speechflow 2.0.4 → 2.1.1

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 ChangeLog
 =========
 
+2.1.1 (2026-02-18)
+------------------
+
+- IMPROVEMENT: add "endpointing" parameter to "a2t-deepgram" node
+- IMPROVEMENT: allow queue handling to be silenced
+- IMPROVEMENT: improve optical appearance of dashboard
+- CLEANUP: remove "preview" feature on "t2t-sentence"
+- BUGFIX: fix "t2t-sentence" processing for intermediate/final kinds and merging functionality
+- UPDATE: upgrade NPM dependencies
+
+2.1.0 (2026-01-27)
+------------------
+
+- BUGFIX: correctly support English ("en") as the source language in t2t-deepl node
+- BUGFIX: improve rendering of LUFS-M audio meter in Dashboard
+- BUGFIX: fix a2a-meter node by correcting the internal chunk buffer handling
+- IMPROVEMENT: add "keywords" parameter to node "a2t-deepgram"
+- IMPROVEMENT: improve readability of subtitle rendering
+- IMPROVEMENT: improve profanity filtering node
+- IMPROVEMENT: add a2a-gtcrn node for GTCRN-based local noise suppression
+- IMPROVEMENT: support intermediate/final text chunk tagging in t2t-sentence node
+- IMPROVEMENT: support timeout handling for incomplete text chunks in t2t-sentence node
+- IMPROVEMENT: support keywords also for Nova-3 mode in a2t-deepgram node
+- UPDATE: upgrade NPM dependencies
+
 2.0.4 (2026-01-16)
 ------------------
 

package/README.md

@@ -31,7 +31,7 @@ speech-to-speech).
 - local Voice Activity Detection (VAD),
 - local voice gender recognition,
 - local audio LUFS-S/RMS metering,
-- local audio Speex and RNNoise noise suppression,
+- local audio Speex, RNNoise, and GTCRN noise suppression,
 - local audio compressor and expander dynamics processing,
 - local audio gain adjustment,
 - local audio pitch shifting and time stretching,
@@ -356,6 +356,7 @@ First a short overview of the available processing nodes:
   **a2a-gender**,
   **a2a-speex**,
   **a2a-rnnoise**,
+  **a2a-gtcrn**,
   **a2a-compressor**,
   **a2a-expander**,
   **a2a-gain**,
@@ -717,6 +718,27 @@ The following nodes process audio chunks only.
   | Parameter | Position  | Default  | Requirement              |
   | --------- | --------- | -------- | ------------------------ |
 
+- Node: **a2a-gtcrn**<br/>
+  Purpose: **GTCRN Deep Learning Noise Suppression node**<br/>
+  Example: `a2a-gtcrn()`
+
+  > This node uses GTCRN (Gated Temporal Convolutional Recurrent Network)
+  > to perform deep learning based noise suppression and speech denoising.
+  > It detects and removes noise from the audio stream while preserving
+  > speech quality. The GTCRN ONNX model is automatically downloaded from
+  > the sherpa-onnx project on first use. NOTICE: This node internally
+  > operates at 16KHz sample rate only. Audio is automatically resampled
+  > from SpeechFlow's internal 48KHz to 16KHz for processing, and then
+  > resampled back to 48KHz for output.
+
+  | Port    | Payload     |
+  | ------- | ----------- |
+  | input   | audio       |
+  | output  | audio       |
+
+  | Parameter | Position  | Default  | Requirement              |
+  | --------- | --------- | -------- | ------------------------ |
+
 - Node: **a2a-compressor**<br/>
   Purpose: **audio compressor node**<br/>
   Example: `a2a-compressor(thresholdDb: -18)`
@@ -867,26 +889,30 @@ The following nodes convert audio to text chunks.
 
 - Node: **a2t-deepgram**<br/>
   Purpose: **Deepgram Speech-to-Text conversion**<br/>
-  Example: `a2t-deepgram(language: "de")`<br/>
+  Example: `a2t-deepgram(language: "de", keywords: "SpeechFlow, TypeScript")`<br/>
   Notice: this node requires an API key!
 
   > This node performs Speech-to-Text (S2T) conversion, i.e., it
   > recognizes speech in the input audio stream and outputs a
-  > corresponding text stream.
+  > corresponding text stream. The optional `keywords` parameter
+  > accepts a comma or space-separated list of words to boost
+  > during recognition, improving accuracy for domain-specific terminology.
 
   | Port    | Payload     |
   | ------- | ----------- |
   | input   | audio       |
   | output  | text        |
 
-  | Parameter    | Position  | Default  | Requirement        |
-  | ------------ | --------- | -------- | ------------------ |
-  | **key**      | *none*    | env.SPEECHFLOW\_DEEPGRAM\_KEY | *none* |
-  | **keyAdm**   | *none*    | env.SPEECHFLOW\_DEEPGRAM\_KEY\_ADM | *none* |
-  | **model**    | 0         | "nova-2" | *none* |
-  | **version**  | 1         | "latest" | *none* |
-  | **language** | 2         | "multi"  | *none* |
-  | **interim**  | 3         | false    | *none* |
+  | Parameter        | Position  | Default  | Requirement        |
+  | ---------------- | --------- | -------- | ------------------ |
+  | **key**          | *none*    | env.SPEECHFLOW\_DEEPGRAM\_KEY | *none* |
+  | **keyAdm**       | *none*    | env.SPEECHFLOW\_DEEPGRAM\_KEY\_ADM | *none* |
+  | **model**        | 0         | "nova-2" | *none* |
+  | **version**      | 1         | "latest" | *none* |
+  | **language**     | 2         | "multi"  | *none* |
+  | **interim**      | 3         | false    | *none* |
+  | **endpointing**  | 4         | 0        | *none* |
+  | **keywords**     | 5         | ""       | *none* |
 
 - Node: **a2t-google**<br/>
   Purpose: **Google Cloud Speech-to-Text conversion**<br/>
@@ -1131,12 +1157,15 @@ The following nodes process text chunks only.
 
 - Node: **t2t-sentence**<br/>
   Purpose: **sentence splitting/merging**<br/>
-  Example: `t2t-sentence()`<br/>
+  Example: `t2t-sentence(timeout: 3000)`<br/>
 
   > This node allows you to ensure that a text stream is split or merged
   > into complete sentences. It is primarily intended to be used after
   > the "a2t-deepgram" node and before "t2t-deepl" or "t2a-elevenlabs" nodes in
-  > order to improve overall quality.
+  > order to improve overall quality. Intermediate text chunks are passed
+  > through immediately, while final chunks are queued for sentence splitting.
+  > If an incomplete sentence remains in the queue longer than the timeout,
+  > it is promoted to a final chunk and emitted.
 
   | Port    | Payload     |
   | ------- | ----------- |
@@ -1145,6 +1174,8 @@ The following nodes process text chunks only.
 
   | Parameter    | Position  | Default  | Requirement        |
   | ------------ | --------- | -------- | ------------------ |
+  | **timeout**  | 0         | 3000     | *none*             |
+  | **interim**  | 1         | false    | *none*             |
 
 - Node: **t2t-subtitle**<br/>
   Purpose: **SRT/VTT Subtitle Generation**<br/>
@@ -1411,8 +1442,9 @@ History
 **SpeechFlow**, as a technical cut-through, was initially created in
 March 2024 for use in the msg Filmstudio context. It was later refined
 into a more complete toolkit in April 2025 and this way the first time
-could be used in production. It was fully refactored in July 2025 in
-order to support timestamps in the streams processing.
+could be used. It was fully refactored in July 2025 in order to support
+timestamps in the streams processing. In February 2026 it was the first
+time used in production in the msg Filmstudio.
 
 Copyright & License
 -------------------

package/etc/speechflow.yaml

@@ -88,58 +88,31 @@ studio-transcription: |
 #   Real-time studio translation from German to English,
 #   including the capturing of all involved inputs and outputs:
 studio-translation: |
-    xio-device(device: env.SPEECHFLOW_DEVICE_MIC, mode: "r") | {
-        a2a-gender() | {
-            a2a-meter(interval: 250, dashboard: "meter1") |
-                a2a-wav(mode: "encode") |
-                    xio-file(path: "program-de.wav", mode: "w", type: "audio"),
-            a2t-deepgram(language: "de", interim: true) | {
-                x2x-trace(name: "trace1", type: "text", dashboard: "text1") |
-                    t2t-subtitle(format: "vtt", words: true) |
-                        xio-file(path: "program-de.vtt", mode: "w", type: "text"),
+    xio-device(device: env.SPEECHFLOW_DEVICE_MIC, mode: "r", chunk: 200) | {
+        a2a-meter(interval: 50, dashboard: "audio-de", mode: "sink"),
+        a2t-deepgram(language: "de", model: "nova-3", interim: true, keywords: env.SPEECHFLOW_KEYWORDS) | {
+            t2t-profanity(lang: "de") | {
+                x2x-trace(type: "text", dashboard: "text-de-interim", mode: "sink"),
+                t2t-subtitle(mode: "render", addr: env.SPEECHFLOW_IPADDR, port: 8585),
                 t2t-sentence() | {
-                    x2x-trace(name: "trace2", type: "text", notify: true, dashboard: "text2") |
+                    x2x-filter(name: "final", type: "text", var: "kind", op: "==", val: "final") | {
                         t2t-format(width: 80) |
-                            xio-file(path: "program-de.txt", mode: "w", type: "text"),
+                            xio-file(path: `${env.SPEECHFLOW_DATADIR}/program-de.txt`, mode: "w", type: "text"),
+                        t2t-subtitle(mode: "export", format: "srt") |
+                            xio-file(path: `${env.SPEECHFLOW_DATADIR}/program-de.srt`, mode: "w", type: "text"),
+                        x2x-trace(type: "text", dashboard: "text-de-final", mode: "sink") /* DE-Final */
+                    },
                     t2t-deepl(src: "de", dst: "en") | {
-                        x2x-trace(name: "trace3", type: "text", dashboard: "text3") | {
+                        x2x-trace(type: "text", dashboard: "text-en-interim", mode: "sink"),
+                        t2t-subtitle(mode: "render", addr: env.SPEECHFLOW_IPADDR, port: 8686),
+                        x2x-filter(name: "final", type: "text", var: "kind", op: "==", val: "final") | {
                             t2t-format(width: 80) |
-                                xio-file(path: "program-en.txt", mode: "w", type: "text"),
-                            t2t-subtitle(format: "vtt", words: false) |
-                                xio-file(path: "program-en.vtt", mode: "w", type: "text"),
-                            {
-                                x2x-filter(name: "S2T-male", type: "text", var: "meta:gender", op: "==", val: "male") |
-                                    t2a-elevenlabs(voice: "Mark", optimize: "latency", speed: 1.05, language: "en"),
-                                x2x-filter(name: "S2T-female", type: "text", var: "meta:gender", op: "==", val: "female") |
-                                    t2a-elevenlabs(voice: "Brittney", optimize: "latency", speed: 1.05, language: "en")
-                            } | {
-                                a2a-meter(interval: 250, dashboard: "meter2"),
-                                a2a-wav(mode: "encode") |
-                                    xio-file(path: "program-en.wav", mode: "w", type: "audio"),
-                                xio-device(device: env.SPEECHFLOW_DEVICE_SPK, mode: "w")
-                            }
-                        }
-                    }
-                }
-            }
-        }
-    }
-
-#   Test-drive for development
-test: |
-    xio-device(device: env.SPEECHFLOW_DEVICE_MIC, mode: "r", chunk: 200) | {
-        a2a-meter(interval: 100, dashboard: "meter1", mode: "sink"),
-        a2a-gender() | {
-            a2t-deepgram(language: "de", model: "nova-2", interim: true) | {
-                x2x-trace(type: "text", mode: "sink", dashboard: "text1"),
-                t2t-subtitle(mode: "render", addr: "127.0.0.1", port: 8585),
-                x2x-filter(name: "final", type: "text", var: "kind", op: "==", val: "final") | {
-                    t2t-sentence() | {
-                        x2x-trace(type: "text", dashboard: "text2", mode: "sink"),
-                        t2t-deepl(src: "de", dst: "en") | {
-                            x2x-trace(type: "text", dashboard: "text3", mode: "sink"),
+                                xio-file(path: `${env.SPEECHFLOW_DATADIR}/program-en.txt`, mode: "w", type: "text"),
+                            t2t-subtitle(mode: "export", format: "srt") |
+                                xio-file(path: `${env.SPEECHFLOW_DATADIR}/program-en.srt`, mode: "w", type: "text"),
+                            x2x-trace(type: "text", dashboard: "text-en-final", mode: "sink"),
                             t2a-elevenlabs(voice: "Mark", optimize: "latency", speed: 1.05, language: "en") | {
-                                a2a-meter(interval: 100, dashboard: "meter2", mode: "sink"),
+                                a2a-meter(interval: 50, dashboard: "audio-en", mode: "sink"),
                                 xio-device(device: env.SPEECHFLOW_DEVICE_SPK, mode: "w")
                             }
                         }
@@ -148,4 +121,3 @@ test: |
             }
         }
     }
-

package/etc/stx.conf

@@ -68,8 +68,8 @@ run:dev
 #   [top-level] test drive
 test
     node --enable-source-maps speechflow-cli/dst/speechflow.js \
-        -v info -c test@etc/speechflow.yaml \
-        -d audio:meter1:DE,text:text1:DE-Interim,text:text2:DE-Final,text:text3:EN,audio:meter2:EN
+        -v info -c studio-translation@etc/speechflow.yaml \
+        -d audio:audio-de:DE,text:text-de-interim:DE-Interim,text:text-en-interim:EN-Interim,text:text-de-final:DE-Final,text:text-en-final:EN-Final,audio:audio-en:EN
 
 #   [top-level] remove all generated artifacts (reverse of "npm start build")
 clean

package/package.json

@@ -1,8 +1,8 @@
 {
     "name":             "speechflow",
-    "version":          "2.0.4",
-    "x-stdver":         "2.0.3-GA",
-    "x-release":        "2025-12-24",
+    "version":          "2.1.1",
+    "x-stdver":         "2.1.1-GA",
+    "x-release":        "2026-02-18",
     "homepage":         "https://github.com/rse/speechflow",
     "description":      "Speech Processing Flow Graph",
     "keywords":         [ "speech", "audio", "flow", "graph" ],
@@ -23,11 +23,11 @@
         "nodemon":      "3.1.11",
         "watch":        "1.0.2",
         "concurrently": "9.2.1",
-        "wait-on":      "9.0.3",
+        "wait-on":      "9.0.4",
         "cross-env":    "10.1.0",
         "shx":          "0.4.0",
-        "secretlint":   "11.3.0",
-        "@secretlint/secretlint-rule-preset-recommend": "11.3.0"
+        "secretlint":   "11.3.1",
+        "@secretlint/secretlint-rule-preset-recommend": "11.3.1"
     },
     "engines": {
         "npm":          ">=10.0.0",

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn-wt.d.ts

@@ -0,0 +1 @@
+export {};

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn-wt.js

@@ -0,0 +1,60 @@
+"use strict";
+/*
+**  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>
+*/
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+/*  standard dependencies  */
+const node_worker_threads_1 = require("node:worker_threads");
+/*  external dependencies  */
+const sherpa_onnx_1 = __importDefault(require("sherpa-onnx"));
+/*  receive model path from parent thread  */
+const modelPath = node_worker_threads_1.workerData.modelPath;
+/*  GTCRN state  */
+let denoiser;
+/*  helper: log message to parent  */
+const log = (level, message) => {
+    node_worker_threads_1.parentPort.postMessage({ type: "log", level, message });
+};
+(async () => {
+    try {
+        /*  create denoiser  */
+        const config = {
+            model: {
+                gtcrn: {
+                    model: modelPath
+                }
+            },
+            numThreads: 1
+        };
+        denoiser = sherpa_onnx_1.default.createOfflineSpeechDenoiser(config);
+        log("info", "GTCRN denoiser initialized");
+        node_worker_threads_1.parentPort.postMessage({ type: "ready" });
+    }
+    catch (err) {
+        node_worker_threads_1.parentPort.postMessage({ type: "failed", message: `failed to initialize GTCRN: ${err}` });
+        process.exit(1);
+    }
+})();
+/*  receive messages  */
+node_worker_threads_1.parentPort.on("message", (msg) => {
+    if (msg.type === "process") {
+        const { id, samples } = msg;
+        /*  process with GTCRN denoiser
+            NOTICE: GTCRN can also resample out input, but will always
+            produces 16KHz output, so we already fixate 16KHz input here!  */
+        const result = denoiser.run(samples, 16000);
+        /*  copy to transferable ArrayBuffer and send back to parent  */
+        const samplesDenoised = new Float32Array(result.samples);
+        node_worker_threads_1.parentPort.postMessage({ type: "process-done", id, data: samplesDenoised }, [samplesDenoised.buffer]);
+    }
+    else if (msg.type === "close") {
+        /*  shutdown this process  */
+        process.exit(0);
+    }
+});
+//# sourceMappingURL=speechflow-node-a2a-gtcrn-wt.js.map

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn-wt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speechflow-node-a2a-gtcrn-wt.js","sourceRoot":"","sources":["../src/speechflow-node-a2a-gtcrn-wt.ts"],"names":[],"mappings":";AAAA;;;;EAIE;;;;;AAEF,6BAA6B;AAC7B,6DAAgE;AAEhE,6BAA6B;AAC7B,8DAAwD;AAMxD,6CAA6C;AAC7C,MAAM,SAAS,GAAW,gCAAU,CAAC,SAAS,CAAA;AAE9C,mBAAmB;AACnB,IAAI,QAAyC,CAAA;AAE7C,qCAAqC;AACrC,MAAM,GAAG,GAAG,CAAC,KAAa,EAAE,OAAe,EAAE,EAAE;IAC3C,gCAAW,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,CAAA;AAC5D,CAAC,CAGA;AAAA,CAAC,KAAK,IAAI,EAAE;IACT,IAAI,CAAC;QACD,uBAAuB;QACvB,MAAM,MAAM,GAA6B;YACrC,KAAK,EAAE;gBACH,KAAK,EAAE;oBACH,KAAK,EAAE,SAAS;iBACnB;aACJ;YACD,UAAU,EAAE,CAAC;SAChB,CAAA;QACD,QAAQ,GAAG,qBAAU,CAAC,2BAA2B,CAAC,MAAM,CAAC,CAAA;QACzD,GAAG,CAAC,MAAM,EAAE,4BAA4B,CAAC,CAAA;QACzC,gCAAW,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,CAAC,CAAA;IAC9C,CAAC;IACD,OAAO,GAAG,EAAE,CAAC;QACT,gCAAW,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,+BAA+B,GAAG,EAAE,EAAE,CAAC,CAAA;QAC1F,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;IACnB,CAAC;AACL,CAAC,CAAC,EAAE,CAAA;AAEJ,wBAAwB;AACxB,gCAAW,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAG,EAAE,EAAE;IAC9B,IAAI,GAAG,CAAC,IAAI,KAAK,SAAS,EAAE,CAAC;QACzB,MAAM,EAAE,EAAE,EAAE,OAAO,EAAE,GAAG,GAAG,CAAA;QAE3B;;6EAEqE;QACrE,MAAM,MAAM,GAAG,QAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,KAAK,CAAC,CAAA;QAE3C,gEAAgE;QAChE,MAAM,eAAe,GAAG,IAAI,YAAY,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;QACxD,gCAAW,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,cAAc,EAAE,EAAE,EAAE,IAAI,EAAE,eAAe,EAAE,EAAE,CAAE,eAAe,CAAC,MAAM,CAAE,CAAC,CAAA;IAC5G,CAAC;SACI,IAAI,GAAG,CAAC,IAAI,KAAK,OAAO,EAAE,CAAC;QAC5B,6BAA6B;QAC7B,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;IACnB,CAAC;AACL,CAAC,CAAC,CAAA"}

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn.d.ts

@@ -0,0 +1,15 @@
+import SpeechFlowNode from "./speechflow-node";
+export default class SpeechFlowNodeA2AGTCRN extends SpeechFlowNode {
+    static name: string;
+    private closing;
+    private worker;
+    private resamplerDown;
+    private resamplerUp;
+    constructor(id: string, cfg: {
+        [id: string]: any;
+    }, opts: {
+        [id: string]: any;
+    }, args: any[]);
+    open(): Promise<void>;
+    close(): Promise<void>;
+}

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn.js

@@ -0,0 +1,234 @@
+"use strict";
+/*
+**  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>
+*/
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+/*  standard dependencies  */
+const node_fs_1 = __importDefault(require("node:fs"));
+const node_path_1 = __importDefault(require("node:path"));
+const node_stream_1 = __importDefault(require("node:stream"));
+const node_worker_threads_1 = require("node:worker_threads");
+/*  external dependencies  */
+const axios_1 = __importDefault(require("axios"));
+const speex_resampler_1 = __importDefault(require("speex-resampler"));
+/*  internal dependencies  */
+const speechflow_node_1 = __importDefault(require("./speechflow-node"));
+const util = __importStar(require("./speechflow-util"));
+/*  SpeechFlow node for GTCRN based noise suppression in audio-to-audio passing  */
+class SpeechFlowNodeA2AGTCRN extends speechflow_node_1.default {
+    /*  declare official node name  */
+    static name = "a2a-gtcrn";
+    /*  internal state  */
+    closing = false;
+    worker = null;
+    resamplerDown = null;
+    resamplerUp = null;
+    /*  construct node  */
+    constructor(id, cfg, opts, args) {
+        super(id, cfg, opts, args);
+        /*  declare node configuration parameters  */
+        this.configure({});
+        /*  declare node input/output format  */
+        this.input = "audio";
+        this.output = "audio";
+    }
+    /*  open node  */
+    async open() {
+        /*  clear destruction flag  */
+        this.closing = false;
+        /*  ensure GTCRN ONNX model is available  */
+        const modelUrl = "https://github.com/k2-fsa/sherpa-onnx/" +
+            "releases/download/speech-enhancement-models/gtcrn_simple.onnx";
+        const modelDir = node_path_1.default.join(this.config.cacheDir, "gtcrn");
+        const modelPath = node_path_1.default.resolve(modelDir, "gtcrn_simple.onnx");
+        const stat = await node_fs_1.default.promises.stat(modelPath).catch(() => null);
+        if (stat === null) {
+            this.log("info", `GTCRN model downloading from "${modelUrl}"`);
+            await node_fs_1.default.promises.mkdir(modelDir, { recursive: true });
+            const response = await axios_1.default.get(modelUrl, {
+                responseType: "arraybuffer",
+                onDownloadProgress: (progressEvent) => {
+                    if (progressEvent.total) {
+                        const percent = (progressEvent.loaded / progressEvent.total) * 100;
+                        this.log("info", `GTCRN model download: ${percent.toFixed(1)}%`);
+                    }
+                }
+            });
+            await node_fs_1.default.promises.writeFile(modelPath, Buffer.from(response.data));
+            this.log("info", `GTCRN model downloaded to "${modelPath}"`);
+        }
+        /*  establish resamplers from SpeechFlow's internal 48KHz
+            to GTCRN's required 16KHz format and back  */
+        this.resamplerDown = new speex_resampler_1.default(1, this.config.audioSampleRate, 16000, 7);
+        this.resamplerUp = new speex_resampler_1.default(1, 16000, this.config.audioSampleRate, 7);
+        /*  initialize worker  */
+        this.worker = new node_worker_threads_1.Worker(node_path_1.default.resolve(__dirname, "speechflow-node-a2a-gtcrn-wt.js"), {
+            workerData: { modelPath }
+        });
+        this.worker.on("error", (err) => {
+            this.log("error", `GTCRN worker thread error: ${err}`);
+            this.stream?.emit("error", err);
+        });
+        this.worker.on("exit", (code) => {
+            if (code !== 0)
+                this.log("error", `GTCRN worker thread exited with error code ${code}`);
+            else
+                this.log("info", `GTCRN worker thread exited with regular code ${code}`);
+        });
+        /*  wait for worker to be ready  */
+        await new Promise((resolve, reject) => {
+            const timeout = setTimeout(() => {
+                reject(new Error("GTCRN worker thread initialization timeout"));
+            }, 60 * 1000);
+            const onMessage = (msg) => {
+                if (typeof msg === "object" && msg !== null && msg.type === "log")
+                    this.log(msg.level, msg.message);
+                else if (typeof msg === "object" && msg !== null && msg.type === "ready") {
+                    clearTimeout(timeout);
+                    this.worker.off("message", onMessage);
+                    resolve();
+                }
+                else if (typeof msg === "object" && msg !== null && msg.type === "failed") {
+                    clearTimeout(timeout);
+                    this.worker.off("message", onMessage);
+                    reject(new Error(msg.message ?? "GTCRN worker thread initialization failed"));
+                }
+            };
+            this.worker.on("message", onMessage);
+            this.worker.once("error", (err) => {
+                clearTimeout(timeout);
+                reject(err);
+            });
+        });
+        /*  receive message from worker  */
+        const pending = new Map();
+        this.worker.on("exit", () => {
+            pending.clear();
+        });
+        this.worker.on("message", (msg) => {
+            if (typeof msg === "object" && msg !== null && msg.type === "process-done") {
+                const cb = pending.get(msg.id);
+                pending.delete(msg.id);
+                if (cb)
+                    cb(msg.data);
+                else
+                    this.log("warning", `GTCRN worker thread sent back unexpected id: ${msg.id}`);
+            }
+            else if (typeof msg === "object" && msg !== null && msg.type === "log")
+                this.log(msg.level, msg.message);
+            else
+                this.log("warning", `GTCRN worker thread sent unexpected message: ${JSON.stringify(msg)}`);
+        });
+        /*  send message to worker  */
+        let seq = 0;
+        const workerProcess = async (samples) => {
+            if (this.closing)
+                return samples;
+            const id = `${seq++}`;
+            return new Promise((resolve) => {
+                pending.set(id, (result) => { resolve(result); });
+                this.worker.postMessage({ type: "process", id, samples }, [samples.buffer]);
+            });
+        };
+        /*  establish a transform stream  */
+        const self = this;
+        this.stream = new node_stream_1.default.Transform({
+            readableObjectMode: true,
+            writableObjectMode: true,
+            decodeStrings: false,
+            transform(chunk, encoding, callback) {
+                if (self.closing) {
+                    callback(new Error("stream already destroyed"));
+                    return;
+                }
+                if (!Buffer.isBuffer(chunk.payload))
+                    callback(new Error("invalid chunk payload type"));
+                else {
+                    /*  resample Buffer from 48KHz (SpeechFlow) to 16KHz (GTCRN)  */
+                    const resampledDown = self.resamplerDown.processChunk(chunk.payload);
+                    /*  convert Buffer into Float32Array  */
+                    const payload = util.convertBufToF32(resampledDown);
+                    /*  process with GTCRN  */
+                    workerProcess(payload).then((result) => {
+                        /*  convert Float32Array into Buffer  */
+                        const buf = util.convertF32ToBuf(result);
+                        /*  resample Buffer from 16KHz (GTCRN) back to 48KHz (SpeechFlow)  */
+                        const resampledUp = self.resamplerUp.processChunk(buf);
+                        /*  update chunk  */
+                        chunk.payload = resampledUp;
+                        /*  forward updated chunk  */
+                        this.push(chunk);
+                        callback();
+                    }).catch((err) => {
+                        const error = util.ensureError(err);
+                        self.log("warning", `processing of chunk failed: ${error.message}`);
+                        callback(error);
+                    });
+                }
+            },
+            final(callback) {
+                callback();
+            }
+        });
+    }
+    /*  close node  */
+    async close() {
+        /*  indicate closing  */
+        this.closing = true;
+        /*  shutdown worker  */
+        if (this.worker !== null) {
+            this.worker.terminate();
+            this.worker = null;
+        }
+        /*  shutdown stream  */
+        if (this.stream !== null) {
+            await util.destroyStream(this.stream);
+            this.stream = null;
+        }
+        /*  destroy resamplers  */
+        if (this.resamplerDown !== null)
+            this.resamplerDown = null;
+        if (this.resamplerUp !== null)
+            this.resamplerUp = null;
+    }
+}
+exports.default = SpeechFlowNodeA2AGTCRN;
+//# sourceMappingURL=speechflow-node-a2a-gtcrn.js.map

package/speechflow-cli/dst/speechflow-node-a2a-gtcrn.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speechflow-node-a2a-gtcrn.js","sourceRoot":"","sources":["../src/speechflow-node-a2a-gtcrn.ts"],"names":[],"mappings":";AAAA;;;;EAIE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAEF,6BAA6B;AAC7B,sDAAyD;AACzD,0DAA2D;AAC3D,8DAA6D;AAC7D,6DAAqE;AAErE,6BAA6B;AAC7B,kDAAuD;AACvD,sEAAiE;AAEjE,6BAA6B;AAC7B,wEAAmE;AACnE,wDAAmE;AAEnE,mFAAmF;AACnF,MAAqB,sBAAuB,SAAQ,yBAAc;IAC9D,kCAAkC;IAC3B,MAAM,CAAC,IAAI,GAAG,WAAW,CAAA;IAEhC,sBAAsB;IACd,OAAO,GAAG,KAAK,CAAA;IACf,MAAM,GAAkB,IAAI,CAAA;IAC5B,aAAa,GAA0B,IAAI,CAAA;IAC3C,WAAW,GAA4B,IAAI,CAAA;IAEnD,sBAAsB;IACtB,YAAa,EAAU,EAAE,GAA4B,EAAE,IAA6B,EAAE,IAAW;QAC7F,KAAK,CAAC,EAAE,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,CAAC,CAAA;QAE1B,6CAA6C;QAC7C,IAAI,CAAC,SAAS,CAAC,EAAE,CAAC,CAAA;QAElB,wCAAwC;QACxC,IAAI,CAAC,KAAK,GAAI,OAAO,CAAA;QACrB,IAAI,CAAC,MAAM,GAAG,OAAO,CAAA;IACzB,CAAC;IAED,iBAAiB;IACjB,KAAK,CAAC,IAAI;QACN,8BAA8B;QAC9B,IAAI,CAAC,OAAO,GAAG,KAAK,CAAA;QAEpB,4CAA4C;QAC5C,MAAM,QAAQ,GAAI,wCAAwC;YACtD,+DAA+D,CAAA;QACnE,MAAM,QAAQ,GAAI,mBAAI,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,QAAQ,EAAE,OAAO,CAAC,CAAA;QAC1D,MAAM,SAAS,GAAG,mBAAI,CAAC,OAAO,CAAC,QAAQ,EAAE,mBAAmB,CAAC,CAAA;QAC7D,MAAM,IAAI,GAAG,MAAM,iBAAE,CAAC,QAAQ,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,CAAA;QAChE,IAAI,IAAI,KAAK,IAAI,EAAE,CAAC;YAChB,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,iCAAiC,QAAQ,GAAG,CAAC,CAAA;YAC9D,MAAM,iBAAE,CAAC,QAAQ,CAAC,KAAK,CAAC,QAAQ,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAA;YACtD,MAAM,QAAQ,GAAG,MAAM,eAAK,CAAC,GAAG,CAAC,QAAQ,EAAE;gBACvC,YAAY,EAAE,aAAa;gBAC3B,kBAAkB,EAAE,CAAC,aAAa,EAAE,EAAE;oBAClC,IAAI,aAAa,CAAC,KAAK,EAAE,CAAC;wBACtB,MAAM,OAAO,GAAG,CAAC,aAAa,CAAC,MAAM,GAAG,aAAa,CAAC,KAAK,CAAC,GAAG,GAAG,CAAA;wBAClE,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,yBAAyB,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAA;oBACpE,CAAC;gBACL,CAAC;aACJ,CAAC,CAAA;YACF,MAAM,iBAAE,CAAC,QAAQ,CAAC,SAAS,CAAC,SAAS,EAAE,MAAM,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAA;YAClE,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,8BAA8B,SAAS,GAAG,CAAC,CAAA;QAChE,CAAC;QAED;yDACiD;QACjD,IAAI,CAAC,aAAa,GAAG,IAAI,yBAAc,CAAC,CAAC,EAAE,IAAI,CAAC,MAAM,CAAC,eAAe,EAAE,KAAK,EAAE,CAAC,CAAC,CAAA;QACjF,IAAI,CAAC,WAAW,GAAK,IAAI,yBAAc,CAAC,CAAC,EAAE,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,eAAe,EAAE,CAAC,CAAC,CAAA;QAEjF,yBAAyB;QACzB,IAAI,CAAC,MAAM,GAAG,IAAI,4BAAM,CAAC,mBAAI,CAAC,OAAO,CAAC,SAAS,EAAE,iCAAiC,CAAC,EAAE;YACjF,UAAU,EAAE,EAAE,SAAS,EAAE;SAC5B,CAAC,CAAA;QACF,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;YAC5B,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,8BAA8B,GAAG,EAAE,CAAC,CAAA;YACtD,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAA;QACnC,CAAC,CAAC,CAAA;QACF,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,EAAE;YAC5B,IAAI,IAAI,KAAK,CAAC;gBACV,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,8CAA8C,IAAI,EAAE,CAAC,CAAA;;gBAEvE,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,gDAAgD,IAAI,EAAE,CAAC,CAAA;QAChF,CAAC,CAAC,CAAA;QAEF,mCAAmC;QACnC,MAAM,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YACxC,MAAM,OAAO,GAAG,UAAU,CAAC,GAAG,EAAE;gBAC5B,MAAM,CAAC,IAAI,KAAK,CAAC,4CAA4C,CAAC,CAAC,CAAA;YACnE,CAAC,EAAE,EAAE,GAAG,IAAI,CAAC,CAAA;YACb,MAAM,SAAS,GAAG,CAAC,GAAQ,EAAE,EAAE;gBAC3B,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,KAAK;oBAC7D,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,EAAE,GAAG,CAAC,OAAO,CAAC,CAAA;qBAC/B,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,OAAO,EAAE,CAAC;oBACvE,YAAY,CAAC,OAAO,CAAC,CAAA;oBACrB,IAAI,CAAC,MAAO,CAAC,GAAG,CAAC,SAAS,EAAE,SAAS,CAAC,CAAA;oBACtC,OAAO,EAAE,CAAA;gBACb,CAAC;qBACI,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;oBACxE,YAAY,CAAC,OAAO,CAAC,CAAA;oBACrB,IAAI,CAAC,MAAO,CAAC,GAAG,CAAC,SAAS,EAAE,SAAS,CAAC,CAAA;oBACtC,MAAM,CAAC,IAAI,KAAK,CAAC,GAAG,CAAC,OAAO,IAAI,2CAA2C,CAAC,CAAC,CAAA;gBACjF,CAAC;YACL,CAAC,CAAA;YACD,IAAI,CAAC,MAAO,CAAC,EAAE,CAAC,SAAS,EAAE,SAAS,CAAC,CAAA;YACrC,IAAI,CAAC,MAAO,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;gBAC/B,YAAY,CAAC,OAAO,CAAC,CAAA;gBACrB,MAAM,CAAC,GAAG,CAAC,CAAA;YACf,CAAC,CAAC,CAAA;QACN,CAAC,CAAC,CAAA;QAEF,mCAAmC;QACnC,MAAM,OAAO,GAAG,IAAI,GAAG,EAAoD,CAAA;QAC3E,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;YACxB,OAAO,CAAC,KAAK,EAAE,CAAA;QACnB,CAAC,CAAC,CAAA;QACF,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,GAAQ,EAAE,EAAE;YACnC,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,cAAc,EAAE,CAAC;gBACzE,MAAM,EAAE,GAAG,OAAO,CAAC,GAAG,CAAC,GAAG,CAAC,EAAE,CAAC,CAAA;gBAC9B,OAAO,CAAC,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAA;gBACtB,IAAI,EAAE;oBACF,EAAE,CAAC,GAAG,CAAC,IAAI,CAAC,CAAA;;oBAEZ,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,gDAAgD,GAAG,CAAC,EAAE,EAAE,CAAC,CAAA;YACrF,CAAC;iBACI,IAAI,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,GAAG,CAAC,IAAI,KAAK,KAAK;gBAClE,IAAI,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,EAAE,GAAG,CAAC,OAAO,CAAC,CAAA;;gBAEhC,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,gDAAgD,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,EAAE,CAAC,CAAA;QAClG,CAAC,CAAC,CAAA;QAEF,8BAA8B;QAC9B,IAAI,GAAG,GAAG,CAAC,CAAA;QACX,MAAM,aAAa,GAAG,KAAK,EAAE,OAAkC,EAAE,EAAE;YAC/D,IAAI,IAAI,CAAC,OAAO;gBACZ,OAAO,OAAO,CAAA;YAClB,MAAM,EAAE,GAAG,GAAG,GAAG,EAAE,EAAE,CAAA;YACrB,OAAO,IAAI,OAAO,CAA4B,CAAC,OAAO,EAAE,EAAE;gBACtD,OAAO,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,MAAM,EAAE,EAAE,GAAG,OAAO,CAAC,MAAM,CAAC,CAAA,CAAC,CAAC,CAAC,CAAA;gBAChD,IAAI,CAAC,MAAO,CAAC,WAAW,CAAC,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,CAAE,OAAO,CAAC,MAAM,CAAE,CAAC,CAAA;YAClF,CAAC,CAAC,CAAA;QACN,CAAC,CAAA;QAED,oCAAoC;QACpC,MAAM,IAAI,GAAG,IAAI,CAAA;QACjB,IAAI,CAAC,MAAM,GAAG,IAAI,qBAAM,CAAC,SAAS,CAAC;YAC/B,kBAAkB,EAAE,IAAI;YACxB,kBAAkB,EAAE,IAAI;YACxB,aAAa,EAAO,KAAK;YACzB,SAAS,CAAE,KAA4C,EAAE,QAAQ,EAAE,QAAQ;gBACvE,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;oBACf,QAAQ,CAAC,IAAI,KAAK,CAAC,0BAA0B,CAAC,CAAC,CAAA;oBAC/C,OAAM;gBACV,CAAC;gBACD,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,OAAO,CAAC;oBAC/B,QAAQ,CAAC,IAAI,KAAK,CAAC,4BAA4B,CAAC,CAAC,CAAA;qBAChD,CAAC;oBACF,gEAAgE;oBAChE,MAAM,aAAa,GAAG,IAAI,CAAC,aAAc,CAAC,YAAY,CAAC,KAAK,CAAC,OAAO,CAAC,CAAA;oBAErE,wCAAwC;oBACxC,MAAM,OAAO,GAAG,IAAI,CAAC,eAAe,CAAC,aAAa,CAAC,CAAA;oBAEnD,0BAA0B;oBAC1B,aAAa,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,CAAC,MAAiC,EAAE,EAAE;wBAC9D,wCAAwC;wBACxC,MAAM,GAAG,GAAG,IAAI,CAAC,eAAe,CAAC,MAAM,CAAC,CAAA;wBAExC,qEAAqE;wBACrE,MAAM,WAAW,GAAG,IAAI,CAAC,WAAY,CAAC,YAAY,CAAC,GAAG,CAAC,CAAA;wBAEvD,oBAAoB;wBACpB,KAAK,CAAC,OAAO,GAAG,WAAW,CAAA;wBAE3B,6BAA6B;wBAC7B,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;wBAChB,QAAQ,EAAE,CAAA;oBACd,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,GAAY,EAAE,EAAE;wBACtB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,CAAA;wBACnC,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,+BAA+B,KAAK,CAAC,OAAO,EAAE,CAAC,CAAA;wBACnE,QAAQ,CAAC,KAAK,CAAC,CAAA;oBACnB,CAAC,CAAC,CAAA;gBACN,CAAC;YACL,CAAC;YACD,KAAK,CAAE,QAAQ;gBACX,QAAQ,EAAE,CAAA;YACd,CAAC;SACJ,CAAC,CAAA;IACN,CAAC;IAED,kBAAkB;IAClB,KAAK,CAAC,KAAK;QACP,wBAAwB;QACxB,IAAI,CAAC,OAAO,GAAG,IAAI,CAAA;QAEnB,uBAAuB;QACvB,IAAI,IAAI,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC;YACvB,IAAI,CAAC,MAAM,CAAC,SAAS,EAAE,CAAA;YACvB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAA;QACtB,CAAC;QAED,uBAAuB;QACvB,IAAI,IAAI,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC;YACvB,MAAM,IAAI,CAAC,aAAa,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;YACrC,IAAI,CAAC,MAAM,GAAG,IAAI,CAAA;QACtB,CAAC;QAED,0BAA0B;QAC1B,IAAI,IAAI,CAAC,aAAa,KAAK,IAAI;YAC3B,IAAI,CAAC,aAAa,GAAG,IAAI,CAAA;QAC7B,IAAI,IAAI,CAAC,WAAW,KAAK,IAAI;YACzB,IAAI,CAAC,WAAW,GAAG,IAAI,CAAA;IAC/B,CAAC;;AApML,yCAqMC"}

package/speechflow-cli/dst/speechflow-node-a2a-meter.js

@@ -103,10 +103,10 @@ class SpeechFlowNodeA2AMeter extends speechflow_node_1.default {
                 return;
             /*  grab the accumulated chunk data  */
             const chunkData = this.chunkBuffer;
-            this.chunkBuffer = new Float32Array(0);
+            this.chunkBuffer = chunkData.subarray(samplesPerChunk);
             /*  update internal audio sample sliding window for LUFS-M  */
             if (chunkData.length > sampleWindow.length)
-                sampleWindow.set(chunkData.subarray(chunkData.length - sampleWindow.length), 0);
+                sampleWindow.set(chunkData.subarray(0, sampleWindow.length), 0);
             else {
                 sampleWindow.set(sampleWindow.subarray(chunkData.length), 0);
                 sampleWindow.set(chunkData, sampleWindow.length - chunkData.length);