speechflow 0.9.8 → 0.9.9

package/src/speechflow-node-t2t-openai.ts

@@ -0,0 +1,246 @@
+/*
+**  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>
+*/
+
+/*  standard dependencies  */
+import Stream           from "node:stream"
+
+/*  external dependencies  */
+import OpenAI           from "openai"
+
+/*  internal dependencies  */
+import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"
+
+/*  internal utility types  */
+type ConfigEntry = { systemPrompt: string, chat: OpenAI.ChatCompletionMessageParam[] }
+type Config      = { [ key: string ]: ConfigEntry }
+
+/*  SpeechFlow node for OpenAI/GPT text-to-text translation  */
+export default class SpeechFlowNodeOpenAI extends SpeechFlowNode {
+    /*  declare official node name  */
+    public static name = "openai"
+
+    /*  internal state  */
+    private openai: OpenAI | null = null
+
+    /*  internal LLM setup  */
+    private setup: Config = {
+        /*  English (EN) spellchecking only  */
+        "en-en": {
+            systemPrompt:
+                "You are a proofreader and spellchecker for English.\n" +
+                "Output only the corrected text.\n" +
+                "Do NOT use markdown.\n" +
+                "Do NOT give any explanations.\n" +
+                "Do NOT give any introduction.\n" +
+                "Do NOT give any comments.\n" +
+                "Do NOT give any preamble.\n" +
+                "Do NOT give any prolog.\n" +
+                "Do NOT give any epilog.\n" +
+                "Do NOT change the gammar.\n" +
+                "Do NOT use synonyms for words.\n" +
+                "Keep all words.\n" +
+                "Fill in missing commas.\n" +
+                "Fill in missing points.\n" +
+                "Fill in missing question marks.\n" +
+                "Fill in missing hyphens.\n" +
+                "Focus ONLY on the word spelling.\n" +
+                "The text you have to correct is:\n",
+            chat: [
+                { role: "user",   content: "I luve my wyfe" },
+                { role: "system", content: "I love my wife." },
+                { role: "user",   content: "The weether is wunderfull!" },
+                { role: "system", content: "The weather is wonderful!" },
+                { role: "user",   content: "The live awesome but I'm hungry." },
+                { role: "system", content: "The live is awesome, but I'm hungry." }
+            ]
+        },
+
+        /*  German (DE) spellchecking only  */
+        "de-de": {
+            systemPrompt:
+                "Du bist ein Korrekturleser und Rechtschreibprüfer für Deutsch.\n" +
+                "Gib nur den korrigierten Text aus.\n" +
+                "Benutze KEIN Markdown.\n" +
+                "Gib KEINE Erklärungen.\n" +
+                "Gib KEINE Einleitung.\n" +
+                "Gib KEINE Kommentare.\n" +
+                "Gib KEINE Preamble.\n" +
+                "Gib KEINEN Prolog.\n" +
+                "Gib KEINEN Epilog.\n" +
+                "Ändere NICHT die Grammatik.\n" +
+                "Verwende KEINE Synonyme für Wörter.\n" +
+                "Behalte alle Wörter bei.\n" +
+                "Füge fehlende Kommas ein.\n" +
+                "Füge fehlende Punkte ein.\n" +
+                "Füge fehlende Fragezeichen ein.\n" +
+                "Füge fehlende Bindestriche ein.\n" +
+                "Füge fehlende Gedankenstriche ein.\n" +
+                "Fokussiere dich NUR auf die Rechtschreibung der Wörter.\n" +
+                "Der von dir zu korrigierende Text ist:\n",
+            chat: [
+                { role: "user",   content: "Ich ljebe meine Frao" },
+                { role: "system", content: "Ich liebe meine Frau." },
+                { role: "user",   content: "Die Wedter ist wunderschoen." },
+                { role: "system", content: "Das Wetter ist wunderschön." },
+                { role: "user",   content: "Das Leben einfach großartig aber ich bin hungrig." },
+                { role: "system", content: "Das Leben ist einfach großartig, aber ich bin hungrig." }
+            ]
+        },
+
+        /*  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",
+            chat: [
+                { role: "user",   content: "I love my wife." },
+                { role: "system", content: "Ich liebe meine Frau." },
+                { role: "user",   content: "The weather is wonderful." },
+                { role: "system", content: "Das Wetter ist wunderschön." },
+                { role: "user",   content: "The live is awesome." },
+                { role: "system", content: "Das Leben ist einfach großartig." }
+            ]
+        },
+
+        /*  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",
+            chat: [
+                { role: "user",   content: "Ich liebe meine Frau." },
+                { role: "system", content: "I love my wife." },
+                { role: "user",   content: "Das Wetter ist wunderschön." },
+                { role: "system", content: "The weather is wonderful." },
+                { role: "user",   content: "Das Leben ist einfach großartig." },
+                { role: "system", content: "The live is awesome." }
+            ]
+        }
+    }
+
+    /*  construct node  */
+    constructor (id: string, cfg: { [ id: string ]: any }, opts: { [ id: string ]: any }, args: any[]) {
+        super(id, cfg, opts, args)
+
+        /*  declare node configuration parameters  */
+        this.configure({
+            src:   { type: "string", pos: 0, val: "de", match: /^(?:de|en)$/   },
+            dst:   { type: "string", pos: 1, val: "en", match: /^(?:de|en)$/   },
+            key:   { type: "string", val: process.env.SPEECHFLOW_KEY_OPENAI },
+            api:   { type: "string", val: "https://api.openai.com/v1", match: /^https?:\/\/.+?:\d+$/ },
+            model: { type: "string", val: "gpt-4o-mini" }
+        })
+
+        /*  tell effective mode  */
+        if (this.params.src === this.params.dst)
+            this.log("info", `OpenAI: operation mode: spellchecking for language "${this.params.src}"`)
+        else
+            this.log("info", `OpenAI: operation mode: translation from language "${this.params.src}"` +
+                ` to language "${this.params.dst}"`)
+
+        /*  declare node input/output format  */
+        this.input  = "text"
+        this.output = "text"
+    }
+
+    /*  open node  */
+    async open () {
+        /*  instantiate OpenAI API  */
+        this.openai = new OpenAI({
+            baseURL: this.params.api,
+            apiKey:  this.params.key,
+            dangerouslyAllowBrowser: true
+        })
+
+        /*  provide text-to-text translation  */
+        const translate = async (text: string) => {
+            const key = `${this.params.src}-${this.params.dst}`
+            const cfg = this.setup[key]
+            const stream = this.openai!.chat.completions.stream({
+                stream:                true,
+                model:                 this.params.model,
+                seed:                  null,
+                temperature:           0.7,
+                n:                     1,
+                messages: [
+                    { role: "system", content: cfg.systemPrompt },
+                    ...cfg.chat,
+                    { role: "user", content: text }
+                ]
+            })
+            const completion = await stream.finalChatCompletion()
+            const translation = completion.choices[0].message.content!
+            if (!stream.ended)
+                stream.abort()
+            return translation
+        }
+
+        /*  establish a duplex stream and connect it to OpenAI  */
+        this.stream = new Stream.Transform({
+            readableObjectMode: true,
+            writableObjectMode: true,
+            decodeStrings:      false,
+            transform (chunk: SpeechFlowChunk, encoding, callback) {
+                if (Buffer.isBuffer(chunk.payload))
+                    callback(new Error("invalid chunk payload type"))
+                else {
+                    if (chunk.payload === "") {
+                        this.push(chunk)
+                        callback()
+                    }
+                    else {
+                        translate(chunk.payload).then((payload) => {
+                            const chunkNew = chunk.clone()
+                            chunkNew.payload = payload
+                            this.push(chunkNew)
+                            callback()
+                        }).catch((err) => {
+                            callback(err)
+                        })
+                    }
+                }
+            },
+            final (callback) {
+                this.push(null)
+                callback()
+            }
+        })
+    }
+
+    /*  close node  */
+    async close () {
+        /*  close stream  */
+        if (this.stream !== null) {
+            this.stream.destroy()
+            this.stream = null
+        }
+
+        /*  shutdown OpenAI  */
+        if (this.openai !== null)
+            this.openai = null
+    }
+}
+

package/src/speechflow-node-t2t-transformers.ts

@@ -0,0 +1,244 @@
+/*
+**  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>
+*/
+
+/*  standard dependencies  */
+import path             from "node:path"
+import Stream           from "node:stream"
+
+/*  external dependencies  */
+import * as Transformers from "@huggingface/transformers"
+
+/*  internal dependencies  */
+import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"
+
+/*  internal utility types  */
+type ConfigEntry = { systemPrompt: string, chat: Array<{ role: string, content: string }> }
+type Config      = { [ key: string ]: ConfigEntry }
+
+/*  SpeechFlow node for Transformers text-to-text translation  */
+export default class SpeechFlowNodeTransformers extends SpeechFlowNode {
+    /*  declare official node name  */
+    public static name = "transformers"
+
+    /*  internal state  */
+    private translator: Transformers.TranslationPipeline    | null = null
+    private generator:  Transformers.TextGenerationPipeline | null = null
+
+    /*  internal LLM setup  */
+    private setup: Config = {
+        /*  SmolLM3: English (EN) to German (DE) translation  */
+        "SmolLM3:en-de": {
+            systemPrompt:
+                "/no_think\n" +
+                "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",
+            chat: [
+                { role: "user",   content: "I love my wife." },
+                { role: "assistant", content: "Ich liebe meine Frau." },
+                { role: "user",   content: "The weather is wonderful." },
+                { role: "assistant", content: "Das Wetter ist wunderschön." },
+                { role: "user",   content: "The live is awesome." },
+                { role: "assistant", content: "Das Leben ist einfach großartig." }
+            ]
+        },
+
+        /*  SmolLM3: German (DE) to English (EN) translation  */
+        "SmolLM3:de-en": {
+            systemPrompt:
+                "/no_think\n" +
+                "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",
+            chat: [
+                { role: "user",   content: "Ich liebe meine Frau." },
+                { role: "assistant", content: "I love my wife." },
+                { role: "user",   content: "Das Wetter ist wunderschön." },
+                { role: "assistant", content: "The weather is wonderful." },
+                { role: "user",   content: "Das Leben ist einfach großartig." },
+                { role: "assistant", content: "The live is awesome." }
+            ]
+        }
+    }
+
+    /*  construct node  */
+    constructor (id: string, cfg: { [ id: string ]: any }, opts: { [ id: string ]: any }, args: any[]) {
+        super(id, cfg, opts, args)
+
+        /*  declare node configuration parameters  */
+        this.configure({
+            src:   { type: "string", pos: 0, val: "de", match: /^(?:de|en)$/ },
+            dst:   { type: "string", pos: 1, val: "en", match: /^(?:de|en)$/ },
+            model: { type: "string", val: "OPUS", match: /^(?:OPUS|SmolLM3)$/ }
+        })
+
+        /*  sanity check parameters  */
+        if (this.params.src === this.params.dst)
+            throw new Error("source and destination languages cannot be the same")
+
+        /*  declare node input/output format  */
+        this.input  = "text"
+        this.output = "text"
+    }
+
+    /*  open node  */
+    async open () {
+        /*  instantiate Transformers engine and model  */
+        let model: string = ""
+        const progressState = new Map<string, number>()
+        const progressCallback = (progress: any) => {
+            let artifact = model
+            if (typeof progress.file === "string")
+                artifact += `:${progress.file}`
+            let percent = 0
+            if (typeof progress.loaded === "number" && typeof progress.total === "number")
+                percent = (progress.loaded as number / progress.total as number) * 100
+            else if (typeof progress.progress === "number")
+                percent = progress.progress
+            if (percent > 0)
+                progressState.set(artifact, percent)
+        }
+        const interval = setInterval(() => {
+            for (const [ artifact, percent ] of progressState) {
+                this.log("info", `downloaded ${percent.toFixed(2)}% of artifact "${artifact}"`)
+                if (percent >= 1.0)
+                    progressState.delete(artifact)
+            }
+        }, 1000)
+        if (this.params.model === "OPUS") {
+            model = `onnx-community/opus-mt-${this.params.src}-${this.params.dst}`
+            this.translator = await Transformers.pipeline("translation", model, {
+                cache_dir: path.join(this.config.cacheDir, "opus"),
+                dtype:     "q4",
+                device:    "gpu",
+                progress_callback: progressCallback
+            })
+            clearInterval(interval)
+            if (this.translator === null)
+                throw new Error("failed to instantiate translator pipeline")
+        }
+        else if (this.params.model === "SmolLM3") {
+            model = "HuggingFaceTB/SmolLM3-3B-ONNX"
+            this.generator = await Transformers.pipeline("text-generation", model, {
+                cache_dir: path.join(this.config.cacheDir, "transformers"),
+                dtype:     "q4",
+                device:    "gpu",
+                progress_callback: progressCallback
+            })
+            clearInterval(interval)
+            if (this.generator === null)
+                throw new Error("failed to instantiate generator pipeline")
+        }
+        else
+            throw new Error("invalid model")
+
+        /*  provide text-to-text translation  */
+        const translate = async (text: string) => {
+            if (this.params.model === "OPUS") {
+                const result = await this.translator!(text)
+                return Array.isArray(result) ?
+                    (result[0] as Transformers.TranslationSingle).translation_text :
+                    (result as Transformers.TranslationSingle).translation_text
+            }
+            else if (this.params.model === "SmolLM3") {
+                const key = `SmolLM3:${this.params.src}-${this.params.dst}`
+                const cfg = this.setup[key]
+                const messages = [
+                    { role: "system", content: cfg.systemPrompt },
+                    ...cfg.chat,
+                    { role: "user", content: text }
+                ]
+                const result = await this.generator!(messages, {
+                    max_new_tokens: 100,
+                    temperature:    0.6,
+                    top_p:          0.95,
+                    streamer: new Transformers.TextStreamer(this.generator!.tokenizer, {
+                        skip_prompt:         true,
+                        skip_special_tokens: true
+                    })
+                })
+                const generatedText = Array.isArray(result) ?
+                    (result[0] as Transformers.TextGenerationSingle).generated_text :
+                    (result as Transformers.TextGenerationSingle).generated_text
+                const response = typeof generatedText === "string" ?
+                    generatedText :
+                    generatedText.at(-1)!.content
+                return response
+            }
+            else
+                throw new Error("invalid model")
+        }
+
+        /*  establish a duplex stream and connect it to Transformers  */
+        this.stream = new Stream.Transform({
+            readableObjectMode: true,
+            writableObjectMode: true,
+            decodeStrings:      false,
+            transform (chunk: SpeechFlowChunk, encoding, callback) {
+                if (Buffer.isBuffer(chunk.payload))
+                    callback(new Error("invalid chunk payload type"))
+                else {
+                    if (chunk.payload === "") {
+                        this.push(chunk)
+                        callback()
+                    }
+                    else {
+                        translate(chunk.payload).then((payload) => {
+                            chunk = chunk.clone()
+                            chunk.payload = payload
+                            this.push(chunk)
+                            callback()
+                        }).catch((err) => {
+                            callback(err)
+                        })
+                    }
+                }
+            },
+            final (callback) {
+                this.push(null)
+                callback()
+            }
+        })
+    }
+
+    /*  close node  */
+    async close () {
+        /*  close stream  */
+        if (this.stream !== null) {
+            this.stream.destroy()
+            this.stream = null
+        }
+
+        /*  shutdown Transformers  */
+        if (this.translator !== null) {
+            this.translator.dispose()
+            this.translator = null
+        }
+        if (this.generator !== null) {
+            this.generator.dispose()
+            this.generator = null
+        }
+    }
+}
+

package/src/speechflow-node-x2x-trace.ts

@@ -52,7 +52,7 @@ export default class SpeechFlowNodeTrace extends SpeechFlowNode {
                 const fmt = (t: Duration) => t.toFormat("hh:mm:ss.SSS")
                 if (Buffer.isBuffer(chunk.payload)) {
                     if (type === "audio")
-                        log("info", `writing ${type} chunk: start=${fmt(chunk.timestampStart)} ` +
+                        log("debug", `writing ${type} chunk: start=${fmt(chunk.timestampStart)} ` +
                             `end=${fmt(chunk.timestampEnd)} kind=${chunk.kind} type=${chunk.type} ` +
                             `payload-type=Buffer payload-bytes=${chunk.payload.byteLength}`)
                     else
@@ -60,7 +60,7 @@ export default class SpeechFlowNodeTrace extends SpeechFlowNode {
                 }
                 else {
                     if (type === "text")
-                        log("info", `writing ${type} chunk: start=${fmt(chunk.timestampStart)} ` +
+                        log("debug", `writing ${type} chunk: start=${fmt(chunk.timestampStart)} ` +
                             `end=${fmt(chunk.timestampEnd)} kind=${chunk.kind} type=${chunk.type}` +
                             `payload-type=String payload-length=${chunk.payload.length} ` +
                             `payload-encoding=${encoding} payload-content="${chunk.payload.toString()}"`)