speechflow 2.0.0 → 2.0.2

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

@@ -5,7 +5,7 @@
 */
 
 /*  standard dependencies  */
-import Stream from "node:stream"
+import Stream                                    from "node:stream"
 
 /*  external dependencies  */
 import { TranslateClient, TranslateTextCommand } from "@aws-sdk/client-translate"
@@ -65,8 +65,6 @@ export default class SpeechFlowNodeT2TAmazon extends SpeechFlowNode {
                 secretAccessKey: this.params.secKey
             }
         })
-        if (this.client === null)
-            throw new Error("failed to establish Amazon Translate client")
 
         /*  provide text-to-text translation  */
         const maxRetries = 10

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

@@ -5,10 +5,10 @@
 */
 
 /*  standard dependencies  */
-import Stream from "node:stream"
+import Stream                              from "node:stream"
 
 /*  external dependencies  */
-import * as DeepL from "deepl-node"
+import * as DeepL                          from "deepl-node"
 
 /*  internal dependencies  */
 import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"

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

@@ -5,7 +5,7 @@
 */
 
 /*  standard dependencies  */
-import Stream from "node:stream"
+import Stream                       from "node:stream"
 
 /*  external dependencies  */
 import { TranslationServiceClient } from "@google-cloud/translate"

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

@@ -68,18 +68,19 @@ export default class SpeechFlowNodeT2TOPUS extends SpeechFlowNode {
         }, 1000)
 
         /*  instantiate Transformers engine and model  */
-        const pipeline = Transformers.pipeline("translation", model, {
-            cache_dir: path.join(this.config.cacheDir, "transformers"),
-            dtype:     "q4",
-            device:    "auto",
-            progress_callback: progressCallback
-        })
-        this.translator = await pipeline
-        if (this.translator === null)
-            throw new Error("failed to instantiate translator pipeline")
-
-        /*  clear progress interval again  */
-        clearInterval(interval)
+        try {
+            const pipeline = Transformers.pipeline("translation", model, {
+                cache_dir: path.join(this.config.cacheDir, "transformers"),
+                dtype:     "q4",
+                device:    "auto",
+                progress_callback: progressCallback
+            })
+            this.translator = await pipeline
+        }
+        finally {
+            /*  clear progress interval again  */
+            clearInterval(interval)
+        }
 
         /*  provide text-to-text translation  */
         const translate = async (text: string) => {
@@ -120,17 +121,17 @@ export default class SpeechFlowNodeT2TOPUS extends SpeechFlowNode {
 
     /*  close node  */
     async close () {
-        /*  shutdown stream  */
-        if (this.stream !== null) {
-            await util.destroyStream(this.stream)
-            this.stream = null
-        }
-
         /*  shutdown Transformers  */
         if (this.translator !== null) {
             this.translator.dispose()
             this.translator = null
         }
+
+        /*  shutdown stream  */
+        if (this.stream !== null) {
+            await util.destroyStream(this.stream)
+            this.stream = null
+        }
     }
 }
 

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

@@ -74,7 +74,7 @@ export default class SpeechFlowNodeT2TPunctuation extends SpeechFlowNode {
                 "Gib KEINE Erklärungen.\n" +
                 "Gib KEINE Einleitung.\n" +
                 "Gib KEINE Kommentare.\n" +
-                "Gib KEINE Preamble.\n" +
+                "Gib KEINE Präambel.\n" +
                 "Gib KEINEN Prolog.\n" +
                 "Gib KEINEN Epilog.\n" +
                 "Ändere NICHT die Wörter.\n" +
@@ -133,8 +133,7 @@ export default class SpeechFlowNodeT2TPunctuation extends SpeechFlowNode {
             api:         this.params.api,
             model:       this.params.model,
             key:         this.params.key,
-            temperature: 0.7,
-            topP:        0.5
+            temperature: 0.7
         })
         this.llm.on("log", (level: string, message: string) => {
             this.log(level as "info" | "warning" | "error", message)

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

@@ -66,7 +66,7 @@ export default class SpeechFlowNodeT2TSpellcheck extends SpeechFlowNode {
                 "Gib KEINE Erklärungen.\n" +
                 "Gib KEINE Einleitung.\n" +
                 "Gib KEINE Kommentare.\n" +
-                "Gib KEINE Preamble.\n" +
+                "Gib KEINE Präambel.\n" +
                 "Gib KEINEN Prolog.\n" +
                 "Gib KEINEN Epilog.\n" +
                 "Ändere NICHT die Grammatik.\n" +
@@ -120,8 +120,7 @@ export default class SpeechFlowNodeT2TSpellcheck extends SpeechFlowNode {
             api:         this.params.api,
             model:       this.params.model,
             key:         this.params.key,
-            temperature: 0.7,
-            topP:        0.5
+            temperature: 0.7
         })
         this.llm.on("log", (level: string, message: string) => {
             this.log(level as "info" | "warning" | "error", message)

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

@@ -191,7 +191,7 @@ export default class SpeechFlowNodeT2TSubtitle extends SpeechFlowNode {
                 for (const block of blocks) {
                     const lines = block.trim().split(/\r?\n/)
                     if (lines.length < 2) {
-                        this.log("warning", "SRT block contains less than 2 lines")
+                        this.log("warning", "SRT block contains fewer than 2 lines")
                         continue
                     }
 
@@ -231,7 +231,7 @@ export default class SpeechFlowNodeT2TSubtitle extends SpeechFlowNode {
                 for (const block of blocks) {
                     const lines = block.trim().split(/\r?\n/)
                     if (lines.length < 1) {
-                        this.log("warning", "VTT block contains less than 1 line")
+                        this.log("warning", "VTT block contains fewer than 1 line")
                         continue
                     }
 
@@ -394,15 +394,18 @@ export default class SpeechFlowNodeT2TSubtitle extends SpeechFlowNode {
                     h.response({}).code(204)
             })
 
+            /*  start HAPI server  */
             await this.hapi.start()
             this.log("info", `HAPI: started REST/WebSocket network service: http://${this.params.addr}:${this.params.port}`)
 
+            /*  helper to emit chunks to WebSocket peers  */
             const emit = (chunk: SpeechFlowChunk) => {
                 const data = JSON.stringify(chunk)
                 for (const info of wsPeers.values())
                     info.ws.send(data)
             }
 
+            /*  establish writable stream  */
             this.stream = new Stream.Writable({
                 objectMode:     true,
                 decodeStrings:  false,

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

@@ -60,7 +60,7 @@ export default class SpeechFlowNodeT2TSummary extends SpeechFlowNode {
                 "Gib KEINE Erklärungen.\n" +
                 "Gib KEINE Einleitung.\n" +
                 "Gib KEINE Kommentare.\n" +
-                "Gib KEINE Prämbel.\n" +
+                "Gib KEINE Präambel.\n" +
                 "Gib KEINEN Prolog.\n" +
                 "Gib KEINEN Epilog.\n" +
                 "Komme auf den Punkt.\n" +
@@ -119,8 +119,7 @@ export default class SpeechFlowNodeT2TSummary extends SpeechFlowNode {
             api:         this.params.api,
             model:       this.params.model,
             key:         this.params.key,
-            temperature: 0.7,
-            topP:        0.5
+            temperature: 0.7
         })
         this.llm.on("log", (level: string, message: string) => {
             this.log(level as "info" | "warning" | "error", message)

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

@@ -112,8 +112,7 @@ export default class SpeechFlowNodeT2TTranslate extends SpeechFlowNode {
             api:         this.params.api,
             model:       this.params.model,
             key:         this.params.key,
-            temperature: 0.7,
-            topP:        0.5
+            temperature: 0.7
         })
         this.llm.on("log", (level: string, message: string) => {
             this.log(level as "info" | "warning" | "error", message)

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

@@ -120,6 +120,8 @@ export default class SpeechFlowNodeX2XFilter extends SpeechFlowNode {
                     val1 = chunk.timestampStart.toMillis()
                 else if (self.params.var === "time:end")
                     val1 = chunk.timestampEnd.toMillis()
+                else
+                    val1 = undefined
                 if (comparison(val1, self.params.op, val2)) {
                     self.log("info", `[${self.params.name}]: passing through ${chunk.type} chunk`)
                     this.push(chunk)

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

@@ -198,6 +198,7 @@ export default class SpeechFlowNodeXIOExec extends SpeechFlowNode {
             this.subprocess.removeAllListeners("error")
             this.subprocess.removeAllListeners("exit")
 
+            /*  clear subprocess reference  */
             this.subprocess = null
         }
 

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

@@ -54,7 +54,7 @@ export default class SpeechFlowNodeXIOFile extends SpeechFlowNode {
     /*  open node  */
     async open () {
         /*  determine how many bytes we need per chunk when
-            the chunk should be of the required duration/size */
+            the chunk should be of the required duration/size  */
         const highWaterMarkAudio = (
             this.config.audioSampleRate *
             (this.config.audioBitDepth / 8)
@@ -139,11 +139,10 @@ export default class SpeechFlowNodeXIOFile extends SpeechFlowNode {
                             const payload = Buffer.isBuffer(chunk.payload) ?
                                 chunk.payload : Buffer.from(chunk.payload)
                             const seekPosition = chunk.meta.get("chunk:seek") as number | undefined
-                            if (seekPosition !== undefined) {
+                            if (seekPosition !== undefined)
                                 /*  seek to specified position and write (overload)  */
                                 fs.write(self.fd!, payload, 0, payload.byteLength, seekPosition, callback)
-                            }
-                            else {
+                            else
                                 /*  append at current position  */
                                 fs.write(self.fd!, payload, 0, payload.byteLength, writePosition, (err) => {
                                     if (err)
@@ -153,7 +152,6 @@ export default class SpeechFlowNodeXIOFile extends SpeechFlowNode {
                                         callback()
                                     }
                                 })
-                            }
                         },
                         final (callback) {
                             callback()

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

@@ -21,8 +21,8 @@ export default class SpeechFlowNodeXIOMQTT extends SpeechFlowNode {
     public static name = "xio-mqtt"
 
     /*  internal state  */
-    private broker: MQTT.MqttClient | null = null
-    private clientId: string = (new UUID(1)).format()
+    private broker:     MQTT.MqttClient | null                   = null
+    private clientId:   string                                   = (new UUID(1)).format()
     private chunkQueue: util.SingleQueue<SpeechFlowChunk> | null = null
 
     /*  construct node  */

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

@@ -29,11 +29,11 @@ export default class SpeechFlowNodeXIOVBAN extends SpeechFlowNode {
     public static name = "xio-vban"
 
     /*  internal state  */
-    private server:     VBANServer                        | null = null
-    private chunkQueue: util.SingleQueue<SpeechFlowChunk> | null = null
-    private frameCounter                                         = 0
-    private targetAddress                                        = ""
-    private targetPort                                           = 0
+    private server:        VBANServer                        | null = null
+    private chunkQueue:    util.SingleQueue<SpeechFlowChunk> | null = null
+    private frameCounter                                            = 0
+    private targetAddress                                           = ""
+    private targetPort                                              = 0
 
     /*  construct node  */
     constructor (id: string, cfg: { [ id: string ]: any }, opts: { [ id: string ]: any }, args: any[]) {

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

@@ -154,6 +154,7 @@ export default class SpeechFlowNodeXIOWebRTC extends SpeechFlowNode {
             this.pcmBuffer = this.pcmBuffer.subarray(this.pcmBuffer.length - maxBufferSize)
         }
 
+        /*  process full Opus frames from buffer  */
         while (this.pcmBuffer.length >= this.OPUS_FRAME_BYTES) {
             const frame = this.pcmBuffer.subarray(0, this.OPUS_FRAME_BYTES)
             this.pcmBuffer = this.pcmBuffer.subarray(this.OPUS_FRAME_BYTES)
@@ -418,6 +419,7 @@ export default class SpeechFlowNodeXIOWebRTC extends SpeechFlowNode {
                 const isPublisher  = hasSendonly || hasSendrecv
                 const isViewer     = hasRecvonly
 
+                /*  handle protocol based on mode  */
                 if (self.params.mode === "r" && isPublisher)
                     /*  in read mode, accept WHIP publishers  */
                     await self.handleWHIP(res, body)

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

@@ -15,7 +15,7 @@ import ReconnWebSocket, { ErrorEvent } from "@opensumi/reconnecting-websocket"
 import SpeechFlowNode, { SpeechFlowChunk } from "./speechflow-node"
 import * as util                           from "./speechflow-util"
 
-/*  SpeechFlow node for Websocket networking  */
+/*  SpeechFlow node for WebSocket networking  */
 export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
     /*  declare official node name  */
     public static name = "xio-websocket"
@@ -38,9 +38,9 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
 
         /*  sanity check parameters  */
         if (this.params.listen !== "" && this.params.connect !== "")
-            throw new Error("Websocket node cannot listen and connect at the same time")
+            throw new Error("WebSocket node cannot listen and connect at the same time")
         else if (this.params.listen === "" && this.params.connect === "")
-            throw new Error("Websocket node requires either listen or connect mode")
+            throw new Error("WebSocket node requires either listen or connect mode")
 
         /*  declare node input/output format  */
         if (this.params.mode === "rw") {
@@ -121,7 +121,7 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
                     else if (chunk.type !== self.params.type)
                         callback(new Error(`written chunk is not of ${self.params.type} type`))
                     else if (websockets.size === 0)
-                        callback(new Error("still no Websocket connections available"))
+                        callback(new Error("still no WebSocket connections available"))
                     else {
                         const data = util.streamChunkEncode(chunk)
                         const results: Promise<void>[] = []
@@ -168,10 +168,10 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
                 connectionTimeout:           4000,
                 minUptime:                   5000
             })
-            this.client.addEventListener("open", (ev) => {
+            this.client.addEventListener("open", (_ev) => {
                 this.log("info", `connection opened to URL ${this.params.connect}`)
             })
-            this.client.addEventListener("close", (ev) => {
+            this.client.addEventListener("close", (_ev) => {
                 this.log("info", `connection closed to URL ${this.params.connect}`)
             })
             this.client.addEventListener("error", (ev: ErrorEvent) => {
@@ -208,7 +208,7 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
                     else if (chunk.type !== self.params.type)
                         callback(new Error(`written chunk is not of ${self.params.type} type`))
                     else if (!self.client!.OPEN)
-                        callback(new Error("still no Websocket connection available"))
+                        callback(new Error("still no WebSocket connection available"))
                     else {
                         const data = util.streamChunkEncode(chunk)
                         self.client!.send(data)
@@ -234,7 +234,7 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
 
     /*  close node  */
     async close () {
-        /*  close Websocket server  */
+        /*  close WebSocket server  */
         if (this.server !== null) {
             await new Promise<void>((resolve, reject) => {
                 this.server!.close((error) => {
@@ -245,7 +245,7 @@ export default class SpeechFlowNodeXIOWebSocket extends SpeechFlowNode {
             this.server = null
         }
 
-        /*  close Websocket client  */
+        /*  close WebSocket client  */
         if (this.client !== null) {
             this.client.close()
             this.client = null

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

@@ -162,6 +162,9 @@ export function updateEnvelopeForChannel (
         else
             currentEnv = alphaR * currentEnv + (1 - alphaR) * det
     }
+
+    /*  store updated envelope value back  */
+    env[chan] = currentEnv
     return Math.sqrt(Math.max(currentEnv, 1e-12))
 }
 
@@ -173,6 +176,7 @@ export function dB2lin (db: number): number {
     return Math.pow(10, db / 20)
 }
 
+/*  Web Audio API wrapper class  */
 export class WebAudio {
     /*  internal state  */
     public audioContext: AudioContext
@@ -281,6 +285,7 @@ export class WebAudio {
         })
     }
 
+    /*  destroy object  */
     public async destroy (): Promise<void> {
         /*  reject all pending promises  */
         shield(() => {

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

@@ -27,7 +27,6 @@ export type LLMConfig = {
     timeout?:       number
     temperature?:   number
     maxTokens?:     number
-    topP?:          number
     cacheDir?:      string
 }
 export type LLMCompleteOptions = {
@@ -61,7 +60,6 @@ export class LLM extends EventEmitter {
             timeout:       30 * 1000,
             temperature:   0.7,
             maxTokens:     1024,
-            topP:          0.5,
             cacheDir:      "",
             ...config
         } as Required<LLMConfig>
@@ -102,7 +100,9 @@ export class LLM extends EventEmitter {
             /*  instantiate Anthropic API  */
             this.anthropic = new Anthropic({
                 ...(this.config.api !== "" ? { baseURL: this.config.api } : {}),
-                apiKey:  this.config.key,
+                ...(this.config.key.match(/^sk-ant-oat/) ?
+                    { authToken: this.config.key } :
+                    { apiKey:    this.config.key }  ),
                 timeout: this.config.timeout
             })
         }
@@ -229,7 +229,6 @@ export class LLM extends EventEmitter {
                 model:       this.config.model,
                 max_tokens:  this.config.maxTokens,
                 temperature: this.config.temperature,
-                top_p:       this.config.topP,
                 messages:    messages as OpenAI.ChatCompletionMessageParam[]
             }).catch((err) => {
                 throw new Error(`failed to perform OpenAI chat completion: ${err}`, { cause: err })
@@ -252,7 +251,6 @@ export class LLM extends EventEmitter {
                 model:       this.config.model,
                 max_tokens:  this.config.maxTokens,
                 temperature: this.config.temperature,
-                top_p:       this.config.topP,
                 system:      systemMessage?.content,
                 messages:    chatMessages as Anthropic.MessageParam[]
             }).catch((err) => {
@@ -283,7 +281,6 @@ export class LLM extends EventEmitter {
                 config: {
                     maxOutputTokens: this.config.maxTokens,
                     temperature:     this.config.temperature,
-                    topP:            this.config.topP,
                     ...(systemInstruction ? { systemInstruction } : {})
                 }
             }).catch((err) => {
@@ -305,8 +302,7 @@ export class LLM extends EventEmitter {
                 keep_alive: "10m",
                 options: {
                     num_predict: this.config.maxTokens,
-                    temperature: this.config.temperature,
-                    top_p:       this.config.topP
+                    temperature: this.config.temperature
                 }
             }).catch((err) => {
                 throw new Error(`failed to perform Ollama chat completion: ${err}`, { cause: err })
@@ -324,7 +320,6 @@ export class LLM extends EventEmitter {
             const result = await this.transformer(messages, {
                 max_new_tokens:  this.config.maxTokens,
                 temperature:     this.config.temperature,
-                top_p:           this.config.topP,
                 do_sample:       true
             }).catch((err) => {
                 throw new Error(`failed to perform HuggingFace Transformers text generation: ${err}`, { cause: err })

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

@@ -6,17 +6,17 @@
 
 /*  standard dependencies  */
 import { EventEmitter }       from "node:events"
-import { type, type Type }    from "arktype"
 
 /*  external dependencies  */
-import { Duration } from "luxon"
+import { type, type Type }    from "arktype"
+import { Duration }           from "luxon"
 import * as IntervalTree      from "node-interval-tree"
 
 /*  internal dependencies  */
 import * as util              from "./speechflow-util"
 
 /*  import an object with parsing and strict error handling  */
-export function importObject<T>(name: string, arg: object | string, validator: Type<T, {}>): T {
+export function importObject<T> (name: string, arg: object | string, validator: Type<T, {}>): T {
     const obj: object = typeof arg === "string" ?
         util.run(`${name}: parsing JSON`, () => JSON.parse(arg)) :
         arg
@@ -289,7 +289,7 @@ export class AsyncQueue<T> {
         if (this.queue.length > 0)
             return this.queue.shift()!
         else
-            return new Promise<T>((resolve, reject) => this.resolvers.push({ resolve, reject }))
+            return new Promise<T>((resolve, reject) => { this.resolvers.push({ resolve, reject }) })
     }
     empty () {
         return this.queue.length === 0

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

@@ -4,6 +4,7 @@
 **  Licensed under GPL 3.0 <https://spdx.org/licenses/GPL-3.0-only>
 */
 
+/*  re-export utility modules  */
 export * from "./speechflow-util-audio"
 export * from "./speechflow-util-error"
 export * from "./speechflow-util-stream"