@goodandready/dsh-tts 0.4.24 → 0.4.25

package/lib/index.js

@@ -24,7 +24,7 @@ import {
 
 // Scoped identity matching package.json and cordis.patch.yml
 export const name = '@goodandready/dsh-tts'
-export const inject = ['tools', 'credentials', 'llm', 'webServer', 'settings']
+export const inject = ['tools', 'credentials', 'llm', 'webServer']
 
 const NS = 'dsh-tts'
 
@@ -45,34 +45,43 @@ const RoleOverride = z.object({
   ssmlStyle: z.string().default(''),
 })
 
+// Unwrap Volatile boxes before any caller reads a value. A volatile field holds a
+// box rather than its value, and the Loader mutates those boxes in place.
+export function plainConfig(value) {
+  if (value === null || typeof value !== 'object') return value
+  if (Array.isArray(value)) return value.map(plainConfig)
+  if (typeof value.get === 'function') return plainConfig(value.get())
+  return Object.fromEntries(Object.entries(value).map(([k, v]) => [k, plainConfig(v)]))
+}
+
 export const Config = z.object({
-  speakReplies: z.boolean().default(false)
+  speakReplies: z.boolean().volatile().default(false)
     .description('When on, speak each finished agent reply in the Web UI.'),
-  enableLocalEngines: z.boolean().default(false)
+  enableLocalEngines: z.boolean().volatile().default(false)
     .description('Enable offline local TTS engines (Kokoro / F5).'),
-  kokoroEnabled: z.boolean().default(false)
+  kokoroEnabled: z.boolean().volatile().default(false)
     .description('Use local Kokoro-82M on CPU when installed.'),
-  f5Enabled: z.boolean().default(false)
+  f5Enabled: z.boolean().volatile().default(false)
     .description('Use local F5-TTS on GPU when installed.'),
-  streamingEnabled: z.boolean().default(true)
+  streamingEnabled: z.boolean().volatile().default(true)
     .description('Stream audio in real-time using AudioWorklet and SSE for sub-300ms latency.'),
-  skipCode: z.boolean().default(true)
+  skipCode: z.boolean().volatile().default(true)
     .description('Replace fenced code blocks with a short spoken notice instead of reading them aloud.'),
-  cache: z.boolean().default(true)
+  cache: z.boolean().volatile().default(true)
     .description('Reuse synthesized audio for repeated phrases instead of paying and waiting again.'),
-  cacheMaxMb: z.number().default(100)
+  cacheMaxMb: z.number().volatile().default(100)
     .description('Disk limit for the synthesis cache; least recently used items are evicted first.'),
-  narrateQuotesOnly: z.boolean().default(false)
+  narrateQuotesOnly: z.boolean().volatile().default(false)
     .description('Speak only quoted fragments.'),
-  skipActions: z.boolean().default(false)
+  skipActions: z.boolean().volatile().default(false)
     .description('Drop *asterisk action* blocks instead of reading them.'),
-  removeRegex: z.string().default('')
+  removeRegex: z.string().volatile().default('')
     .description('Custom global regex whose matches are removed before synthesis.'),
-  autoDetect: z.boolean().default(false)
+  autoDetect: z.boolean().volatile().default(false)
     .description('Guess ru/en per spoken piece instead of using the language setting.'),
-  customBaseUrl: z.string().default('')
+  customBaseUrl: z.string().volatile().default('')
     .description('Origin of an OpenAI-compatible /audio/speech endpoint; empty skips it.'),
-  customKeyEnv: z.string().default('CUSTOM_TTS_API_KEY'),
+  customKeyEnv: z.string().volatile().default('CUSTOM_TTS_API_KEY'),
   pronunciation: z.array(z.object({
     from: z.string().default(''),
     to: z.string().default(''),
@@ -80,29 +89,29 @@ export const Config = z.object({
     lang: z.string().default(''),
   })).default([])
     .description('Say this instead of that. Applied top-down; /regex/ form allowed in from.'),
-  enableItDictionary: z.boolean().default(true)
+  enableItDictionary: z.boolean().volatile().default(true)
     .description('Auto-correct common IT terminology pronunciation (SQL, Nginx, K8s, etc.).'),
-  longReply: z.string().default('truncate')
+  longReply: z.string().volatile().default('truncate')
     .description('What to do when a reply exceeds the length limit.'),
-  summaryModel: z.string().default('')
+  summaryModel: z.string().volatile().default('')
     .description('provider/model for the spoken retelling; empty uses the conversation default.'),
-  summarySentences: z.number().default(3)
+  summarySentences: z.number().volatile().default(3)
     .description('Sentences in the spoken retelling.'),
-  roles: z.dict(RoleOverride).default({
+  roles: z.dict(RoleOverride).volatile().default({
     reply: { provider: '', model: '', voice: '', chime: '', ssmlStyle: '' },
     approval: { provider: '', model: '', voice: '', chime: '', ssmlStyle: '' },
     error: { provider: '', model: '', voice: '', chime: '', ssmlStyle: '' },
   })
     .description('Per-role and subagent overrides; empty fields inherit the main chain.'),
-  voiceDuplexEnabled: z.boolean().default(false)
+  voiceDuplexEnabled: z.boolean().volatile().default(false)
     .description('Full-duplex voice conversation with dsh-voice.'),
-  vadBargeIn: z.boolean().default(true)
+  vadBargeIn: z.boolean().volatile().default(true)
     .description('Mute speech synthesis immediately when voice activity is detected.'),
-  messengerTtsEnabled: z.boolean().default(false)
+  messengerTtsEnabled: z.boolean().volatile().default(false)
     .description('Send synthesized voice notes to messengers via @goodandready/dsh-messenger-gateway.'),
-  autoDetectSubagent: z.boolean().default(true)
+  autoDetectSubagent: z.boolean().volatile().default(true)
     .description('Auto-detect subagent roles in multi-agent sessions and route to dedicated voice profiles.'),
-  language: z.string().default('ru'),
+  language: z.string().volatile().default('ru'),
   speakAsItGoes: z
     .boolean()
     .description('Speak each reply as it lands instead of waiting for the whole turn to finish. '
@@ -138,43 +147,43 @@ export const Config = z.object({
     .string()
     .description('Short sound before an announcement: ding, beep or none.')
     .default('ding'),
-  maxChars: z.number().default(4000)
+  maxChars: z.number().volatile().default(4000)
     .description('Longer replies are truncated before synthesis.'),
-  chain: z.array(ChainEntry)
+  chain: z.array(ChainEntry).volatile()
     .default([
       { provider: 'edge', model: '', voice: 'ru-RU-SvetlanaNeural' },
       { provider: 'piper', model: '', voice: '' },
       { provider: 'espeak', model: '', voice: 'ru' },
     ])
     .description('Fallback chain. Order is the order of attempts. A provider without a key is skipped.'),
-  openaiKeyEnv: z.string().default('OPENAI_API_KEY'),
-  openaiBaseUrl: z.string().default('https://api.openai.com/v1'),
-  elevenlabsKeyEnv: z.string().default('ELEVENLABS_API_KEY'),
-  googleKeyEnv: z.string().default('GEMINI_API_KEY'),
-  azureKeyEnv: z.string().default('AZURE_SPEECH_KEY'),
-  azureRegion: z.string().default('')
+  openaiKeyEnv: z.string().volatile().default('OPENAI_API_KEY'),
+  openaiBaseUrl: z.string().volatile().default('https://api.openai.com/v1'),
+  elevenlabsKeyEnv: z.string().volatile().default('ELEVENLABS_API_KEY'),
+  googleKeyEnv: z.string().volatile().default('GEMINI_API_KEY'),
+  azureKeyEnv: z.string().volatile().default('AZURE_SPEECH_KEY'),
+  azureRegion: z.string().volatile().default('')
     .description('Azure Speech region, e.g. eastus. Required for the azure provider.'),
-  mimoKeyEnv: z.string().description('Credential holding the Xiaomi MiMo key.').default('MIMO_API_KEY'),
-  mimoBaseUrl: z.string().description('MiMo API root. Synthesis goes through its chat endpoint.').default('https://api.xiaomimimo.com/v1'),
-  mimoFormat: z.string().description('MiMo output format: mp3 or wav.').default('mp3'),
-  minimaxBin: z.string()
+  mimoKeyEnv: z.string().volatile().description('Credential holding the Xiaomi MiMo key.').default('MIMO_API_KEY'),
+  mimoBaseUrl: z.string().volatile().description('MiMo API root. Synthesis goes through its chat endpoint.').default('https://api.xiaomimimo.com/v1'),
+  mimoFormat: z.string().volatile().description('MiMo output format: mp3 or wav.').default('mp3'),
+  minimaxBin: z.string().volatile()
     .description('MiniMax CLI, looked up in PATH unless absolute. Install and log in separately; '
       + 'without it the provider declines and the chain moves on.')
     .default('mmx'),
-  siliconflowKeyEnv: z.string().description('Credential holding the SiliconFlow key.').default('SILICONFLOW_API_KEY'),
-  deepinfraKeyEnv: z.string().description('Credential holding the DeepInfra key.').default('DEEPINFRA_API_KEY'),
-  fireworksKeyEnv: z.string().description('Credential holding the Fireworks key.').default('FIREWORKS_API_KEY'),
-  groqKeyEnv: z.string().default('GROQ_API_KEY'),
-  deepgramKeyEnv: z.string().default('DEEPGRAM_API_KEY'),
-  openrouterKeyEnv: z.string().default('OPENROUTER_API_KEY'),
-  edgeBin: z.string().default('edge-tts')
+  siliconflowKeyEnv: z.string().volatile().description('Credential holding the SiliconFlow key.').default('SILICONFLOW_API_KEY'),
+  deepinfraKeyEnv: z.string().volatile().description('Credential holding the DeepInfra key.').default('DEEPINFRA_API_KEY'),
+  fireworksKeyEnv: z.string().volatile().description('Credential holding the Fireworks key.').default('FIREWORKS_API_KEY'),
+  groqKeyEnv: z.string().volatile().default('GROQ_API_KEY'),
+  deepgramKeyEnv: z.string().volatile().default('DEEPGRAM_API_KEY'),
+  openrouterKeyEnv: z.string().volatile().default('OPENROUTER_API_KEY'),
+  edgeBin: z.string().volatile().default('edge-tts')
     .description('edge-tts CLI. Looked up in PATH unless an absolute path is given.'),
-  piperBin: z.string().default('piper'),
-  piperModel: z.string().default('')
+  piperBin: z.string().volatile().default('piper'),
+  piperModel: z.string().volatile().default('')
     .description('Path to a Piper ONNX model. The piper provider is skipped while this is empty.'),
-  espeakBin: z.string().default('espeak-ng'),
-  timeoutMs: z.number().default(60000),
-  maxQueue: z.number().default(8),
+  espeakBin: z.string().volatile().default('espeak-ng'),
+  timeoutMs: z.number().volatile().default(60000),
+  maxQueue: z.number().volatile().default(8),
 })
 
 let utteranceSeq = 0
@@ -183,18 +192,11 @@ export function apply(ctx, config) {
   const providerBreaker = createProviderBreaker()
 
   let getConfig = () => config
-  const live = () => Config(structuredClone(getConfig() ?? {})) ?? config
-
-  let settingsApi
-  ctx.inject(['settings'], (sctx) => {
-    const scope = sctx.settings.register(NS, Config, { base: config })
-    settingsApi = scope
-    getConfig = () => scope.get() ?? config
-    sctx.effect(() => () => {
-      settingsApi = undefined
-      getConfig = () => config
-    })
-  })
+  const live = () => plainConfig(Config(structuredClone(getConfig() ?? {})) ?? config)
+
+  // settings.register was removed before 0.1.7-rc.2. The host hands the profile entry
+  // config to apply() directly; live() below re-resolves it through the schema and
+  // unwraps, so the Loader's in-place box updates are picked up without a service.
 
   const pending = []
   const collectors = new Map()

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-tts",
-  "version": "0.4.24",
+  "version": "0.4.25",
   "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",
@@ -67,7 +67,7 @@
     "@deepseek-ai/dsh-credentials": "^0.1.7-rc.2 || ^0.2.0-rc.1",
     "@deepseek-ai/dsh-host-webserver": "^0.1.7-rc.2 || ^0.2.0-rc.1",
     "@deepseek-ai/dsh-settings": "^0.1.7-rc.2 || ^0.2.0-rc.1",
-    "@deepseek-ai/schemastery": "^3.18.1"
+    "@deepseek-ai/schemastery": "^3.18.4"
   },
   "publishConfig": {
     "access": "public"