speechflow 0.9.7 → 0.9.9

package/src/speechflow.ts

@@ -13,6 +13,7 @@ import { EventEmitter }         from "node:events"
 import { DateTime }             from "luxon"
 import CLIio                    from "cli-io"
 import yargs                    from "yargs"
+import { hideBin }              from "yargs/helpers"
 import jsYAML                   from "js-yaml"
 import FlowLink                 from "flowlink"
 import objectPath               from "object-path"
@@ -36,6 +37,7 @@ let cli: CLIio | null = null
     })
 
     /*  parse command-line arguments  */
+    const coerce = (arg: string) => Array.isArray(arg) ? arg[arg.length - 1] : arg
     const args = await yargs()
         /* eslint @stylistic/indent: off */
         .usage(
@@ -49,28 +51,69 @@ let cli: CLIio | null = null
             "[-c|--config <id>@<yaml-config-file>] " +
             "[<argument> [...]]"
         )
-        .help("h").alias("h", "help").default("h", false)
-            .describe("h", "show usage help")
-        .boolean("V").alias("V", "version").default("V", false)
-            .describe("V", "show program version information")
-        .string("v").nargs("v", 1).alias("v", "log-level").default("v", "warning")
-            .describe("v", "level for verbose logging ('none', 'error', 'warning', 'info', 'debug')")
-        .string("C").nargs("C", 1).alias("C", "cache").default("C", path.join(dataDir, "cache"))
-            .describe("C", "directory for cached files (primarily AI model files)")
-        .string("e").nargs("e", 1).alias("e", "expression").default("e", "")
-            .describe("e", "FlowLink expression string")
-        .string("f").nargs("f", 1).alias("f", "file").default("f", "")
-            .describe("f", "FlowLink expression file")
-        .string("c").nargs("c", 1).alias("c", "config-file").default("c", "")
-            .describe("c", "FlowLink expression reference into YAML file (in format <id>@<file>)")
+        .option("V", {
+            alias:    "version",
+            type:     "boolean",
+            array:    false,
+            coerce,
+            default:  false,
+            describe: "show program version information"
+        })
+        .option("v", {
+            alias:    "log-level",
+            type:     "string",
+            array:    false,
+            coerce,
+            nargs:    1,
+            default:  "warning",
+            describe: "level for verbose logging ('none', 'error', 'warning', 'info', 'debug')"
+        })
+        .option("C", {
+            alias:    "cache",
+            type:     "string",
+            array:    false,
+            coerce,
+            nargs:    1,
+            default:  path.join(dataDir, "cache"),
+            describe: "directory for cached files (primarily AI model files)"
+        })
+        .option("e", {
+            alias:    "expression",
+            type:     "string",
+            array:    false,
+            coerce,
+            nargs:    1,
+            default:  "",
+            describe: "FlowLink expression string"
+        })
+        .option("f", {
+            alias:    "file",
+            type:     "string",
+            array:    false,
+            coerce,
+            nargs:    1,
+            default:  "",
+            describe: "FlowLink expression file"
+        })
+        .option("c", {
+            alias:    "config",
+            type:     "string",
+            array:    false,
+            coerce,
+            nargs:    1,
+            default:  "",
+            describe: "FlowLink expression reference into YAML file (in format <id>@<file>)"
+        })
+        .help("h", "show usage help")
+        .alias("h", "help")
+        .showHelpOnFail(true)
         .version(false)
         .strict()
-        .showHelpOnFail(true)
         .demand(0)
-        .parse(process.argv.slice(2))
+        .parse(hideBin(process.argv))
 
     /*  short-circuit version request  */
-    if (args.version) {
+    if (args.V) {
         process.stderr.write(`SpeechFlow ${pkg["x-stdver"]} (${pkg["x-release"]}) <${pkg.homepage}>\n`)
         process.stderr.write(`${pkg.description}\n`)
         process.stderr.write(`Copyright (c) 2024-2025 ${pkg.author.name} <${pkg.author.url}>\n`)
@@ -81,7 +124,7 @@ let cli: CLIio | null = null
     /*  establish CLI environment  */
     cli = new CLIio({
         encoding:  "utf8",
-        logLevel:  args.logLevel,
+        logLevel:  args.v,
         logTime:   true,
         logPrefix: pkg.name
     })
@@ -112,28 +155,28 @@ let cli: CLIio | null = null
 
     /*  sanity check usage  */
     let n = 0
-    if (typeof args.expression     === "string" && args.expression     !== "") n++
-    if (typeof args.expressionFile === "string" && args.expressionFile !== "") n++
-    if (typeof args.configFile     === "string" && args.configFile     !== "") n++
+    if (typeof args.e === "string" && args.e !== "") n++
+    if (typeof args.f === "string" && args.f !== "") n++
+    if (typeof args.c === "string" && args.c !== "") n++
     if (n !== 1)
         throw new Error("cannot use more than one FlowLink specification source (either option -e, -f or -c)")
 
     /*  read configuration  */
     let config = ""
-    if (typeof args.expression === "string" && args.expression !== "")
-        config = args.expression
-    else if (typeof args.expressionFile === "string" && args.expressionFile !== "")
-        config = await cli.input(args.expressionFile, { encoding: "utf8" })
-    else if (typeof args.configFile === "string" && args.configFile !== "") {
-        const m = args.configFile.match(/^(.+?)@(.+)$/)
+    if (typeof args.e === "string" && args.e !== "")
+        config = args.e
+    else if (typeof args.f === "string" && args.f !== "")
+        config = await cli.input(args.f, { encoding: "utf8" })
+    else if (typeof args.c === "string" && args.c !== "") {
+        const m = args.c.match(/^(.+?)@(.+)$/)
         if (m === null)
-            throw new Error("invalid configuration file specification (expected \"<key>@<yaml-config-file>\")")
-        const [ , key, file ] = m
+            throw new Error("invalid configuration file specification (expected \"<id>@<yaml-config-file>\")")
+        const [ , id, file ] = m
         const yaml = await cli.input(file, { encoding: "utf8" })
         const obj: any = jsYAML.load(yaml)
-        if (obj[key] === undefined)
-            throw new Error(`no such key "${key}" found in configuration file`)
-        config = obj[key] as string
+        if (obj[id] === undefined)
+            throw new Error(`no such id "${id}" found in configuration file`)
+        config = obj[id] as string
     }
 
     /*  track the available SpeechFlow nodes  */
@@ -145,11 +188,14 @@ let cli: CLIio | null = null
         "./speechflow-node-a2a-wav.js",
         "./speechflow-node-a2t-deepgram.js",
         "./speechflow-node-t2a-elevenlabs.js",
+        "./speechflow-node-t2a-kokoro.js",
         "./speechflow-node-t2t-deepl.js",
-        "./speechflow-node-t2t-format.js",
-        "./speechflow-node-t2t-gemma.js",
+        "./speechflow-node-t2t-openai.js",
+        "./speechflow-node-t2t-ollama.js",
+        "./speechflow-node-t2t-transformers.js",
         "./speechflow-node-t2t-opus.js",
         "./speechflow-node-t2t-subtitle.js",
+        "./speechflow-node-t2t-format.js",
         "./speechflow-node-x2x-trace.js",
         "./speechflow-node-xio-device.js",
         "./speechflow-node-xio-file.js",
@@ -200,7 +246,7 @@ let cli: CLIio | null = null
         audioLittleEndian: true,
         audioSampleRate:   48000,
         textEncoding:      "utf8",
-        cacheDir:          args.cache
+        cacheDir:          args.C
     }
     let ast: unknown
     try {
@@ -254,9 +300,9 @@ let cli: CLIio | null = null
     }
     catch (err) {
         if (err instanceof Error && err.name === "FlowLinkError")
-            cli!.log("error", `failed to materialize SpeechFlow configuration: ${err.toString()}"`)
+            cli!.log("error", `failed to materialize SpeechFlow configuration: ${err.toString()}`)
         else if (err instanceof Error)
-            cli!.log("error", `failed to materialize SpeechFlow configuration: ${err.message}"`)
+            cli!.log("error", `failed to materialize SpeechFlow configuration: ${err.message}`)
         else
             cli!.log("error", "failed to materialize SpeechFlow configuration: internal error")
         process.exit(1)
@@ -358,7 +404,7 @@ let cli: CLIio | null = null
     }
 
     /*  start of internal stream processing  */
-    cli!.log("info", "everything established -- stream processing in SpeechFlow graph starts")
+    cli!.log("info", "**** everything established -- stream processing in SpeechFlow graph starts ****")
 
     /*  gracefully shutdown process  */
     let shuttingDown = false
@@ -367,9 +413,9 @@ let cli: CLIio | null = null
             return
         shuttingDown = true
         if (signal === "finished")
-            cli!.log("info", "streams of all nodes finished -- shutting down service")
+            cli!.log("info", "**** streams of all nodes finished -- shutting down service ****")
         else
-            cli!.log("warning", `received signal ${signal} -- shutting down service`)
+            cli!.log("warning", `**** received signal ${signal} -- shutting down service ****`)
 
         /*  graph processing: PASS 1: disconnect node streams  */
         for (const node of graphNodes) {

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

@@ -1,111 +0,0 @@
-/*
-**  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"
-
-/*  SpeechFlow node for OPUS text-to-text translation  */
-export default class SpeechFlowNodeOPUS extends SpeechFlowNode {
-    /*  declare official node name  */
-    public static name = "opus"
-
-    /*  internal state  */
-    private translator: Transformers.TranslationPipeline | null = null
-
-    /*  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)$/ },
-            dst: { type: "string", pos: 1, val: "en", match: /^(?:de|en)$/ }
-        })
-
-        /*  sanity check situation  */
-        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 OPUS  */
-        const 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"
-        })
-        if (this.translator === null)
-            throw new Error("failed to instantiate translator pipeline")
-
-        /*  provide text-to-text translation  */
-        const translate = async (text: string) => {
-            const result = await this.translator!(text)
-            return Array.isArray(result) ?
-                (result[0] as Transformers.TranslationSingle).translation_text :
-                (result as Transformers.TranslationSingle).translation_text
-        }
-
-        /*  establish a duplex stream and connect it to Ollama  */
-        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 OPUS  */
-        if (this.translator !== null) {
-            this.translator.dispose()
-            this.translator = null
-        }
-    }
-}
-