@goodandready/dsh-tts 0.4.6 → 0.4.7

package/lib/text.js

@@ -16,11 +16,14 @@ export function blocksToText(content) {
 export const SPEECH_PHRASES = {
   en: { codeBlock: 'code block, {n} lines', table: 'table, {n} rows', summaryIntro: 'Summary of the reply.' },
   ru: { codeBlock: 'блок кода, {n} строк', table: 'таблица, {n} строк', summaryIntro: 'Пересказ ответа.' },
+  zh: { codeBlock: '代码块，共 {n} 行', table: '表格，共 {n} 行', summaryIntro: '回复摘要。' },
 }
 
 export function speechPhrases(language) {
   const lang = String(language || '').toLowerCase()
-  return lang.startsWith('ru') ? SPEECH_PHRASES.ru : SPEECH_PHRASES.en
+  if (lang.startsWith('zh')) return SPEECH_PHRASES.zh
+  if (lang.startsWith('ru')) return SPEECH_PHRASES.ru
+  return SPEECH_PHRASES.en
 }
 
 // Narration filters run on top of general scrubbing, before pronunciation.
@@ -43,8 +46,10 @@ export function applyNarrationFilters(text, cfg) {
 // Rough language heuristic: whichever script has more letters wins.
 export function detectLang(text) {
   const s = String(text || '')
+  const cjk = (s.match(/[\u4E00-\u9FFF]/g) || []).length
   const cyr = (s.match(/[\u0400-\u04FF]/g) || []).length
   const lat = (s.match(/[A-Za-z]/g) || []).length
+  if (cjk >= cyr && cjk >= lat && cjk > 0) return 'zh'
   return cyr > lat ? 'ru' : 'en'
 }
 
@@ -165,7 +170,14 @@ export function protectAbbreviations(text) {
     const rx = new RegExp('(?<=^|[\\s"«(])' + escaped + '(?=[\\s"»)\\]]|$)', 'gi')
     res = res.replace(rx, (m) => m.replace(/\./g, '\u2024'))
   }
+  // Decimal numbers and versions: e.g. 3.14, v0.4.6, 192.168.1.111
   res = res.replace(/(?<=[vV]?\d+)\.(?=\d+)/g, '\u2024')
+  // List numbering at line/word starts: 1. item, 2. item
+  res = res.replace(/(?<=^|[\s\n])(\d+)\.(?=\s+[A-Za-zА-Яа-я\u4E00-\u9FFF])/g, '$1\u2024')
+  // Common file extensions: e.g. package.json, index.js, script.py
+  res = res.replace(/(?<=[a-zA-Z0-9_-])\.(?:js|mjs|cjs|ts|tsx|jsx|json|ya?ml|md|py|sh|tgz|tar|gz|zip|html?|css|rs|go|c|cpp|h|txt)(?=[ \t\n"«'\)\]\},;\.\?!]|$)/gi, (m) => m.replace(/\./g, '\u2024'))
+  // Hostnames and domains: e.g. goodandready.app, github.com
+  res = res.replace(/(?<=[a-zA-Z0-9])\.(?:com|org|net|io|dev|app|ru|cn|local)(?=[ \t\n/:"«'\)\]\},;\.\?!]|$)/gi, (m) => m.replace(/\./g, '\u2024'))
   return res
 }
 
@@ -186,20 +198,29 @@ export function restoreAbbreviations(text) {
  */
 export function splitSentences(text, maxChars) {
   const limit = Number(maxChars) > 0 ? Number(maxChars) : 320
-  const min = Math.min(60, Math.floor(limit / 3))
   const raw = String(text || '').trim()
   if (!raw) return []
 
+  const isCjk = /[\u4E00-\u9FFF]/.test(raw)
+  // CJK characters carry much more meaning per character than Latin/Cyrillic words.
+  const min = isCjk ? Math.min(10, Math.floor(limit / 5)) : Math.min(60, Math.floor(limit / 3))
+
   const source = protectAbbreviations(raw)
   const pieces = []
   let current = ''
-  // Boundaries are period, question, exclamation, ellipsis, and newline.
-  for (const part of source.split(/(?<=[.!?…])\s+|\n+/)) {
+  // Boundaries are period, question, exclamation, ellipsis, Chinese full-width punctuation, and newline.
+  for (const part of source.split(/(?<=[.!?…\u3002\uFF01\uFF1F])\s*|\n+/)) {
     const chunk = String(part || '').trim()
     if (!chunk) continue
-    if (!current) { current = chunk } else if (current.length < min) { current += ' ' + chunk }
-    else { pieces.push(current); current = chunk }
-    // Oversized pieces are split on words: providers choke on walls of text.
+    if (!current) {
+      current = chunk
+    } else if (current.length < min) {
+      current += (isCjk ? '' : ' ') + chunk
+    } else {
+      pieces.push(current)
+      current = chunk
+    }
+    // Oversized pieces are split on words or characters: providers choke on walls of text.
     while (current.length > limit) {
       let cut = current.lastIndexOf(' ', limit)
       if (cut < min) cut = limit

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-tts",
-  "version": "0.4.6",
+  "version": "0.4.7",
   "description": "Text-to-speech for DeepSeek Harness: speak agent replies in the Web UI with a provider fallback chain (OpenAI, ElevenLabs, Google, Azure, Groq, Deepgram, OpenRouter, Edge, Piper, eSpeak).",
   "license": "MIT",
   "type": "module",