SoundTouch worklet module  */
         const packagePath = path.join(__dirname, "../node_modules/@soundtouchjs/audio-worklet")
-        const workletPath = path.join(packagePath, "dist/soundtouch-worklet.js")
+        const workletPath = path.join(packagePath, ".dist/soundtouch-processor.js")
         await this.audioContext.audioWorklet.addModule(workletPath)
 
-        /*  create SoundTouch worklet node  */
+        /*  create SoundTouch worklet node (the processor always operates on
+            stereo internally, but mixes down to the requested channel count)  */
         this.pitchNode = new AudioWorkletNode(this.audioContext, "soundtouch-processor", {
             numberOfInputs:  1,
             numberOfOutputs: 1,
@@ -62,10 +61,9 @@ class AudioPitchShifter extends util.WebAudio {
 
         /*  set initial parameter values  */
         const params = this.pitchNode.parameters as Map<string, AudioParam>
-        params.get("rate")!.value           = this.config.rate
-        params.get("tempo")!.value          = this.config.tempo
-        params.get("pitch")!.value          = this.config.pitch
-        params.get("pitchSemitones")!.value = this.config.semitones
+        params.get("playbackRate")!.value    = this.config.rate
+        params.get("pitch")!.value           = this.config.pitch
+        params.get("pitchSemitones")!.value  = this.config.semitones
 
         /*  connect nodes: source -> pitch -> capture  */
         this.sourceNode!.connect(this.pitchNode)
@@ -83,14 +81,10 @@ class AudioPitchShifter extends util.WebAudio {
         this.config[configField] = value
     }
 
-    /*  update rate value  */
+    /*  update rate value (compensates the pitch for an upstream playback
+        rate change, i.e. the resulting pitch is divided by this value)  */
     public setRate (rate: number): void {
-        this.updateParameter("rate", rate, "rate")
-    }
-
-    /*  update tempo value  */
-    public setTempo (tempo: number): void {
-        this.updateParameter("tempo", tempo, "tempo")
+        this.updateParameter("playbackRate", rate, "rate")
     }
 
     /*  update pitch shift value  */
@@ -131,10 +125,9 @@ export default class SpeechFlowNodeA2APitch extends SpeechFlowNode {
 
         /*  declare node configuration parameters  */
         this.configure({
-            rate:      { type: "number",  val: 1.0,  match: (n: number) => n >= 0.25 && n <= 4.0 },
-            tempo:     { type: "number",  val: 1.0,  match: (n: number) => n >= 0.25 && n <= 4.0 },
-            pitch:     { type: "number",  val: 1.0,  match: (n: number) => n >= 0.25 && n <= 4.0 },
-            semitones: { type: "number",  val: 0.0,  match: (n: number) => n >= -24  && n <= 24  }
+            rate:      { type: "number",  val: 1.0,  match: (n: number) => n >= 0.1 && n <= 8.0 },
+            pitch:     { type: "number",  val: 1.0,  match: (n: number) => n >= 0.1 && n <= 8.0 },
+            semitones: { type: "number",  val: 0.0,  match: (n: number) => n >= -24 && n <= 24  }
         })
 
         /*  declare node input/output format  */
@@ -152,7 +145,6 @@ export default class SpeechFlowNodeA2APitch extends SpeechFlowNode {
             this.config.audioSampleRate,
             this.config.audioChannels, {
                 rate:      this.params.rate,
-                tempo:     this.params.tempo,
                 pitch:     this.params.pitch,
                 semitones: this.params.semitones
             }