UNPKG

speechflow

Version:

Speech Processing Flow Graph

250 lines 10.7 kB
"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_stream_1 = __importDefault(require("node:stream")); /* external dependencies */ const luxon_1 = require("luxon"); /* internal dependencies */ const speechflow_node_1 = __importDefault(require("./speechflow-node")); const util = __importStar(require("./speechflow-util")); /* SpeechFlow node for sentence splitting */ class SpeechFlowNodeT2TSentence extends speechflow_node_1.default { /* declare official node name */ static name = "t2t-sentence"; /* internal state */ queue = new util.Queue(); queueRecv = this.queue.pointerUse("recv"); queueSplit = this.queue.pointerUse("split"); queueSend = this.queue.pointerUse("send"); destroyed = false; workingOffTimer = 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 = "text"; this.output = "text"; } /* open node */ async open() { /* clear destruction flag */ this.destroyed = false; /* work off queued text frames */ let workingOff = false; const workOffQueue = async () => { if (this.destroyed) return; /* control working off round */ if (workingOff) return; workingOff = true; if (this.workingOffTimer !== null) { clearTimeout(this.workingOffTimer); this.workingOffTimer = null; } this.queue.off("write", workOffQueue); /* try to work off one or more chunks */ while (!this.destroyed) { const element = this.queueSplit.peek(); if (element === undefined) break; if (element.type === "text-eof") { this.queueSplit.walk(+1); break; } const chunk = element.chunk; const payload = chunk.payload; const m = payload.match(/^((?:.|\r?\n)+?[.;?!])\s*((?:.|\r?\n)*)$/); if (m !== null) { /* contains a sentence */ const [, sentence, rest] = m; if (rest !== "") { /* contains more than a sentence */ const chunk2 = chunk.clone(); const duration = luxon_1.Duration.fromMillis(chunk.timestampEnd.minus(chunk.timestampStart).toMillis() * (sentence.length / payload.length)); chunk2.timestampStart = chunk.timestampStart.plus(duration); chunk.timestampEnd = chunk2.timestampStart; chunk.payload = sentence; chunk2.payload = rest; element.complete = true; this.queueSplit.touch(); this.queueSplit.walk(+1); this.queueSplit.insert({ type: "text-frame", chunk: chunk2 }); } else { /* contains just the sentence */ element.complete = true; this.queueSplit.touch(); this.queueSplit.walk(+1); } } else { /* contains less than a sentence */ const position = this.queueSplit.position(); if (position < this.queueSplit.maxPosition() - 1) { /* merge into following chunk */ const element2 = this.queueSplit.peek(position + 1); if (element2 === undefined) break; if (element2.type === "text-eof") { element.complete = true; this.queueSplit.touch(); this.queueSplit.walk(+1); break; } element2.chunk.timestampStart = element.chunk.timestampStart; element2.chunk.payload = element.chunk.payload + " " + element2.chunk.payload; this.queueSplit.delete(); this.queueSplit.touch(); } else break; } } /* re-initiate working off round (if still not destroyed) */ workingOff = false; if (!this.destroyed) { this.workingOffTimer = setTimeout(workOffQueue, 100); this.queue.once("write", workOffQueue); } }; this.queue.once("write", workOffQueue); /* provide Duplex stream and internally attach to classifier */ const self = this; this.stream = new node_stream_1.default.Duplex({ writableObjectMode: true, readableObjectMode: true, decodeStrings: false, highWaterMark: 1, /* receive text chunk (writable side of stream) */ write(chunk, encoding, callback) { if (self.destroyed) callback(new Error("stream already destroyed")); else if (Buffer.isBuffer(chunk.payload)) callback(new Error("expected text input as string chunks")); else if (chunk.payload.length === 0) callback(); else { self.log("info", `received text: ${JSON.stringify(chunk.payload)}`); self.queueRecv.append({ type: "text-frame", chunk }); callback(); } }, /* receive no more text chunks (writable side of stream) */ final(callback) { if (self.destroyed) { callback(); return; } /* signal end of file */ self.queueRecv.append({ type: "text-eof" }); callback(); }, /* send text chunk(s) (readable side of stream) */ read(_size) { /* flush pending text chunks */ const flushPendingChunks = () => { if (self.destroyed) { this.push(null); return; } const element = self.queueSend.peek(); if (element !== undefined && element.type === "text-eof") { this.push(null); self.queueSend.walk(+1); } else if (element !== undefined && element.type === "text-frame" && element.complete === true) { while (true) { const nextElement = self.queueSend.peek(); if (nextElement === undefined) break; else if (nextElement.type === "text-eof") { this.push(null); self.queueSend.walk(+1); break; } else if (nextElement.type === "text-frame" && nextElement.complete !== true) break; self.log("info", `send text: ${JSON.stringify(nextElement.chunk.payload)}`); this.push(nextElement.chunk); self.queueSend.walk(+1); self.queue.trim(); } } else if (!self.destroyed) self.queue.once("write", flushPendingChunks); }; flushPendingChunks(); } }); } /* close node */ async close() { /* indicate destruction */ this.destroyed = true; /* clean up timer */ if (this.workingOffTimer !== null) { clearTimeout(this.workingOffTimer); this.workingOffTimer = null; } /* remove any pending event listeners */ this.queue.removeAllListeners("write"); /* close stream */ if (this.stream !== null) { this.stream.destroy(); this.stream = null; } } } exports.default = SpeechFlowNodeT2TSentence; //# sourceMappingURL=speechflow-node-t2t-sentence.js.map