@goodandready/dsh-tts 0.3.10 → 0.3.12

package/lib/client.js

@@ -21,7 +21,7 @@ window.__ModuleLoader__.load({
       'edge', 'piper', 'espeak',
     ]
     const CLOUD = {
-      openai: 1, elevenlabs: 1, google: 1, azure: 1, groq: 1, deepgram: 1, openrouter: 1,
+      openai: 1, elevenlabs: 1, google: 1, azure: 1, groq: 1, deepgram: 1, openrouter: 1, custom: 1,
     }
     const MODEL_HINT = {
       openai: 'gpt-4o-mini-tts',
@@ -213,6 +213,13 @@ window.__ModuleLoader__.load({
       playerChanged()
     }
 
+    function skipCurrent() {
+      if (player.audio) { try { player.audio.pause() } catch (already) { /* стоит */ } }
+      player.audio = null
+      playerChanged()
+      if (player.queue.length) drainQueue()
+    }
+
     // Человек заговорил — замолкаем.
     //
     // Событие приходит из плагина голосового ввода, но никакой связи с ним
@@ -251,12 +258,55 @@ window.__ModuleLoader__.load({
           // опроса, иначе порядок фраз рассыплется.
           if (item.kind === 'reserved') break
           player.after = item.id
+          if (item.text) recent.add(item.text)
           player.queue.push(item)
         }
         if (player.queue.length) drainQueue()
       } catch (hostRestarting) { /* хост перезапускается */ }
     }
 
+    // Недавние озвучки: тексты последних реплик и избранное (localStorage).
+    function recentStore() {
+      const KEY = 'dsh-tts/recent'
+      let st
+      try {
+        const raw = typeof localStorage !== 'undefined' ? localStorage.getItem(KEY) : null
+        st = JSON.parse(raw) || { items: [], favs: [] }
+      } catch (broken) { st = { items: [], favs: [] } }
+      const persist = () => { try { if (typeof localStorage !== 'undefined') localStorage.setItem(KEY, JSON.stringify(st)) } catch (full) { /* приватный режим */ } }
+      return {
+        add(text) {
+          if (!text || text.length < 2) return
+          st.items = [text, ...st.items.filter((x) => x !== text)].slice(0, 30)
+          persist()
+        },
+        toggleFav(text) {
+          st.favs = st.favs.includes(text)
+            ? st.favs.filter((x) => x !== text)
+            : [text, ...st.favs].slice(0, 20)
+          persist()
+        },
+        list() { return { items: st.items, favs: st.favs } },
+      }
+    }
+    const recent = recentStore()
+
+    // Разовое проигрывание вне очереди: реплай из списка недавних.
+    function playStandalone(text) {
+      fetch('/dsh-tts/speak', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ text }),
+      })
+        .then((r) => r.json())
+        .then((d) => {
+          if (!d || !d.ok) return
+          const audio = new Audio('data:' + (d.mime || 'audio/mpeg') + ';base64,' + d.audioBase64)
+          audio.play().catch(() => {})
+        })
+        .catch(() => {})
+    }
+
     function credOf(map, provider) {
       return (map && map[provider]) || { configured: false, writable: true, ref: '' }
     }
@@ -302,6 +352,7 @@ window.__ModuleLoader__.load({
               }),
               React.createElement('button', { type: 'button', className: 'dts-mini', title: t('up'), disabled: !props.writable, onClick: () => move(i, -1) }, '\u2191'),
               React.createElement('button', { type: 'button', className: 'dts-mini', title: t('down'), disabled: !props.writable, onClick: () => move(i, 1) }, '\u2193'),
+              React.createElement('button', { type: 'button', className: 'dts-mini', title: t('preview'), disabled: !props.writable, onClick: () => props.onPreview(row.provider, row.model, row.voice) }, '\u25b6'),
               React.createElement('button', { type: 'button', className: 'dts-mini', title: t('remove'), disabled: !props.writable, onClick: () => remove(i) }, '\u00d7'),
             ),
             cloud ? React.createElement('div', { className: 'dts-row' },
@@ -403,6 +454,25 @@ window.__ModuleLoader__.load({
       'summaryModelHint': 'provider/model, empty = conversation default.',
       'summarySentences': 'Retelling length',
       'summarySentencesHint': 'Sentences in the spoken retelling.',
+      'preview': 'Preview',
+      'statsTitle': 'Statistics',
+      'statTotal': 'Total',
+      'statHits': 'Cache hits',
+      'statErrors': 'Errors',
+      'statReset': 'Reset stats',
+
+      'autoDetect': 'Auto language per piece',
+      'autoDetectHint': 'Guess ru/en per spoken piece instead of the language setting.',
+      'narrateQuotesOnly': 'Only quotes',
+      'narrateQuotesHint': 'Read only «quoted» fragments of a reply.',
+      'skipActions': 'Skip *actions*',
+      'skipActionsHint': 'Drop asterisk action blocks.',
+      'removeRegex': 'Remove regex',
+      'removeRegexHint': 'Global regex; matches are removed before synthesis.',
+      'customBaseUrl': 'Custom TTS endpoint',
+      'customBaseUrlHint': 'OpenAI-compatible origin, e.g. http://localhost:8880/v1; empty = off.',
+      'customKeyEnv': 'Custom key env',
+      'customKeyEnvHint': 'Credential name for the custom endpoint bearer key.',
     }
     const ru = {
       'general': 'Общее',
@@ -476,6 +546,25 @@ window.__ModuleLoader__.load({
       'summaryModelHint': 'provider/model, пусто — как у разговора.',
       'summarySentences': 'Длина пересказа',
       'summarySentencesHint': 'Предложений в озвучиваемом пересказе.',
+      'preview': 'Проба',
+      'statsTitle': 'Статистика',
+      'statTotal': 'Всего',
+      'statHits': 'Кэш-хиты',
+      'statErrors': 'Ошибки',
+      'statReset': 'Сбросить',
+
+      'autoDetect': 'Автоязык по куску',
+      'autoDetectHint': 'Подсказывать провайдеру ru/en по каждому куску вместо общей настройки.',
+      'narrateQuotesOnly': 'Только цитаты',
+      'narrateQuotesHint': 'Читать только фрагменты в кавычках.',
+      'skipActions': 'Пропускать *действия*',
+      'skipActionsHint': 'Вырезать блоки в звёздочках.',
+      'removeRegex': 'Регулярка вырезания',
+      'removeRegexHint': 'Глобальная регулярка: совпадения удаляются перед синтезом.',
+      'customBaseUrl': 'Свой TTS endpoint',
+      'customBaseUrlHint': 'OpenAI-совместимый адрес, например http://localhost:8880/v1; пусто = выключен.',
+      'customKeyEnv': 'Имя ключа для endpoint',
+      'customKeyEnvHint': 'Имя credential для Bearer-ключа своего endpointа.',
     }
 
     function TtsSection(props) {
@@ -575,6 +664,19 @@ window.__ModuleLoader__.load({
         clearCacheNow().catch((e) => setErr(String(e && e.message ? e.message : e)))
       }
 
+      const onPreview = (provider, model, voice) => {
+        fetch('/dsh-tts/preview', {
+          method: 'POST', headers: { 'Content-Type': 'application/json' },
+          body: JSON.stringify({ provider, model, voice }),
+        })
+          .then((r) => r.json())
+          .then((d) => {
+            if (!d.ok) throw new Error((d.error && d.error.message) || 'HTTP error')
+            new Audio('data:' + d.mime + ';base64,' + d.audioBase64).play().catch(() => {})
+          })
+          .catch((e) => setErr(String(e && e.message ? e.message : e)))
+      }
+
       const boolField = (key, label, hint) => React.createElement('label', { className: 'dts-field' }, label,
         React.createElement('input', {
           type: 'checkbox', checked: !!(draft && draft[key]), disabled: !writable,
@@ -625,11 +727,15 @@ window.__ModuleLoader__.load({
         React.createElement('div', { className: 'dts-block' },
           React.createElement('div', { className: 'dts-h' }, t('rolesTitle')),
           React.createElement(RolesEditor, {
-            t: t, writable: writable,
+            t: t, writable: writable, onPreview: onPreview,
             value: draft && draft.roles ? draft.roles : {},
             onChange: (v) => setTop('roles', v),
           }),
         ),
+        React.createElement('div', { className: 'dts-block' },
+          React.createElement('div', { className: 'dts-h' }, t('statsTitle')),
+          React.createElement(StatsPanel, { t: t }),
+        ),
         React.createElement('div', { className: 'dts-block' },
           React.createElement('div', { className: 'dts-h' }, t('chainTitle')),
           React.createElement('div', { className: 'dts-sub' },
@@ -642,6 +748,7 @@ window.__ModuleLoader__.load({
             onDraft: setDraftKey,
             onCommitKey: onCommitKey,
             onClearKey: clearKey,
+            onPreview: onPreview,
           }),
         ),
         React.createElement('div', { className: 'dts-block' },
@@ -666,6 +773,20 @@ window.__ModuleLoader__.load({
             React.createElement('span', { className: 'dts-sub' }, t('longReplyHint'))),
           textField('summaryModel', t('summaryModel'), t('summaryModelHint')),
           numberField('summarySentences', t('summarySentences'), t('summarySentencesHint'), 1),
+          React.createElement('label', { className: 'dts-field' }, t('longReply'),
+            React.createElement('select', {
+              className: 'dts-input', value: (draft && draft.longReply) || 'truncate', disabled: !writable,
+              onChange: (e) => setTop('longReply', e.target.value),
+            }, ['truncate', 'summarize', 'full'].map((mode) => React.createElement('option', { key: mode, value: mode }, mode))),
+            React.createElement('span', { className: 'dts-sub' }, t('longReplyHint'))),
+          textField('summaryModel', t('summaryModel'), t('summaryModelHint')),
+          numberField('summarySentences', t('summarySentences'), t('summarySentencesHint'), 1),
+          boolField('autoDetect', t('autoDetect'), t('autoDetectHint')),
+          boolField('narrateQuotesOnly', t('narrateQuotesOnly'), t('narrateQuotesHint')),
+          boolField('skipActions', t('skipActions'), t('skipActionsHint')),
+          textField('removeRegex', t('removeRegex'), t('removeRegexHint')),
+          textField('customBaseUrl', t('customBaseUrl'), t('customBaseUrlHint')),
+          textField('customKeyEnv', t('customKeyEnv'), t('customKeyEnvHint')),
           textField('approvalText', t('approvalText'), t('approvalHint')),
           textField('questionText', t('questionText'), t('questionHint')),
           textField('chime', t('chime'), t('chimeHint')),
@@ -760,10 +881,37 @@ window.__ModuleLoader__.load({
             value: (value[r] && value[r].chime) || '', disabled: !props.writable,
             onChange: (e) => set(r, { chime: e.target.value }),
           }),
+          React.createElement('input', {
+            className: 'dts-input dts-grow', placeholder: 'ssml',
+            value: (value[r] && value[r].ssmlStyle) || '', disabled: !props.writable,
+            onChange: (e) => set(r, { ssmlStyle: e.target.value }),
+          }),
+          React.createElement('button', { type: 'button', className: 'dts-mini', disabled: !props.writable, onClick: () => props.onPreview((value[r] && value[r].provider) || '', (value[r] && value[r].model) || '', (value[r] && value[r].voice) || '') }, '\u25b6'),
         )),
       )
     }
 
+    // Панель статистики синтеза: живые счётчики хоста.
+    function StatsPanel(props) {
+      const t = props.t || ((key) => key)
+      const [data, setData] = React.useState(null)
+      const loadStats = () => {
+        fetch('/dsh-tts/stats', { cache: 'no-store' }).then((r) => r.json()).then(setData).catch(() => {})
+      }
+      React.useEffect(() => { loadStats() }, [])
+      const reset = () => {
+        fetch('/dsh-tts/stats', { method: 'DELETE' }).then(loadStats).catch(() => {})
+      }
+      return React.createElement('div', { className: 'dts-block' },
+        React.createElement('span', { className: 'dts-sub' }, data
+          ? t('statTotal') + ': ' + data.total + ' · ' + t('statHits') + ': ' + data.cacheHits + ' · ' + t('statErrors') + ': ' + data.errors
+          : t('loading')),
+        data && Object.keys(data.providers || {}).map((p) => React.createElement('span', { className: 'dts-sub', key: p },
+          p + ': ' + data.providers[p].n + ' / ' + data.providers[p].ms + 'ms')),
+        React.createElement('button', { type: 'button', className: 'dts-link', onClick: reset }, t('statReset')),
+      )
+    }
+
     // Карточка во вкладке «Настройки плагинов» — разметка и стили по
     // единому канону (issue #19): li в списке ядра, шапка со сбросом темы,
     // тело с чертой-разделителем. Шеврон — наш, ядро его не даёт.
@@ -799,12 +947,41 @@ window.__ModuleLoader__.load({
       )
     }
 
-    // Кнопка в строке ввода: видно, что сейчас читается, и можно остановить.
+    // Кнопка в строке ввода: видно, что сейчас читается, можно остановить и
+    // открыть недавние озвучки для реплая.
     function SpeakerControl() {
       const p = usePlayer()
+      const [showList, setShowList] = React.useState(false)
       const active = !!p.audio || p.queue.length > 0 || p.busy
-      if (!active) return null
-      return React.createElement('div', { style: { display: 'inline-flex', gap: '4px', alignItems: 'center' } },
+      const [, force] = React.useReducer((n) => n + 1, 0)
+      if (!active && !showList) return null
+      const lists = recent.list()
+      const row = (text, starred) => React.createElement('div', { className: 'dts-row', key: text.slice(0, 24) + starred },
+        React.createElement('button', {
+          type: 'button', className: 'dts-link', style: { flex: 1, textAlign: 'left', overflow: 'hidden', textOverflow: 'ellipsis', whiteSpace: 'nowrap' },
+          title: text,
+          onClick: () => playStandalone(text),
+        }, text.slice(0, 80)),
+        React.createElement('button', {
+          type: 'button', className: 'dts-link', title: '★',
+          onClick: () => { recent.toggleFav(text); force() },
+        }, starred ? '★' : '☆'),
+      )
+      return React.createElement('div', { style: { position: 'relative', display: 'inline-flex', gap: '4px', alignItems: 'center' } },
+        React.createElement('button', {
+          type: 'button', className: 'dts-link', title: 'Недавние',
+          onClick: () => setShowList((v) => !v),
+        }, '☰'),
+        showList ? React.createElement('div', {
+          className: 'dts-wrap',
+          style: { position: 'absolute', bottom: '36px', right: 0, zIndex: 30, maxHeight: '320px', overflowY: 'auto', background: 'var(--dsw-alias-bg-layer-2)', border: '1px solid var(--dsw-alias-border-l2)', borderRadius: '12px', padding: '10px 12px', minWidth: '260px', flexDirection: 'column', gap: '6px', display: 'flex' },
+        },
+          lists.favs.length ? React.createElement('div', { className: 'dts-h' }, '★') : null,
+          lists.favs.map((text) => row(text, true)),
+          React.createElement('div', { className: 'dts-h' }, '⟳'),
+          lists.items.filter((x) => !lists.favs.includes(x)).map((text) => row(text, false)),
+          !lists.items.length ? React.createElement('span', { className: 'dts-sub' }, '—') : null,
+        ) : null,
         React.createElement('button', {
           type: 'button',
           className: 'dts-link',
@@ -887,6 +1064,16 @@ window.__ModuleLoader__.load({
       registerSettings(ctx)
       registerSpeaker(ctx)
       ctx.effect(() => listenForVoice(), 'dsh-tts: замолкать, когда человек заговорил')
+      ctx.effect(() => {
+        if (typeof window === 'undefined') return () => {}
+        const onKey = (e) => {
+          if (e.ctrlKey && e.key === 'Escape') { togglePause(); e.preventDefault() }
+          else if (e.altKey && (e.key === 's' || e.key === 'S' || e.key === 'ы')) { stopPlayback(); e.preventDefault() }
+          else if (e.altKey && e.key === 'ArrowRight') { skipCurrent(); e.preventDefault() }
+        }
+        window.addEventListener('keydown', onKey)
+        return () => window.removeEventListener('keydown', onKey)
+      }, 'dsh-tts: горячие клавиши')
       ctx.effect(() => {
         const timer = setInterval(pollPending, 1000)
         return () => clearInterval(timer)

package/lib/index.js

@@ -3,7 +3,7 @@ import { defineTool } from '@deepseek-ai/dsh-tools'
 import { credentialRef } from '@deepseek-ai/dsh-credentials'
 import { runChain } from './chain.js'
 import { makeProviders, PROVIDER_KEYS, DEFAULT_MODELS, DEFAULT_VOICES } from './providers.js'
-import { applyPronunciation, assistantText, speechPhrases, splitSentences, stripForSpeech } from './text.js'
+import { applyNarrationFilters, applyPronunciation, assistantText, detectLang, speechPhrases, splitSentences, stripForSpeech } from './text.js'
 import { cacheKey, createSpeechCache } from './cache.js'
 import fs from 'node:fs'
 import os from 'node:os'
@@ -37,6 +37,7 @@ const RoleOverride = z.object({
   model: z.string().default(''),
   voice: z.string().default(''),
   chime: z.string().default(''),
+  ssmlStyle: z.string().default(''),
 })
 
 export const Config = z.object({
@@ -48,6 +49,17 @@ export const Config = z.object({
     .description('Reuse synthesized audio for repeated phrases instead of paying and waiting again.'),
   cacheMaxMb: z.number().default(100)
     .description('Disk limit for the synthesis cache; least recently used items are evicted first.'),
+  narrateQuotesOnly: z.boolean().default(false)
+    .description('Speak only quoted fragments.'),
+  skipActions: z.boolean().default(false)
+    .description('Drop *asterisk action* blocks instead of reading them.'),
+  removeRegex: z.string().default('')
+    .description('Custom global regex whose matches are removed before synthesis.'),
+  autoDetect: z.boolean().default(false)
+    .description('Guess ru/en per spoken piece instead of using the language setting.'),
+  customBaseUrl: z.string().default('')
+    .description('Origin of an OpenAI-compatible /audio/speech endpoint; empty skips it.'),
+  customKeyEnv: z.string().default('CUSTOM_TTS_API_KEY'),
   pronunciation: z.array(z.object({
     from: z.string().default(''),
     to: z.string().default(''),
@@ -263,6 +275,9 @@ export function apply(ctx, config) {
 
   const llm = ctx.llm
 
+  // Счётчики синтеза: в памяти, обнуляются рестартом и DELETE /stats.
+  const stats = { total: 0, cacheHits: 0, errors: 0, providers: {} }
+
   // Кэш синтеза живёт рядом с прочими данными харнесса.
   const speechCache = createSpeechCache({
     root: process.env.DSH_HOME || path.join(os.homedir(), '.dsh'),
@@ -280,7 +295,7 @@ export function apply(ctx, config) {
     return { skipCode: cfg.skipCode !== false, phrases: speechPhrases(cfg.language) }
   }
   function cleanText(raw, cfg) {
-    return applyPronunciation(stripForSpeech(raw, cfg.maxChars, cleanOpts(cfg)), cfg.pronunciation, cfg.language)
+    return applyPronunciation(applyNarrationFilters(stripForSpeech(raw, cfg.maxChars, cleanOpts(cfg)), cfg), cfg.pronunciation, cfg.language)
   }
 
   // Пересказ быстрой моделью. Любая неудача — пустая строка, вызывающая
@@ -342,18 +357,28 @@ export function apply(ctx, config) {
     const cacheAllowed = cfg.cache !== false && text.length <= CACHE_MAX_TEXT
     if (cacheAllowed) {
       for (const provider of order) {
-        const hit = await speechCache.get(cacheKey([text, provider, models[provider], voices[provider], role]))
-        if (hit) return { provider, mime: hit.mime, audio: hit.audio, tookMs: 0, cached: true }
+        const hit = await speechCache.get(cacheKey([text, provider, models[provider], voices[provider], role, ov.ssmlStyle]))
+        if (hit) {
+          stats.total++
+          stats.cacheHits++
+          const ps = stats.providers[provider] || (stats.providers[provider] = { n: 0, e: 0, ms: 0 })
+          ps.n++
+          return { provider, mime: hit.mime, audio: hit.audio, tookMs: 0, cached: true }
+        }
       }
     }
     const providers = makeProviders(
       { resolveKey, fetchImpl: fetch, cfg },
-      { text, lang: cfg.language, signal, models, voices },
+      { text, lang: cfg.autoDetect ? detectLang(text) : cfg.language, signal, models, voices, role },
     )
     const out = await runChain(order, providers)
+    stats.total++
+    const ps = stats.providers[out.provider] || (stats.providers[out.provider] = { n: 0, e: 0, ms: 0 })
+    ps.n++
+    ps.ms += out.tookMs || 0
     if (cacheAllowed) {
       const p = out.provider
-      speechCache.put(cacheKey([text, p, models[p], voices[p], role]), out.mime, out.audio).catch(() => {})
+      speechCache.put(cacheKey([text, p, models[p], voices[p], role, ov.ssmlStyle]), out.mime, out.audio).catch(() => {})
     }
     return out
   }
@@ -373,12 +398,16 @@ export function apply(ctx, config) {
         id,
         sessionId: sid,
         kind: kind || 'speech',
+        text,
         provider: out.provider,
         mime: out.mime,
         audioBase64: Buffer.from(out.audio).toString('base64'),
         tookMs: out.tookMs,
       }))
-      .catch((err) => settle(id, { id, sessionId: sid, kind: kind || 'speech', text, error: String(err && err.message || err) }))
+      .catch((err) => {
+        stats.errors++
+        settle(id, { id, sessionId: sid, kind: 'error', text, error: String(err && err.message || err) })
+      })
       .finally(() => clearTimeout(timer))
   }
 
@@ -519,6 +548,65 @@ export function apply(ctx, config) {
     },
   }), 'dsh-tts: /cache route')
 
+  ctx.effect(() => ctx.webServer.register({
+    kind: 'exact',
+    path: '/dsh-tts/preview',
+    handler: async (req, res) => {
+      if (req.method !== 'POST') {
+        writeJson(res, 405, { ok: false, error: { code: 'method', message: 'POST only' } })
+        return
+      }
+      if (!isTrustedSettingsRequest(req)) {
+        writeJson(res, 403, { ok: false, error: { code: 'forbidden', message: 'same-origin only' } })
+        return
+      }
+      let body = {}
+      try {
+        const chunks = []
+        for await (const c of req) chunks.push(c)
+        body = JSON.parse(Buffer.concat(chunks).toString('utf8') || '{}')
+      } catch (e) {
+        writeJson(res, 400, { ok: false, error: { code: 'body', message: e.message } })
+        return
+      }
+      const cfg = live()
+      const phrase = String(cfg.language || '').toLowerCase().startsWith('ru') ? 'Проверка голоса.' : 'Voice check.'
+      try {
+        const probe = Object.assign({}, cfg, {
+          chain: [{ provider: body.provider || '', model: body.model || '', voice: body.voice || '' }],
+        })
+        const out = await synthesize(String(body.text || phrase), probe, null, 'reply')
+        writeJson(res, 200, { ok: true, mime: out.mime, audioBase64: Buffer.from(out.audio).toString('base64') })
+      } catch (e) {
+        writeJson(res, 502, { ok: false, error: { code: 'tts', message: String(e && e.message || e) } })
+      }
+    },
+  }), 'dsh-tts: /preview route')
+
+  ctx.effect(() => ctx.webServer.register({
+    kind: 'exact',
+    path: '/dsh-tts/stats',
+    handler: async (req, res) => {
+      if (req.method === 'GET') {
+        writeJson(res, 200, { ok: true, total: stats.total, cacheHits: stats.cacheHits, errors: stats.errors, providers: stats.providers })
+        return
+      }
+      if (req.method !== 'DELETE') {
+        writeJson(res, 405, { ok: false, error: { code: 'method', message: 'GET or DELETE' } })
+        return
+      }
+      if (!isTrustedSettingsRequest(req)) {
+        writeJson(res, 403, { ok: false, error: { code: 'forbidden', message: 'same-origin only' } })
+        return
+      }
+      stats.total = 0
+      stats.cacheHits = 0
+      stats.errors = 0
+      stats.providers = {}
+      writeJson(res, 200, { ok: true })
+    },
+  }), 'dsh-tts: /stats route')
+
   ctx.effect(() => ctx.webServer.register({
     kind: 'exact',
     path: '/dsh-tts/config',

package/lib/keys.js

@@ -1,5 +1,5 @@
 export const CLOUD_PROVIDERS = [
-  'openai', 'elevenlabs', 'google', 'azure', 'groq', 'deepgram', 'openrouter',
+  'openai', 'elevenlabs', 'google', 'azure', 'groq', 'deepgram', 'openrouter', 'custom',
 ]
 
 export const KEY_ENV_FIELD = {
@@ -10,6 +10,7 @@ export const KEY_ENV_FIELD = {
   groq: 'groqKeyEnv',
   deepgram: 'deepgramKeyEnv',
   openrouter: 'openrouterKeyEnv',
+  custom: 'customKeyEnv',
 }
 
 export const DEFAULT_KEY_ENV = {
@@ -20,6 +21,7 @@ export const DEFAULT_KEY_ENV = {
   groq: 'GROQ_API_KEY',
   deepgram: 'DEEPGRAM_API_KEY',
   openrouter: 'OPENROUTER_API_KEY',
+  custom: 'CUSTOM_TTS_API_KEY',
 }
 
 const REF_PATTERN = /^[A-Za-z_][A-Za-z0-9_]*$/

package/lib/providers.js

@@ -2,7 +2,7 @@ import { edgeSpeak, espeakSpeak, minimaxSpeak, piperSpeak } from './local.js'
 
 export const PROVIDER_KEYS = [
   'openai', 'elevenlabs', 'google', 'azure', 'groq', 'deepgram', 'openrouter',
-  'siliconflow', 'deepinfra', 'fireworks', 'mimo',
+  'siliconflow', 'deepinfra', 'fireworks', 'mimo', 'custom',
   'edge', 'piper', 'espeak', 'minimax',
 ]
 
@@ -13,6 +13,8 @@ export const OPENAI_COMPATIBLE = {
   siliconflow: { baseUrl: 'https://api.siliconflow.cn/v1', keyEnv: 'SILICONFLOW_API_KEY' },
   deepinfra: { baseUrl: 'https://api.deepinfra.com/v1/openai', keyEnv: 'DEEPINFRA_API_KEY' },
   fireworks: { baseUrl: 'https://api.fireworks.ai/inference/v1', keyEnv: 'FIREWORKS_API_KEY' },
+  // Пользовательский endpoint: адрес и имя ключа приходят из настроек.
+  custom: { baseUrl: '', keyEnv: 'CUSTOM_TTS_API_KEY' },
 }
 
 export const DEFAULT_MODELS = {
@@ -28,6 +30,7 @@ export const DEFAULT_MODELS = {
   groq: 'playai-tts',
   deepgram: 'aura-asteria-en',
   openrouter: 'openai/gpt-4o-mini-tts-2025-12-15',
+  custom: '',
   edge: 'ru-RU-SvetlanaNeural',
   piper: '',
   espeak: 'ru',
@@ -47,6 +50,7 @@ export const DEFAULT_VOICES = {
   groq: 'Fritz-PlayAI',
   deepgram: '',
   openrouter: 'alloy',
+  custom: '',
   edge: 'ru-RU-SvetlanaNeural',
   piper: '',
   espeak: 'ru',
@@ -94,6 +98,7 @@ export function makeProviders(deps, req) {
       const token = await resolveKey(keyEnv)
       if (!token) return { ok: false, provider: key, reason: `no ${keyEnv}` }
       const base = cfg[key + 'BaseUrl'] || spec.baseUrl
+      if (!base) return { ok: false, provider: key, reason: 'no base url configured' }
       const res = await fetchImpl(base + '/audio/speech', {
         method: 'POST',
         headers: { authorization: `Bearer ${token}`, 'content-type': 'application/json' },
@@ -217,9 +222,15 @@ export function makeProviders(deps, req) {
     const region = (cfg.azureRegion || '').trim()
     if (!region) return { ok: false, provider: 'azure', reason: 'azureRegion is empty' }
     const voice = pick(DEFAULT_VOICES, voices, 'azure')
+    const roleCfg = req.role && cfg.roles && cfg.roles[req.role]
+    const style = roleCfg && roleCfg.ssmlStyle
+    const safeStyle = style ? String(style).replace(/["<>&]/g, '') : ''
+    const inner = safeStyle
+      ? `<mstts:express-as style="${safeStyle}">${escapeXml(text)}</mstts:express-as>`
+      : escapeXml(text)
     const ssml = `<speak version="1.0" xml:lang="${lang || 'en-US'}">`
       + `<voice xml:lang="${lang || 'en-US'}" name="${voice}">`
-      + `${escapeXml(text)}</voice></speak>`
+      + `${inner}</voice></speak>`
     const res = await fetchImpl(`https://${region}.tts.speech.microsoft.com/cognitiveservices/v1`, {
       method: 'POST',
       headers: {

package/lib/text.js

@@ -23,6 +23,31 @@ export function speechPhrases(language) {
   return lang.startsWith('ru') ? SPEECH_PHRASES.ru : SPEECH_PHRASES.en
 }
 
+// Фильтры наррации: вырезания поверх общей чистки, до словаря произношения.
+// Порядок: своя регулярка -> блоки *действий* -> режим «только цитаты».
+export function applyNarrationFilters(text, cfg) {
+  let out = String(text || '')
+  if (cfg && cfg.removeRegex) {
+    try {
+      out = out.replace(new RegExp(cfg.removeRegex, 'g'), ' ')
+    } catch (badUserRegex) { /* кривая регулярка не роняет озвучку */ }
+  }
+  if (cfg && cfg.skipActions) out = out.replace(/(^|\s)\*[^*\n]+\*/g, '$1')
+  if (cfg && cfg.narrateQuotesOnly) {
+    const quotes = [...out.matchAll(/«[^»]*»|"[^"]*"/g)].map((m) => m[0]).join(' ')
+    if (quotes) out = quotes
+  }
+  return out
+}
+
+// Грубая эвристика языка куска: где букв больше, тот язык и подсказываем.
+export function detectLang(text) {
+  const s = String(text || '')
+  const cyr = (s.match(/[\u0400-\u04FF]/g) || []).length
+  const lat = (s.match(/[A-Za-z]/g) || []).length
+  return cyr > lat ? 'ru' : 'en'
+}
+
 // Словарь произношения: правила применяются сверху вниз, каждое по одному
 // разу ко всему тексту. Левая часть — обычный текст; запись вида /…/флаги
 // трактуется как регулярное выражение для тех, кто умеет. whole=true требует

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-tts",
-  "version": "0.3.10",
+  "version": "0.3.12",
   "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",