@goodandready/dsh-tts 0.4.2 → 0.4.5

package/lib/client-src/10-styles.js

@@ -41,7 +41,7 @@
       '.dts-alert{padding:10px 14px;border-radius:8px;font-size:13px}' +
       '.dts-alert-err{background:color-mix(in srgb, var(--dsw-alias-state-error-primary) 10%, transparent);color:var(--dsw-alias-state-error-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-error-primary) 20%, transparent)}' +
       '.dts-alert-warn{background:color-mix(in srgb, var(--dsw-alias-state-warning-primary) 10%, transparent);color:var(--dsw-alias-state-warning-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-warning-primary) 20%, transparent)}' +
-      '.dts-alert-ok{background:color-mix(in srgb, var(--dsw-alias-state-success-primary) 10%, transparent);color:var(--dsw-alias-state-success-primary);border:1px solid rgba(16,185,129,0.2)}' +
+      '.dts-alert-ok{background:color-mix(in srgb, var(--dsw-alias-state-success-primary) 10%, transparent);color:var(--dsw-alias-state-success-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-success-primary) 20%, transparent)}' +
       '.dts-grow{flex:1;min-width:100px}'
 
     const setCssId = 'dsh-tts/settings.module.css'

package/lib/client-src/20-player.js

@@ -18,7 +18,7 @@
         const AC = window.AudioContext || window.webkitAudioContext
         if (AC) {
           const ctx = new AC()
-          if (ctx.state === 'suspended') ctx.resume().catch(() => {})
+          if (ctx.state === 'suspended') ctx.resume().catch(() => { /* audio context unlock failure ignored */ })
           setTimeout(() => { try { ctx.close() } catch { /* closed */ } }, 300)
         }
       } catch { /* no AudioContext */ }
@@ -317,9 +317,9 @@
         .then((d) => {
           if (!d || !d.ok) return
           const audio = new Audio('data:' + (d.mime || 'audio/mpeg') + ';base64,' + d.audioBase64)
-          audio.play().catch(() => {})
+          audio.play().catch(() => { /* chime autoplay blocked or interrupted */ })
         })
-        .catch(() => {})
+        .catch(() => { /* chime fetch failed or not found */ })
     }
 
     function credOf(map, provider) {

package/lib/client-src/40-editors.js

@@ -39,7 +39,7 @@
               }),
               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('testProvider') || 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' },
@@ -60,7 +60,7 @@
             (() => {
               const hit = Array.isArray(props.breaker) ? props.breaker.find((b) => b && b.id === row.provider) : null
               if (!hit || (!hit.open && !hit.lastError)) return null
-              return React.createElement('div', { className: 'dts-sub' },
+              return React.createElement('div', { className: 'dts-sub dts-bad' },
                 (hit.open ? t('breakerOpen') + ' · ' : '') + (hit.lastError || '')
               )
             })(),

package/lib/client-src/50-settings.js

@@ -9,7 +9,7 @@
       const [integrations, setIntegrations] = React.useState({ voice: { installed: false }, messenger: { installed: false } })
       const [telemetry, setTelemetry] = React.useState({ online: true, cacheHits: 0, sseOk: true, total: 0, sseClients: 0, breaker: [], provider: '', queue: 0 })
       const [advancedOpen, setAdvancedOpen] = React.useState(() => {
-        try { return localStorage.getItem('dsh-tts/advancedOpen') === '1' } catch { return false }
+        try { return localStorage.getItem('dsh-tts/advancedOpen') === '1' } catch { /* localStorage unavailable */ return false }
       })
       const autoplay = useAutoplayGate()
 
@@ -19,7 +19,7 @@
         if (!s || !s.bind) return null
         try {
           return s.bind({ namespace: NS })
-        } catch (_) {
+        } catch (_) { /* settings scope binding failed */ 
           return null
         }
       }, [ctx])
@@ -29,7 +29,7 @@
           if (!scope || !scope.subscribe) return () => {}
           try {
             return scope.subscribe(cb) || (() => {})
-          } catch (_) {
+          } catch (_) { /* settings scope binding failed */ 
             return () => {}
           }
         }
@@ -39,7 +39,7 @@
         if (!scope || !scope.getSnapshot) return SNAPSHOT_LOADING
         try {
           return scope.getSnapshot() || SNAPSHOT_LOADING
-        } catch (_) {
+        } catch (_) { /* settings scope binding failed */ 
           return SNAPSHOT_LOADING
         }
       }, [scope])
@@ -119,7 +119,7 @@
           headers: { 'Content-Type': 'application/json' },
           body: JSON.stringify({ provider: provider, value: value }),
         })
-        const data = await res.json().catch(() => ({}))
+        const data = await res.json().catch(() => ({ /* parse error */ }))
         if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
         if (data && data.credentials) setCredentials(data.credentials)
         setKeyDrafts((d) => Object.assign({}, d, { [provider]: '' }))
@@ -133,7 +133,7 @@
             headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify({ provider: provider }),
           })
-          const data = await res.json().catch(() => ({}))
+          const data = await res.json().catch(() => ({ /* parse error */ }))
           if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
           if (data && data.credentials) setCredentials(data.credentials)
           setKeyDrafts((d) => Object.assign({}, d, { [provider]: '' }))
@@ -154,7 +154,7 @@
             headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify(Object.assign({}, draft, { keys: keys })),
           })
-          const data = await res.json().catch(() => ({}))
+          const data = await res.json().catch(() => ({ /* parse error */ }))
           if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
           applyPayload(data)
           setKeyDrafts({})
@@ -187,7 +187,7 @@
           .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(() => {})
+            new Audio('data:' + d.mime + ';base64,' + d.audioBase64).play().catch(() => { /* autoplay blocked */ })
           })
           .catch((e) => setErr(String(e && e.message ? e.message : e)))
       }

package/lib/client.js

@@ -126,7 +126,7 @@ window.__ModuleLoader__.load({
       '.dts-alert{padding:10px 14px;border-radius:8px;font-size:13px}' +
       '.dts-alert-err{background:color-mix(in srgb, var(--dsw-alias-state-error-primary) 10%, transparent);color:var(--dsw-alias-state-error-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-error-primary) 20%, transparent)}' +
       '.dts-alert-warn{background:color-mix(in srgb, var(--dsw-alias-state-warning-primary) 10%, transparent);color:var(--dsw-alias-state-warning-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-warning-primary) 20%, transparent)}' +
-      '.dts-alert-ok{background:color-mix(in srgb, var(--dsw-alias-state-success-primary) 10%, transparent);color:var(--dsw-alias-state-success-primary);border:1px solid rgba(16,185,129,0.2)}' +
+      '.dts-alert-ok{background:color-mix(in srgb, var(--dsw-alias-state-success-primary) 10%, transparent);color:var(--dsw-alias-state-success-primary);border:1px solid color-mix(in srgb, var(--dsw-alias-state-success-primary) 20%, transparent)}' +
       '.dts-grow{flex:1;min-width:100px}'
 
     const setCssId = 'dsh-tts/settings.module.css'
@@ -157,7 +157,7 @@ window.__ModuleLoader__.load({
         const AC = window.AudioContext || window.webkitAudioContext
         if (AC) {
           const ctx = new AC()
-          if (ctx.state === 'suspended') ctx.resume().catch(() => {})
+          if (ctx.state === 'suspended') ctx.resume().catch(() => { /* audio context unlock failure ignored */ })
           setTimeout(() => { try { ctx.close() } catch { /* closed */ } }, 300)
         }
       } catch { /* no AudioContext */ }
@@ -456,9 +456,9 @@ window.__ModuleLoader__.load({
         .then((d) => {
           if (!d || !d.ok) return
           const audio = new Audio('data:' + (d.mime || 'audio/mpeg') + ';base64,' + d.audioBase64)
-          audio.play().catch(() => {})
+          audio.play().catch(() => { /* chime autoplay blocked or interrupted */ })
         })
-        .catch(() => {})
+        .catch(() => { /* chime fetch failed or not found */ })
     }
 
     function credOf(map, provider) {
@@ -867,7 +867,7 @@ 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('testProvider') || 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' },
@@ -888,7 +888,7 @@ window.__ModuleLoader__.load({
             (() => {
               const hit = Array.isArray(props.breaker) ? props.breaker.find((b) => b && b.id === row.provider) : null
               if (!hit || (!hit.open && !hit.lastError)) return null
-              return React.createElement('div', { className: 'dts-sub' },
+              return React.createElement('div', { className: 'dts-sub dts-bad' },
                 (hit.open ? t('breakerOpen') + ' · ' : '') + (hit.lastError || '')
               )
             })(),
@@ -1227,7 +1227,7 @@ window.__ModuleLoader__.load({
       const [integrations, setIntegrations] = React.useState({ voice: { installed: false }, messenger: { installed: false } })
       const [telemetry, setTelemetry] = React.useState({ online: true, cacheHits: 0, sseOk: true, total: 0, sseClients: 0, breaker: [], provider: '', queue: 0 })
       const [advancedOpen, setAdvancedOpen] = React.useState(() => {
-        try { return localStorage.getItem('dsh-tts/advancedOpen') === '1' } catch { return false }
+        try { return localStorage.getItem('dsh-tts/advancedOpen') === '1' } catch { /* localStorage unavailable */ return false }
       })
       const autoplay = useAutoplayGate()
 
@@ -1237,7 +1237,7 @@ window.__ModuleLoader__.load({
         if (!s || !s.bind) return null
         try {
           return s.bind({ namespace: NS })
-        } catch (_) {
+        } catch (_) { /* settings scope binding failed */ 
           return null
         }
       }, [ctx])
@@ -1247,7 +1247,7 @@ window.__ModuleLoader__.load({
           if (!scope || !scope.subscribe) return () => {}
           try {
             return scope.subscribe(cb) || (() => {})
-          } catch (_) {
+          } catch (_) { /* settings scope binding failed */ 
             return () => {}
           }
         }
@@ -1257,7 +1257,7 @@ window.__ModuleLoader__.load({
         if (!scope || !scope.getSnapshot) return SNAPSHOT_LOADING
         try {
           return scope.getSnapshot() || SNAPSHOT_LOADING
-        } catch (_) {
+        } catch (_) { /* settings scope binding failed */ 
           return SNAPSHOT_LOADING
         }
       }, [scope])
@@ -1337,7 +1337,7 @@ window.__ModuleLoader__.load({
           headers: { 'Content-Type': 'application/json' },
           body: JSON.stringify({ provider: provider, value: value }),
         })
-        const data = await res.json().catch(() => ({}))
+        const data = await res.json().catch(() => ({ /* parse error */ }))
         if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
         if (data && data.credentials) setCredentials(data.credentials)
         setKeyDrafts((d) => Object.assign({}, d, { [provider]: '' }))
@@ -1351,7 +1351,7 @@ window.__ModuleLoader__.load({
             headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify({ provider: provider }),
           })
-          const data = await res.json().catch(() => ({}))
+          const data = await res.json().catch(() => ({ /* parse error */ }))
           if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
           if (data && data.credentials) setCredentials(data.credentials)
           setKeyDrafts((d) => Object.assign({}, d, { [provider]: '' }))
@@ -1372,7 +1372,7 @@ window.__ModuleLoader__.load({
             headers: { 'Content-Type': 'application/json' },
             body: JSON.stringify(Object.assign({}, draft, { keys: keys })),
           })
-          const data = await res.json().catch(() => ({}))
+          const data = await res.json().catch(() => ({ /* parse error */ }))
           if (!res.ok) throw new Error((data && data.error && data.error.message) || ('HTTP ' + res.status))
           applyPayload(data)
           setKeyDrafts({})
@@ -1405,7 +1405,7 @@ window.__ModuleLoader__.load({
           .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(() => {})
+            new Audio('data:' + d.mime + ';base64,' + d.audioBase64).play().catch(() => { /* autoplay blocked */ })
           })
           .catch((e) => setErr(String(e && e.message ? e.message : e)))
       }

package/lib/providers/cloud.js

@@ -0,0 +1,278 @@
+import { DEFAULT_MODELS, DEFAULT_VOICES, OPENAI_COMPATIBLE } from "./constants.js"
+
+function pick(map, models, key) {
+  const chosen = models && typeof models[key] === "string" ? models[key].trim() : ""
+  return chosen || map[key]
+}
+
+function asBuffer(body) {
+  if (Buffer.isBuffer(body)) return body
+  return Buffer.from(body)
+}
+
+function escapeXml(s) {
+  return String(s)
+    .replace(/&/g, '&')
+    .replace(/</g, '&lt;')
+    .replace(/>/g, '&gt;')
+    .replace(/"/g, '&quot;')
+}
+
+export function makeCloudProviders(deps, req, withCloudTimeout) {
+  const { resolveKey, fetchImpl, cfg } = deps
+  const { text, lang, models, voices, role } = req
+
+  async function openai() {
+    const key = await resolveKey(cfg.openaiKeyEnv)
+    if (!key) return { ok: false, provider: 'openai', reason: `no ${cfg.openaiKeyEnv}` }
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl((cfg.openaiBaseUrl || 'https://api.openai.com/v1') + '/audio/speech', {
+      method: 'POST',
+      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
+      body: JSON.stringify({
+        model: pick(DEFAULT_MODELS, models, 'openai'),
+        input: text,
+        voice: pick(DEFAULT_VOICES, voices, 'openai'),
+        response_format: 'mp3',
+      }),
+      signal,
+    })
+    if (!res.ok) throw new Error(`OpenAI HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'openai', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  function openaiCompatible(key) {
+    const spec = OPENAI_COMPATIBLE[key]
+    return async function speak() {
+      const keyEnv = (cfg[key + 'KeyEnv'] || spec.keyEnv)
+      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 { signal, cleanup } = withCloudTimeout()
+      try {
+        const res = await fetchImpl(base + '/audio/speech', {
+        method: 'POST',
+        headers: { authorization: `Bearer ${token}`, 'content-type': 'application/json' },
+        body: JSON.stringify({
+          model: pick(DEFAULT_MODELS, models, key),
+          input: text,
+          voice: pick(DEFAULT_VOICES, voices, key),
+          response_format: 'mp3',
+        }),
+        signal,
+      })
+      if (!res.ok) throw new Error(`${key} HTTP ${res.status}`)
+      const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+      return { ok: audio.length > 0, provider: key, audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+      } finally {
+        cleanup()
+      }
+    }
+  }
+  async function mimo() {
+    const key = await resolveKey(cfg.mimoKeyEnv)
+    if (!key) return { ok: false, provider: 'mimo', reason: `no ${cfg.mimoKeyEnv}` }
+    const base = (cfg.mimoBaseUrl || 'https://api.xiaomimimo.com/v1').replace(/\/+$/, '')
+    const format = cfg.mimoFormat === 'wav' ? 'wav' : 'mp3'
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl(base + '/chat/completions', {
+      method: 'POST',
+      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json', accept: 'application/json' },
+      body: JSON.stringify({
+        model: pick(DEFAULT_MODELS, models, 'mimo'),
+        messages: [{ role: 'assistant', content: text }],
+        audio: { format, voice: pick(DEFAULT_VOICES, voices, 'mimo') },
+        stream: false,
+      }),
+      signal,
+    })
+    const data = await res.json().catch(() => ({}))
+    if (!res.ok) {
+      const detail = data && data.error && data.error.message
+      throw new Error(`MiMo HTTP ${res.status}${detail ? ': ' + detail : ''}`)
+    }
+    const encoded = data && data.choices && data.choices[0]
+      && data.choices[0].message && data.choices[0].message.audio
+      && data.choices[0].message.audio.data
+    if (typeof encoded !== 'string' || !encoded) {
+      return { ok: false, provider: 'mimo', reason: 'response missing choices[0].message.audio.data' }
+    }
+    const audio = Buffer.from(encoded, 'base64')
+    return {
+      ok: audio.length > 0,
+      provider: 'mimo',
+      audio,
+      mime: format === 'wav' ? 'audio/wav' : 'audio/mpeg',
+      reason: audio.length ? '' : 'empty audio',
+    }
+    } finally {
+      cleanup()
+    }
+  }
+  async function elevenlabs() {
+    const key = await resolveKey(cfg.elevenlabsKeyEnv)
+    if (!key) return { ok: false, provider: 'elevenlabs', reason: `no ${cfg.elevenlabsKeyEnv}` }
+    const voice = pick(DEFAULT_VOICES, voices, 'elevenlabs')
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl(`https://api.elevenlabs.io/v1/text-to-speech/${encodeURIComponent(voice)}`, {
+      method: 'POST',
+      headers: { 'xi-api-key': key, 'content-type': 'application/json', accept: 'audio/mpeg' },
+      body: JSON.stringify({
+        text,
+        model_id: pick(DEFAULT_MODELS, models, 'elevenlabs'),
+      }),
+      signal,
+    })
+    if (!res.ok) throw new Error(`ElevenLabs HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'elevenlabs', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  async function google() {
+    const key = await resolveKey(cfg.googleKeyEnv)
+    if (!key) return { ok: false, provider: 'google', reason: `no ${cfg.googleKeyEnv}` }
+    const model = pick(DEFAULT_MODELS, models, 'google')
+    const voiceName = pick(DEFAULT_VOICES, voices, 'google')
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl(
+      `https://generativelanguage.googleapis.com/v1beta/models/${encodeURIComponent(model)}:generateContent?key=${encodeURIComponent(key)}`,
+      {
+        method: 'POST',
+        headers: { 'content-type': 'application/json' },
+        body: JSON.stringify({
+          contents: [{ parts: [{ text }] }],
+          generationConfig: {
+            responseModalities: ['AUDIO'],
+            speechConfig: { voiceConfig: { prebuiltVoiceConfig: { voiceName } } },
+          },
+        }),
+        signal,
+      },
+    )
+    if (!res.ok) throw new Error(`Google HTTP ${res.status}`)
+    const data = await res.json()
+    const b64 = data?.candidates?.[0]?.content?.parts?.find((p) => p.inlineData)?.inlineData?.data
+    if (!b64) return { ok: false, provider: 'google', reason: 'no inline audio' }
+    const audio = Buffer.from(b64, 'base64')
+    return { ok: audio.length > 0, provider: 'google', audio, mime: 'audio/wav', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  async function azure() {
+    const key = await resolveKey(cfg.azureKeyEnv)
+    if (!key) return { ok: false, provider: 'azure', reason: `no ${cfg.azureKeyEnv}` }
+    const region = (cfg.azureRegion || '').trim()
+    if (!region) return { ok: false, provider: 'azure', reason: 'azureRegion is empty' }
+    const voice = pick(DEFAULT_VOICES, voices, 'azure')
+    const roleCfg = req.role && cfg.roles && cfg.roles[req.role]
+    const style = roleCfg && roleCfg.ssmlStyle
+    const safeStyle = style ? String(style).replace(/["<>&]/g, '') : ''
+    const inner = safeStyle
+      ? `<mstts:express-as style="${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}">`
+      + `${inner}</voice></speak>`
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl(`https://${region}.tts.speech.microsoft.com/cognitiveservices/v1`, {
+      method: 'POST',
+      headers: {
+        'Ocp-Apim-Subscription-Key': key,
+        'Content-Type': 'application/ssml+xml',
+        'X-Microsoft-OutputFormat': 'audio-16khz-128kbitrate-mono-mp3',
+      },
+      body: ssml,
+      signal,
+    })
+    if (!res.ok) throw new Error(`Azure HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'azure', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  async function groq() {
+    const key = await resolveKey(cfg.groqKeyEnv)
+    if (!key) return { ok: false, provider: 'groq', reason: `no ${cfg.groqKeyEnv}` }
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl('https://api.groq.com/openai/v1/audio/speech', {
+      method: 'POST',
+      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
+      body: JSON.stringify({
+        model: pick(DEFAULT_MODELS, models, 'groq'),
+        input: text,
+        voice: pick(DEFAULT_VOICES, voices, 'groq'),
+        response_format: 'mp3',
+      }),
+      signal,
+    })
+    if (!res.ok) throw new Error(`Groq HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'groq', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  async function deepgram() {
+    const key = await resolveKey(cfg.deepgramKeyEnv)
+    if (!key) return { ok: false, provider: 'deepgram', reason: `no ${cfg.deepgramKeyEnv}` }
+    const model = pick(DEFAULT_MODELS, models, 'deepgram')
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl(`https://api.deepgram.com/v1/speak?model=${encodeURIComponent(model)}`, {
+      method: 'POST',
+      headers: { authorization: `Token ${key}`, 'content-type': 'application/json' },
+      body: JSON.stringify({ text }),
+      signal,
+    })
+    if (!res.ok) throw new Error(`Deepgram HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'deepgram', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+  async function openrouter() {
+    const key = await resolveKey(cfg.openrouterKeyEnv)
+    if (!key) return { ok: false, provider: 'openrouter', reason: `no ${cfg.openrouterKeyEnv}` }
+    const { signal, cleanup } = withCloudTimeout()
+    try {
+      const res = await fetchImpl('https://openrouter.ai/api/v1/audio/speech', {
+      method: 'POST',
+      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
+      body: JSON.stringify({
+        model: pick(DEFAULT_MODELS, models, 'openrouter'),
+        input: text,
+        voice: pick(DEFAULT_VOICES, voices, 'openrouter'),
+        response_format: 'mp3',
+      }),
+      signal,
+    })
+    if (!res.ok) throw new Error(`OpenRouter HTTP ${res.status}`)
+    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
+    return { ok: audio.length > 0, provider: 'openrouter', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
+    } finally {
+      cleanup()
+    }
+  }
+
+  const compatible = {}
+  for (const key of Object.keys(OPENAI_COMPATIBLE)) {
+    compatible[key] = openaiCompatible(key)
+  }
+
+  return { openai, elevenlabs, google, azure, groq, deepgram, openrouter, mimo, ...compatible }
+}

package/lib/providers/local.js

@@ -0,0 +1,99 @@
+import { edgeSpeak, espeakSpeak, minimaxSpeak, piperSpeak } from "../local.js"
+import { KokoroEngine } from "../engines/kokoro.js"
+import { F5Engine } from "../engines/f5.js"
+import { DEFAULT_MODELS, DEFAULT_VOICES } from "./constants.js"
+
+function pick(map, models, key) {
+  const chosen = models && typeof models[key] === "string" ? models[key].trim() : ""
+  return chosen || map[key]
+}
+
+export function makeLocalProviders(deps, req) {
+  const { cfg, modelManager } = deps
+  const { text, lang, models, voices } = req
+
+  async function edge() {
+    try {
+      const voice = pick(DEFAULT_VOICES, voices, 'edge') || pick(DEFAULT_MODELS, models, 'edge')
+      const out = await edgeSpeak(text, { bin: cfg.edgeBin, voice, timeoutMs: cfg.timeoutMs })
+      return { ok: true, provider: 'edge', audio: out.audio, mime: out.mime }
+    } catch (e) {
+      return { ok: false, provider: 'edge', reason: String(e && e.message || e) }
+    }
+  }
+  async function piper() {
+    try {
+      const model = pick(DEFAULT_MODELS, models, 'piper') || cfg.piperModel
+      const out = await piperSpeak(text, { bin: cfg.piperBin, model, timeoutMs: cfg.timeoutMs })
+      return { ok: true, provider: 'piper', audio: out.audio, mime: out.mime }
+    } catch (e) {
+      return { ok: false, provider: 'piper', reason: String(e && e.message || e) }
+    }
+  }
+  async function espeak() {
+    try {
+      const voice = pick(DEFAULT_VOICES, voices, 'espeak') || lang || 'ru'
+      const out = await espeakSpeak(text, { bin: cfg.espeakBin, voice, timeoutMs: cfg.timeoutMs })
+      return { ok: true, provider: 'espeak', audio: out.audio, mime: out.mime }
+    } catch (e) {
+      return { ok: false, provider: 'espeak', reason: String(e && e.message || e) }
+    }
+  }
+  async function kokoro() {
+    try {
+      const modelPath = deps.modelManager ? deps.modelManager.getModelPath('kokoro') : (cfg.kokoroModelPath || '')
+      const engine = new KokoroEngine({ modelPath })
+      if (!engine.isInstalled()) {
+        return { ok: false, provider: 'kokoro', reason: 'Kokoro ONNX model not installed. Use Edge, Piper, or eSpeak for offline TTS.' }
+      }
+      if (!engine.isRuntimeAvailable()) {
+        return {
+          ok: false,
+          provider: 'kokoro',
+          reason: 'Kokoro ONNX runtime is not bundled. Weights alone do not produce speech. Use Edge, Piper, or eSpeak for offline TTS.',
+        }
+      }
+      const voice = pick(DEFAULT_VOICES, voices, 'kokoro') || 'af_bella'
+      const audio = await engine.synthesizeWav(text, voice)
+      return { ok: audio && audio.length > 0, provider: 'kokoro', audio, mime: 'audio/wav' }
+    } catch (e) {
+      return { ok: false, provider: 'kokoro', reason: String(e && e.message || e) }
+    }
+  }
+  async function f5() {
+    try {
+      const modelPath = deps.modelManager ? deps.modelManager.getModelPath('f5') : ''
+      const engine = new F5Engine({ modelPath })
+      const ok = await engine.ping()
+      if (!ok) {
+        return { ok: false, provider: 'f5', reason: 'F5 daemon unavailable. Use Edge, Piper, or eSpeak for offline TTS.' }
+      }
+      if (!engine.isRuntimeAvailable()) {
+        return {
+          ok: false,
+          provider: 'f5',
+          reason: 'F5-TTS synthesis is not bundled. Use Edge, Piper, or eSpeak for offline TTS.',
+        }
+      }
+      const audio = await engine.synthesize(text)
+      return { ok: !!(audio && audio.length), provider: 'f5', audio, mime: 'audio/wav' }
+    } catch (e) {
+      return { ok: false, provider: 'f5', reason: String(e && e.message || e) }
+    }
+  }
+  async function minimax() {
+    try {
+      const out = await minimaxSpeak(text, {
+        bin: cfg.minimaxBin,
+        voice: pick(DEFAULT_VOICES, voices, 'minimax'),
+        model: pick(DEFAULT_MODELS, models, 'minimax'),
+        timeoutMs: cfg.timeoutMs,
+      })
+      return { ok: true, provider: 'minimax', audio: out.audio, mime: out.mime }
+    } catch (e) {
+      return { ok: false, provider: 'minimax', reason: String(e && e.message || e) }
+    }
+  }
+
+  return { kokoro, f5, edge, piper, espeak, minimax }
+}

package/lib/providers.js

@@ -1,22 +1,12 @@
-import { edgeSpeak, espeakSpeak, minimaxSpeak, piperSpeak } from './local.js'
-import { KokoroEngine } from './engines/kokoro.js'
-import { F5Engine } from './engines/f5.js'
-import { PROVIDER_KEYS, OPENAI_COMPATIBLE, DEFAULT_MODELS, DEFAULT_VOICES } from './providers/constants.js'
-export { PROVIDER_KEYS, OPENAI_COMPATIBLE, DEFAULT_MODELS, DEFAULT_VOICES }
-
-function pick(map, models, key) {
-  const chosen = models && typeof models[key] === 'string' ? models[key].trim() : ''
-  return chosen || map[key]
-}
+import { PROVIDER_KEYS, OPENAI_COMPATIBLE, DEFAULT_MODELS, DEFAULT_VOICES } from "./providers/constants.js"
+import { makeCloudProviders } from "./providers/cloud.js"
+import { makeLocalProviders } from "./providers/local.js"
 
-function asBuffer(body) {
-  if (Buffer.isBuffer(body)) return body
-  return Buffer.from(body)
-}
+export { PROVIDER_KEYS, OPENAI_COMPATIBLE, DEFAULT_MODELS, DEFAULT_VOICES }
 
 export function makeProviders(deps, req) {
-  const { resolveKey, fetchImpl, cfg } = deps
-  const { text, lang, signal: parentSignal, models, voices } = req
+  const { cfg } = deps
+  const { signal: parentSignal } = req
   const cloudTimeoutMs = Number(cfg.cloudTimeoutMs) > 0 ? Number(cfg.cloudTimeoutMs) : 10000
 
   function withCloudTimeout() {
@@ -24,11 +14,11 @@ export function makeProviders(deps, req) {
     let timedOut = false
     const timer = setTimeout(() => {
       timedOut = true
-      controller.abort(new Error(`Cloud TTS timeout after ${cloudTimeoutMs}ms`))
+      controller.abort(new Error())
     }, cloudTimeoutMs)
     if (parentSignal) {
       if (parentSignal.aborted) controller.abort(parentSignal.reason)
-      else parentSignal.addEventListener('abort', () => controller.abort(parentSignal.reason), { once: true })
+      else parentSignal.addEventListener(abort, () => controller.abort(parentSignal.reason), { once: true })
     }
     return {
       signal: controller.signal,
@@ -36,364 +26,8 @@ export function makeProviders(deps, req) {
     }
   }
 
-  async function openai() {
-    const key = await resolveKey(cfg.openaiKeyEnv)
-    if (!key) return { ok: false, provider: 'openai', reason: `no ${cfg.openaiKeyEnv}` }
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl((cfg.openaiBaseUrl || 'https://api.openai.com/v1') + '/audio/speech', {
-      method: 'POST',
-      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
-      body: JSON.stringify({
-        model: pick(DEFAULT_MODELS, models, 'openai'),
-        input: text,
-        voice: pick(DEFAULT_VOICES, voices, 'openai'),
-        response_format: 'mp3',
-      }),
-      signal,
-    })
-    if (!res.ok) throw new Error(`OpenAI HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'openai', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-    // One implementation for all OpenAI-compatible providers: base URL and key
-    // name come from the table; everything else is identical.
-  function openaiCompatible(key) {
-    const spec = OPENAI_COMPATIBLE[key]
-    return async function speak() {
-      const keyEnv = (cfg[key + 'KeyEnv'] || spec.keyEnv)
-      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 { signal, cleanup } = withCloudTimeout()
-      try {
-        const res = await fetchImpl(base + '/audio/speech', {
-        method: 'POST',
-        headers: { authorization: `Bearer ${token}`, 'content-type': 'application/json' },
-        body: JSON.stringify({
-          model: pick(DEFAULT_MODELS, models, key),
-          input: text,
-          voice: pick(DEFAULT_VOICES, voices, key),
-          response_format: 'mp3',
-        }),
-        signal,
-      })
-      if (!res.ok) throw new Error(`${key} HTTP ${res.status}`)
-      const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-      return { ok: audio.length > 0, provider: key, audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-      } finally {
-        cleanup()
-      }
-    }
-  }
-
-    // Xiaomi MiMo. Synthesis goes through chat: text as a message, voice/format
-    // as a separate field; audio returns base64 in the response.
-  async function mimo() {
-    const key = await resolveKey(cfg.mimoKeyEnv)
-    if (!key) return { ok: false, provider: 'mimo', reason: `no ${cfg.mimoKeyEnv}` }
-    const base = (cfg.mimoBaseUrl || 'https://api.xiaomimimo.com/v1').replace(/\/+$/, '')
-    const format = cfg.mimoFormat === 'wav' ? 'wav' : 'mp3'
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl(base + '/chat/completions', {
-      method: 'POST',
-      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json', accept: 'application/json' },
-      body: JSON.stringify({
-        model: pick(DEFAULT_MODELS, models, 'mimo'),
-        messages: [{ role: 'assistant', content: text }],
-        audio: { format, voice: pick(DEFAULT_VOICES, voices, 'mimo') },
-        stream: false,
-      }),
-      signal,
-    })
-    const data = await res.json().catch(() => ({}))
-    if (!res.ok) {
-      const detail = data && data.error && data.error.message
-      throw new Error(`MiMo HTTP ${res.status}${detail ? ': ' + detail : ''}`)
-    }
-    const encoded = data && data.choices && data.choices[0]
-      && data.choices[0].message && data.choices[0].message.audio
-      && data.choices[0].message.audio.data
-    if (typeof encoded !== 'string' || !encoded) {
-      return { ok: false, provider: 'mimo', reason: 'response missing choices[0].message.audio.data' }
-    }
-    const audio = Buffer.from(encoded, 'base64')
-    return {
-      ok: audio.length > 0,
-      provider: 'mimo',
-      audio,
-      mime: format === 'wav' ? 'audio/wav' : 'audio/mpeg',
-      reason: audio.length ? '' : 'empty audio',
-    }
-    } finally {
-      cleanup()
-    }
-  }
-
-  // MiniMax uses their CLI on purpose; no custom protocol here.
-  async function minimax() {
-    try {
-      const out = await minimaxSpeak(text, {
-        bin: cfg.minimaxBin,
-        voice: pick(DEFAULT_VOICES, voices, 'minimax'),
-        model: pick(DEFAULT_MODELS, models, 'minimax'),
-        timeoutMs: cfg.timeoutMs,
-      })
-      return { ok: true, provider: 'minimax', audio: out.audio, mime: out.mime }
-    } catch (e) {
-      return { ok: false, provider: 'minimax', reason: String(e && e.message || e) }
-    }
-  }
-
-  async function elevenlabs() {
-    const key = await resolveKey(cfg.elevenlabsKeyEnv)
-    if (!key) return { ok: false, provider: 'elevenlabs', reason: `no ${cfg.elevenlabsKeyEnv}` }
-    const voice = pick(DEFAULT_VOICES, voices, 'elevenlabs')
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl(`https://api.elevenlabs.io/v1/text-to-speech/${encodeURIComponent(voice)}`, {
-      method: 'POST',
-      headers: { 'xi-api-key': key, 'content-type': 'application/json', accept: 'audio/mpeg' },
-      body: JSON.stringify({
-        text,
-        model_id: pick(DEFAULT_MODELS, models, 'elevenlabs'),
-      }),
-      signal,
-    })
-    if (!res.ok) throw new Error(`ElevenLabs HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'elevenlabs', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function google() {
-    const key = await resolveKey(cfg.googleKeyEnv)
-    if (!key) return { ok: false, provider: 'google', reason: `no ${cfg.googleKeyEnv}` }
-    const model = pick(DEFAULT_MODELS, models, 'google')
-    const voiceName = pick(DEFAULT_VOICES, voices, 'google')
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl(
-      `https://generativelanguage.googleapis.com/v1beta/models/${encodeURIComponent(model)}:generateContent?key=${encodeURIComponent(key)}`,
-      {
-        method: 'POST',
-        headers: { 'content-type': 'application/json' },
-        body: JSON.stringify({
-          contents: [{ parts: [{ text }] }],
-          generationConfig: {
-            responseModalities: ['AUDIO'],
-            speechConfig: { voiceConfig: { prebuiltVoiceConfig: { voiceName } } },
-          },
-        }),
-        signal,
-      },
-    )
-    if (!res.ok) throw new Error(`Google HTTP ${res.status}`)
-    const data = await res.json()
-    const b64 = data?.candidates?.[0]?.content?.parts?.find((p) => p.inlineData)?.inlineData?.data
-    if (!b64) return { ok: false, provider: 'google', reason: 'no inline audio' }
-    const audio = Buffer.from(b64, 'base64')
-    return { ok: audio.length > 0, provider: 'google', audio, mime: 'audio/wav', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function azure() {
-    const key = await resolveKey(cfg.azureKeyEnv)
-    if (!key) return { ok: false, provider: 'azure', reason: `no ${cfg.azureKeyEnv}` }
-    const region = (cfg.azureRegion || '').trim()
-    if (!region) return { ok: false, provider: 'azure', reason: 'azureRegion is empty' }
-    const voice = pick(DEFAULT_VOICES, voices, 'azure')
-    const roleCfg = req.role && cfg.roles && cfg.roles[req.role]
-    const style = roleCfg && roleCfg.ssmlStyle
-    const safeStyle = style ? String(style).replace(/["<>&]/g, '') : ''
-    const inner = safeStyle
-      ? `<mstts:express-as style="${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}">`
-      + `${inner}</voice></speak>`
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl(`https://${region}.tts.speech.microsoft.com/cognitiveservices/v1`, {
-      method: 'POST',
-      headers: {
-        'Ocp-Apim-Subscription-Key': key,
-        'Content-Type': 'application/ssml+xml',
-        'X-Microsoft-OutputFormat': 'audio-16khz-128kbitrate-mono-mp3',
-      },
-      body: ssml,
-      signal,
-    })
-    if (!res.ok) throw new Error(`Azure HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'azure', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function groq() {
-    const key = await resolveKey(cfg.groqKeyEnv)
-    if (!key) return { ok: false, provider: 'groq', reason: `no ${cfg.groqKeyEnv}` }
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl('https://api.groq.com/openai/v1/audio/speech', {
-      method: 'POST',
-      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
-      body: JSON.stringify({
-        model: pick(DEFAULT_MODELS, models, 'groq'),
-        input: text,
-        voice: pick(DEFAULT_VOICES, voices, 'groq'),
-        response_format: 'mp3',
-      }),
-      signal,
-    })
-    if (!res.ok) throw new Error(`Groq HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'groq', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function deepgram() {
-    const key = await resolveKey(cfg.deepgramKeyEnv)
-    if (!key) return { ok: false, provider: 'deepgram', reason: `no ${cfg.deepgramKeyEnv}` }
-    const model = pick(DEFAULT_MODELS, models, 'deepgram')
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl(`https://api.deepgram.com/v1/speak?model=${encodeURIComponent(model)}`, {
-      method: 'POST',
-      headers: { authorization: `Token ${key}`, 'content-type': 'application/json' },
-      body: JSON.stringify({ text }),
-      signal,
-    })
-    if (!res.ok) throw new Error(`Deepgram HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'deepgram', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function openrouter() {
-    const key = await resolveKey(cfg.openrouterKeyEnv)
-    if (!key) return { ok: false, provider: 'openrouter', reason: `no ${cfg.openrouterKeyEnv}` }
-    const { signal, cleanup } = withCloudTimeout()
-    try {
-      const res = await fetchImpl('https://openrouter.ai/api/v1/audio/speech', {
-      method: 'POST',
-      headers: { authorization: `Bearer ${key}`, 'content-type': 'application/json' },
-      body: JSON.stringify({
-        model: pick(DEFAULT_MODELS, models, 'openrouter'),
-        input: text,
-        voice: pick(DEFAULT_VOICES, voices, 'openrouter'),
-        response_format: 'mp3',
-      }),
-      signal,
-    })
-    if (!res.ok) throw new Error(`OpenRouter HTTP ${res.status}`)
-    const audio = asBuffer(Buffer.from(await res.arrayBuffer()))
-    return { ok: audio.length > 0, provider: 'openrouter', audio, mime: 'audio/mpeg', reason: audio.length ? '' : 'empty audio' }
-    } finally {
-      cleanup()
-    }
-  }
-
-  async function edge() {
-    try {
-      const voice = pick(DEFAULT_VOICES, voices, 'edge') || pick(DEFAULT_MODELS, models, 'edge')
-      const out = await edgeSpeak(text, { bin: cfg.edgeBin, voice, timeoutMs: cfg.timeoutMs })
-      return { ok: true, provider: 'edge', audio: out.audio, mime: out.mime }
-    } catch (e) {
-      return { ok: false, provider: 'edge', reason: String(e && e.message || e) }
-    }
-  }
-
-  async function piper() {
-    try {
-      const model = pick(DEFAULT_MODELS, models, 'piper') || cfg.piperModel
-      const out = await piperSpeak(text, { bin: cfg.piperBin, model, timeoutMs: cfg.timeoutMs })
-      return { ok: true, provider: 'piper', audio: out.audio, mime: out.mime }
-    } catch (e) {
-      return { ok: false, provider: 'piper', reason: String(e && e.message || e) }
-    }
-  }
-
-  async function espeak() {
-    try {
-      const voice = pick(DEFAULT_VOICES, voices, 'espeak') || lang || 'ru'
-      const out = await espeakSpeak(text, { bin: cfg.espeakBin, voice, timeoutMs: cfg.timeoutMs })
-      return { ok: true, provider: 'espeak', audio: out.audio, mime: out.mime }
-    } catch (e) {
-      return { ok: false, provider: 'espeak', reason: String(e && e.message || e) }
-    }
-  }
-
-  async function kokoro() {
-    try {
-      const modelPath = deps.modelManager ? deps.modelManager.getModelPath('kokoro') : (cfg.kokoroModelPath || '')
-      const engine = new KokoroEngine({ modelPath })
-      if (!engine.isInstalled()) {
-        return { ok: false, provider: 'kokoro', reason: 'Kokoro ONNX model not installed. Use Edge, Piper, or eSpeak for offline TTS.' }
-      }
-      if (!engine.isRuntimeAvailable()) {
-        return {
-          ok: false,
-          provider: 'kokoro',
-          reason: 'Kokoro ONNX runtime is not bundled. Weights alone do not produce speech. Use Edge, Piper, or eSpeak for offline TTS.',
-        }
-      }
-      const voice = pick(DEFAULT_VOICES, voices, 'kokoro') || 'af_bella'
-      const audio = await engine.synthesizeWav(text, voice)
-      return { ok: audio && audio.length > 0, provider: 'kokoro', audio, mime: 'audio/wav' }
-    } catch (e) {
-      return { ok: false, provider: 'kokoro', reason: String(e && e.message || e) }
-    }
-  }
-
-  async function f5() {
-    try {
-      const modelPath = deps.modelManager ? deps.modelManager.getModelPath('f5') : ''
-      const engine = new F5Engine({ modelPath })
-      const ok = await engine.ping()
-      if (!ok) {
-        return { ok: false, provider: 'f5', reason: 'F5 daemon unavailable. Use Edge, Piper, or eSpeak for offline TTS.' }
-      }
-      if (!engine.isRuntimeAvailable()) {
-        return {
-          ok: false,
-          provider: 'f5',
-          reason: 'F5-TTS synthesis is not bundled. Use Edge, Piper, or eSpeak for offline TTS.',
-        }
-      }
-      const audio = await engine.synthesize(text)
-      return { ok: !!(audio && audio.length), provider: 'f5', audio, mime: 'audio/wav' }
-    } catch (e) {
-      return { ok: false, provider: 'f5', reason: String(e && e.message || e) }
-    }
-  }
-
-  const compatible = {}
-  for (const key of Object.keys(OPENAI_COMPATIBLE)) compatible[key] = openaiCompatible(key)
-
-  return { kokoro, f5, openai, elevenlabs, google, azure, groq, deepgram, openrouter, mimo, ...compatible, edge, piper, espeak, minimax }
-}
+  const cloud = makeCloudProviders(deps, req, withCloudTimeout)
+  const local = makeLocalProviders(deps, req)
 
-function escapeXml(s) {
-  return String(s)
-    .replace(/&/g, '&amp;')
-    .replace(/</g, '&lt;')
-    .replace(/>/g, '&gt;')
-    .replace(/"/g, '&quot;')
+  return Object.assign({}, cloud, local)
 }

package/lib/routes.js

@@ -65,7 +65,7 @@ export function registerHttpRoutes(ctx, api) {
         res.write(': connected\n\n')
         addStreamSubscriber(res)
         req.on('close', () => removeStreamSubscriber(res))
-      } catch {
+      } catch { /* SSE connection failed */ 
         removeStreamSubscriber(res)
       }
     },
@@ -110,7 +110,7 @@ export function registerHttpRoutes(ctx, api) {
       }
       try {
         // Fire installation in background or await initial start
-        modelManager.installModel(engine).catch(() => {})
+        modelManager.installModel(engine).catch(() => { /* background install error tracked in manager state */ })
         const current = await modelManager.getStatus(engine)
         writeJson(res, 200, {
           ok: true,

package/lib/text.js

@@ -176,12 +176,12 @@ export function restoreAbbreviations(text) {
 /**
  * Split text into chunks on sentence boundaries.
  *
- * Нужно, чтобы чтение начиналось почти сразу: длинный ответ синтезируется
- * целиком долго, а первое предложение — быстро. Куски короче нижнего порога
- * прилипают к следующему: отдельная фраза «Да.» звучит рвано.
+ * Enables low latency playback: synthesizing a whole long reply takes seconds,
+ * whereas the first sentence is ready almost immediately. Chunks shorter than
+ * the min threshold attach to the next one so speech sounds fluid.
  *
- * @param text {string} уже очищенный текст
- * @param maxChars {number} верхняя граница куска
+ * @param text {string} pre-cleaned text
+ * @param maxChars {number} upper bound for one chunk
  * @returns {string[]}
  */
 export function splitSentences(text, maxChars) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-tts",
-  "version": "0.4.2",
+  "version": "0.4.5",
   "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",