speechflow 2.0.2 → 2.0.4

package/speechflow-cli/src/speechflow-node-t2t-translate.ts

@@ -13,7 +13,7 @@ import * as util                           from "./speechflow-util"
 import { LLM, type LLMCompleteMessage }    from "./speechflow-util-llm"
 
 /*  internal utility types  */
-type ConfigEntry = { systemPrompt: string, chat: LLMCompleteMessage[] }
+type ConfigEntry = { systemPrompt: { [ type: string ]: string }, chat: LLMCompleteMessage[] }
 type Config      = { [ key: string ]: ConfigEntry }
 
 /*  SpeechFlow node for LLM-based text-to-text translation  */
@@ -28,19 +28,30 @@ export default class SpeechFlowNodeT2TTranslate extends SpeechFlowNode {
     private setup: Config = {
         /*  English (EN) to German (DE) translation  */
         "en-de": {
-            systemPrompt:
-                "You are a translator.\n" +
-                "Output only the requested text.\n" +
-                "Do not use markdown.\n" +
-                "Do not chat.\n" +
-                "Do not show any explanations.\n" +
-                "Do not show any introduction.\n" +
-                "Do not show any preamble.\n" +
-                "Do not show any prolog.\n" +
-                "Do not show any epilog.\n" +
-                "Get to the point.\n" +
-                "Preserve the original meaning, tone, and nuance.\n" +
-                "Directly translate text from English (EN) to fluent and natural German (DE) language.\n",
+            systemPrompt: {
+                "any":
+                    "You are a translator.\n" +
+                    "Output only the requested text.\n" +
+                    "Do not use markdown.\n" +
+                    "Do not chat.\n" +
+                    "Do not show any explanations.\n" +
+                    "Do not show any introduction.\n" +
+                    "Do not show any preamble.\n" +
+                    "Do not show any prolog.\n" +
+                    "Do not show any epilog.\n" +
+                    "Get to the point.\n" +
+                    "Preserve the original meaning, tone, and nuance.\n" +
+                    "Directly translate text from English (EN) to fluent and natural German (DE) language.\n",
+                "translategemma":
+                    /*  ATTENTION: do not change this prompt, as TranslateGemma requires this fixed format!  */
+                    "You are a professional English (en) to German (de) translator. " +
+                    "Your goal is to accurately convey the meaning and nuances of the original " +
+                    "English text while adhering to German grammar, vocabulary, and cultural sensitivities. " +
+                    "Produce only the German translation, without any additional explanations or commentary. " +
+                    "Please translate the following English text into German:\n" +
+                    "\n" +
+                    "\n"
+            },
             chat: [
                 { role: "user",      content: "I love my wife." },
                 { role: "assistant", content: "Ich liebe meine Frau." },
@@ -53,19 +64,30 @@ export default class SpeechFlowNodeT2TTranslate extends SpeechFlowNode {
 
         /*  German (DE) to English (EN) translation  */
         "de-en": {
-            systemPrompt:
-                "You are a translator.\n" +
-                "Output only the requested text.\n" +
-                "Do not use markdown.\n" +
-                "Do not chat.\n" +
-                "Do not show any explanations.\n" +
-                "Do not show any introduction.\n" +
-                "Do not show any preamble.\n" +
-                "Do not show any prolog.\n" +
-                "Do not show any epilog.\n" +
-                "Get to the point.\n" +
-                "Preserve the original meaning, tone, and nuance.\n" +
-                "Directly translate text from German (DE) to fluent and natural English (EN) language.\n",
+            systemPrompt: {
+                "any":
+                    "You are a translator.\n" +
+                    "Output only the requested text.\n" +
+                    "Do not use markdown.\n" +
+                    "Do not chat.\n" +
+                    "Do not show any explanations.\n" +
+                    "Do not show any introduction.\n" +
+                    "Do not show any preamble.\n" +
+                    "Do not show any prolog.\n" +
+                    "Do not show any epilog.\n" +
+                    "Get to the point.\n" +
+                    "Preserve the original meaning, tone, and nuance.\n" +
+                    "Directly translate text from German (DE) to fluent and natural English (EN) language.\n",
+                "translategemma":
+                    /*  ATTENTION: do not change this prompt, as TranslateGemma requires this fixed format!  */
+                    "You are a professional German (de) to English (en) translator. " +
+                    "Your goal is to accurately convey the meaning and nuances of the original " +
+                    "German text while adhering to English grammar, vocabulary, and cultural sensitivities. " +
+                    "Produce only the English translation, without any additional explanations or commentary. " +
+                    "Please translate the following German text into English:\n" +
+                    "\n" +
+                    "\n"
+            },
             chat: [
                 { role: "user",      content: "Ich liebe meine Frau." },
                 { role: "assistant", content: "I love my wife." },
@@ -120,14 +142,17 @@ export default class SpeechFlowNodeT2TTranslate extends SpeechFlowNode {
         await this.llm.open()
 
         /*  provide text-to-text translation  */
-        const llm = this.llm!
+        const llm = this.llm
         const translate = async (text: string) => {
             const key = `${this.params.src}-${this.params.dst}`
             const cfg = this.setup[key]
             if (!cfg)
                 throw new Error(`unsupported language pair: ${key}`)
+            let systemPrompt = cfg.systemPrompt["any"]
+            if (this.params.model.match(/^translategemma/))
+                systemPrompt = cfg.systemPrompt["translategemma"]
             return llm.complete({
-                system:   cfg.systemPrompt,
+                system:   systemPrompt,
                 messages: cfg.chat,
                 prompt:   text
             })

package/speechflow-cli/src/speechflow-node-x2x-filter.ts

@@ -83,10 +83,10 @@ export default class SpeechFlowNodeX2XFilter extends SpeechFlowNode {
                 const num2 = coerceNum(val2)
                 return (
                     op === "<"  ? (num1 <  num2) :
-                    op === "<=" ? (num1 <= num2) :
-                    op === ">=" ? (num1 >= num2) :
-                    op === ">"  ? (num1 >  num2) :
-                    false
+                        op === "<=" ? (num1 <= num2) :
+                            op === ">=" ? (num1 >= num2) :
+                                op === ">"  ? (num1 >  num2) :
+                                    false
                 )
             }
         }

package/speechflow-cli/src/speechflow-node-xio-exec.ts

@@ -151,8 +151,8 @@ export default class SpeechFlowNodeXIOExec extends SpeechFlowNode {
         /*  terminate subprocess  */
         if (this.subprocess !== null) {
             /*  gracefully end stdin if in write or read/write mode  */
-            if ((this.params.mode === "w" || this.params.mode === "rw") && this.subprocess.stdin &&
-                !this.subprocess.stdin.destroyed && !this.subprocess.stdin.writableEnded) {
+            if ((this.params.mode === "w" || this.params.mode === "rw") && this.subprocess.stdin
+                && !this.subprocess.stdin.destroyed && !this.subprocess.stdin.writableEnded) {
                 await Promise.race([
                     new Promise<void>((resolve, reject) => {
                         this.subprocess!.stdin!.end((err?: Error) => {

package/speechflow-cli/src/speechflow-node-xio-file.ts

@@ -197,8 +197,8 @@ export default class SpeechFlowNodeXIOFile extends SpeechFlowNode {
             else {
                 /*  for stdio streams, just end without destroying  */
                 const stream = this.stream
-                if ((stream instanceof Stream.Writable || stream instanceof Stream.Duplex) &&
-                    (!stream.writableEnded && !stream.destroyed)) {
+                if ((stream instanceof Stream.Writable || stream instanceof Stream.Duplex)
+                    && (!stream.writableEnded && !stream.destroyed)) {
                     await Promise.race([
                         new Promise<void>((resolve, reject) => {
                             stream.end((err?: Error) => {

package/speechflow-cli/src/speechflow-node-xio-vban.ts

@@ -9,8 +9,10 @@ import Stream from "node:stream"
 
 /*  external dependencies  */
 import { DateTime }                        from "luxon"
-import { VBANServer, VBANAudioPacket,
-         EBitsResolutions, ECodecs }       from "vban"
+import {
+    VBANServer, VBANAudioPacket,
+    EBitsResolutions, ECodecs
+} from "vban"
 
 /*  internal dependencies  */
 import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"

package/speechflow-cli/src/speechflow-node-xio-webrtc.ts

@@ -306,7 +306,7 @@ export default class SpeechFlowNodeXIOWebRTC extends SpeechFlowNode {
             pc.ontrack = (event: { track: MediaStreamTrack }) => {
                 const track = event.track
                 if (track.kind === "audio") {
-                    this.log("info", `WebRTC audio track received from publisher`)
+                    this.log("info", "WebRTC audio track received from publisher")
 
                     /*  subscribe to incoming RTP packets  */
                     track.onReceiveRtp.subscribe((rtpPacket: RtpPacket) => {

package/speechflow-cli/src/speechflow-util-audio.ts

@@ -168,10 +168,12 @@ export function updateEnvelopeForChannel (
     return Math.sqrt(Math.max(currentEnv, 1e-12))
 }
 
-/*  helper functions for linear/decibel conversions  */
+/*  helper function for linear to decibel conversion  */
 export function lin2dB (x: number): number {
     return 20 * Math.log10(Math.max(x, 1e-12))
 }
+
+/*  helper function for decibel to linear conversion  */
 export function dB2lin (db: number): number {
     return Math.pow(10, db / 20)
 }
@@ -187,6 +189,7 @@ export class WebAudio {
         reject: (error: Error) => void
         timeout: ReturnType<typeof setTimeout>
     }>()
+    private captureListener: ((event: MessageEvent) => void) | null = null
 
     /*  construct object  */
     constructor (
@@ -222,7 +225,7 @@ export class WebAudio {
             numberOfInputs:  1,
             numberOfOutputs: 0
         })
-        this.captureNode.port.addEventListener("message", (event) => {
+        this.captureListener = (event) => {
             const { type, chunkId, data } = event.data ?? {}
             if (type === "capture-complete") {
                 const promise = this.pendingPromises.get(chunkId)
@@ -235,7 +238,8 @@ export class WebAudio {
                     promise.resolve(int16Data)
                 }
             }
-        })
+        }
+        this.captureNode.port.addEventListener("message", this.captureListener)
 
         /*  start ports  */
         this.sourceNode.port.start()
@@ -302,6 +306,10 @@ export class WebAudio {
             this.sourceNode = null
         }
         if (this.captureNode !== null) {
+            if (this.captureListener !== null) {
+                this.captureNode.port.removeEventListener("message", this.captureListener)
+                this.captureListener = null
+            }
             this.captureNode.disconnect()
             this.captureNode = null
         }

package/speechflow-cli/src/speechflow-util-stream.ts

@@ -208,11 +208,10 @@ export async function destroyStream (
     stream: Stream.Readable | Stream.Writable | Stream.Duplex | Stream.Transform
 ) {
     /*  signal the end for a writable stream  */
-    if ((stream instanceof Stream.Duplex    ||
-         stream instanceof Stream.Transform ||
-         stream instanceof Stream.Writable    ) &&
-        (!stream.writableEnded &&
-         !stream.destroyed       )                )
+    if ((  stream instanceof Stream.Duplex
+        || stream instanceof Stream.Transform
+        || stream instanceof Stream.Writable )
+        && (!stream.writableEnded && !stream.destroyed))
         await Promise.race([
             new Promise<void>((resolve) => {
                 stream.end(() => { resolve() })