@xopcai/xopc 0.0.115 → 0.0.116

package/dist/src/tunnel/frpc-extract.js

@@ -1,7 +1,7 @@
 import { createLogger } from "../utils/logger/index.js";
 import { init_logger } from "../utils/logger.js";
-import { dirname, join } from "node:path";
 import { existsSync, mkdirSync, readFileSync, rmSync, writeFileSync } from "node:fs";
+import { dirname, join } from "node:path";
 import { spawn } from "node:child_process";
 import AdmZip from "adm-zip";
 import { gunzipSync } from "node:zlib";

package/dist/src/tunnel/tunnel-state.js

@@ -1,7 +1,7 @@
 import { resolveStateDir } from "../config/paths-state.js";
 import { init_paths } from "../config/paths.js";
-import { join } from "node:path";
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
+import { join } from "node:path";
 //#region src/tunnel/tunnel-state.ts
 init_paths();
 const TUNNEL_STATE_FILE = "tunnel.json";

package/dist/src/utils/logger/audit.js

@@ -1,7 +1,7 @@
 import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
 import { getLogDir, init_config } from "./config.js";
-import path from "path";
 import { existsSync, mkdirSync } from "fs";
+import path from "path";
 import { appendFile } from "fs/promises";
 //#region src/utils/logger/audit.ts
 /**

package/dist/src/utils/logger/log-store.js

@@ -1,6 +1,6 @@
 import { init_pino_record, logEntrySearchText, pinoRecordToLogEntry } from "./pino-record.js";
-import { basename, join } from "path";
 import { createReadStream, existsSync, mkdirSync, readdirSync, statSync } from "fs";
+import { basename, join } from "path";
 import { Readable } from "stream";
 import { readFile } from "fs/promises";
 import { gunzip } from "zlib";

package/dist/src/utils/logger/rotation.js

@@ -1,7 +1,7 @@
 import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
 import { config, getLogDir, init_config } from "./config.js";
-import { join } from "path";
 import { readFileSync, readdirSync, statSync, unlinkSync } from "fs";
+import { join } from "path";
 import { writeFile } from "fs/promises";
 import { gzip } from "zlib";
 import { promisify } from "util";

package/dist/src/utils/logger/streams.js

@@ -2,8 +2,8 @@ import { __esmMin } from "../../../_virtual/_rolldown/runtime.js";
 import { config, getLogDir, init_config } from "./config.js";
 import { emitLogEntry, hasSubscribers, init_log_stream } from "./log-stream.js";
 import { init_pino_record, pinoRecordToLogEntry } from "./pino-record.js";
-import path from "path";
 import { createWriteStream, existsSync, mkdirSync } from "fs";
+import path from "path";
 import { Writable } from "stream";
 //#region src/utils/logger/streams.ts
 /**

package/dist/src/voice/tts/audio.js

@@ -1,8 +1,8 @@
 import { createLogger } from "../../utils/logger/index.js";
 import { init_logger } from "../../utils/logger.js";
+import { spawn } from "child_process";
 import { join } from "path";
 import { unlink, writeFile } from "fs/promises";
-import { spawn } from "child_process";
 import { tmpdir } from "os";
 //#region src/voice/tts/audio.ts
 /**

package/dist/src/voice/tts/openai-compatible-speech.d.ts

@@ -42,7 +42,7 @@ export interface OpenAiCompatibleSpeechProviderExtraJsonBodyField {
 export interface OpenAiCompatibleSpeechProviderOptions {
     /** Canonical provider id (e.g. "openai", "minimax"). */
     id: string;
-    /** Optional aliases for back-compat config strings. */
+    /** Optional aliases for provider lookup. */
     aliases?: readonly string[];
     /** Human-readable label for error messages and logs (e.g. "OpenAI", "MiniMax"). */
     label: string;

package/dist/src/voice/tts/openai-compatible-speech.js.map

@@ -1 +1 @@
-{"version":3,"file":"openai-compatible-speech.js","names":["normalizeBaseUrlGeneric"],"sources":["../../../../src/voice/tts/openai-compatible-speech.ts"],"sourcesContent":["/**\n * Factory: createOpenAiCompatibleSpeechProvider\n *\n * Most TTS providers (OpenAI, MiniMax, ElevenLabs, OpenRouter, ...) speak the\n * OpenAI `/audio/speech` POST shape. This factory eliminates the boilerplate so\n * a new compatible vendor is ~30 lines of declaration instead of a full provider\n * implementation.\n *\n * Ported from openclaw/src/tts/openai-compatible-speech-provider.ts (commit\n * baseline 2026-05-08), with the following INTENTIONAL OMISSIONS per\n * docs/voice-rearchitecture.md §15.3:\n *   - resolveTalkConfig / resolveTalkOverrides (talk-mode is excluded)\n *   - Persona resolution context (persona system is excluded)\n *   - Custom dispatcherPolicy (xopc uses Node 22 native fetch — see ssrf-guard.ts)\n *\n * DECISION: We use `Record<string, unknown>` for the extra-config slot rather\n * than openclaw's generic `<ExtraConfig extends Record<string, unknown>>`. The\n * generic adds significant TypeScript complexity but no runtime safety; we\n * declare per-provider config types in each provider file's own normalizer.\n *\n * DECISION: `voiceCompatibleResponseFormats` defaults to `['opus']` if omitted,\n * because Telegram voice notes require opus/ogg. Providers that natively support\n * opus output (OpenAI, ElevenLabs) can rely on this; others (MiniMax mp3-only)\n * pass an empty array and downstream ffmpeg-compresses to opus.\n */\n\nimport {\n  ProviderHttpError,\n  fetchWithTimeoutGuarded,\n  normalizeBaseUrl as normalizeBaseUrlGeneric,\n  postJsonRequest,\n} from '../../media-shared/http/index.js';\nimport { createLogger } from '../../utils/logger.js';\n\nimport type {\n  SpeechDirectiveTokenParseContext,\n  SpeechDirectiveTokenParseResult,\n  SpeechProviderConfig,\n  SpeechProviderConfiguredContext,\n  SpeechProviderOverrides,\n  SpeechProviderPlugin,\n  SpeechProviderResolveConfigContext,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechVoiceOption,\n} from './speech-provider-types.js';\n\nconst log = createLogger('SpeechProvider:OpenAiCompat');\n\ninterface OpenAiCompatibleBaseConfig {\n  apiKey?: string;\n  baseUrl?: string;\n  model: string;\n  voice: string;\n  speed?: number;\n  responseFormat?: string;\n}\n\nexport type OpenAiCompatibleSpeechProviderConfig = OpenAiCompatibleBaseConfig &\n  Record<string, unknown>;\n\nexport interface OpenAiCompatibleSpeechProviderExtraJsonBodyField {\n  /** Key to read from normalized config. */\n  configKey: string;\n  /** Key to send in the HTTP body. Defaults to `configKey`. */\n  requestKey?: string;\n}\n\nexport interface OpenAiCompatibleSpeechProviderOptions {\n  /** Canonical provider id (e.g. \"openai\", \"minimax\"). */\n  id: string;\n  /** Optional aliases for back-compat config strings. */\n  aliases?: readonly string[];\n  /** Human-readable label for error messages and logs (e.g. \"OpenAI\", \"MiniMax\"). */\n  label: string;\n  /** Sort key for UI auto-fallback ordering. Lower = higher priority. */\n  autoSelectOrder: number;\n  /** Allowed model ids. Used by listVoices and validation. */\n  models: readonly string[];\n  /** Allowed voice ids. */\n  voices: readonly string[];\n  /** Default model when config does not specify. */\n  defaultModel: string;\n  /** Default voice. */\n  defaultVoice: string;\n  /** Default base URL (e.g. \"https://api.openai.com/v1\"). */\n  defaultBaseUrl: string;\n  /** Env var name for the api key (e.g. \"OPENAI_API_KEY\"). */\n  envKey: string;\n  /** Allowed response formats. */\n  responseFormats: readonly string[];\n  /** Default response format (e.g. \"opus\" for OpenAI, \"mp3\" for MiniMax). */\n  defaultResponseFormat: string;\n  /** Response formats that produce a voice-note-compatible buffer (no ffmpeg needed). */\n  voiceCompatibleResponseFormats?: readonly string[];\n  /** Extra static headers (e.g. `{ \"OpenAI-Beta\": \"...\" }`). Auth header is added automatically. */\n  extraHeaders?: Record<string, string>;\n  /** Read provider-specific extra config keys (e.g. MiniMax's `groupId`). */\n  readExtraConfig?: (raw: Record<string, unknown> | undefined) => Record<string, unknown>;\n  /** Provider-specific extra body fields to copy from config into request JSON. */\n  extraJsonBodyFields?: readonly OpenAiCompatibleSpeechProviderExtraJsonBodyField[];\n  /** Override the API error label (defaults to `${label} TTS API error`). */\n  apiErrorLabel?: string;\n  /** Override the missing-key error message. */\n  missingApiKeyError?: string;\n}\n\nfunction asObject(value: unknown): Record<string, unknown> | undefined {\n  return typeof value === 'object' && value !== null && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined;\n}\n\nfunction trimToUndefined(value: unknown): string | undefined {\n  if (typeof value !== 'string') {\n    return undefined;\n  }\n  const trimmed = value.trim();\n  return trimmed.length > 0 ? trimmed : undefined;\n}\n\nfunction asFiniteNumber(value: unknown): number | undefined {\n  if (typeof value === 'number' && Number.isFinite(value)) {\n    return value;\n  }\n  if (typeof value === 'string') {\n    const parsed = Number.parseFloat(value);\n    if (Number.isFinite(parsed)) {\n      return parsed;\n    }\n  }\n  return undefined;\n}\n\nfunction compactKey(value: string): string {\n  return value.replace(/[^a-z0-9]+/giu, '').toLowerCase();\n}\n\nfunction normalizeResponseFormat(label: string, formats: readonly string[], value: unknown): string | undefined {\n  const next = trimToUndefined(value)?.toLowerCase();\n  if (!next) {\n    return undefined;\n  }\n  if (formats.includes(next)) {\n    return next;\n  }\n  throw new Error(`Invalid ${label} speech responseFormat: ${next}`);\n}\n\nfunction buildExtraJsonBodyFields(\n  config: OpenAiCompatibleSpeechProviderConfig,\n  fields: readonly OpenAiCompatibleSpeechProviderExtraJsonBodyField[] | undefined,\n): Record<string, unknown> {\n  const body: Record<string, unknown> = {};\n  for (const field of fields ?? []) {\n    const value = config[field.configKey];\n    if (value !== undefined && value !== null && value !== '') {\n      body[field.requestKey ?? field.configKey] = value;\n    }\n  }\n  return body;\n}\n\nfunction readSpeechOverrides(overrides: SpeechProviderOverrides | undefined): {\n  model?: string;\n  voice?: string;\n  speed?: number;\n} {\n  if (!overrides) {\n    return {};\n  }\n  return {\n    model: trimToUndefined(overrides.model ?? overrides.modelId),\n    voice: trimToUndefined(overrides.voice ?? overrides.voiceId),\n    speed: asFiniteNumber(overrides.speed),\n  };\n}\n\nfunction parseDirectiveTokenInternal(\n  ctx: SpeechDirectiveTokenParseContext,\n  providerConfigKey: string,\n): SpeechDirectiveTokenParseResult {\n  const compactProviderKey = compactKey(providerConfigKey);\n  switch (ctx.key) {\n    case 'voice':\n    case 'voice_id':\n    case 'voiceid':\n    case `${providerConfigKey}_voice`:\n    case `${compactProviderKey}voice`:\n      if (!ctx.policy.allowVoice) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { voice: ctx.value } };\n    case 'model':\n    case 'model_id':\n    case 'modelid':\n    case `${providerConfigKey}_model`:\n    case `${compactProviderKey}model`:\n      if (!ctx.policy.allowModelId) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { model: ctx.value } };\n    case 'speed': {\n      const speed = asFiniteNumber(ctx.value);\n      if (speed === undefined) {\n        return { handled: true, warnings: [`Invalid speed value: ${ctx.value}`] };\n      }\n      return { handled: true, overrides: { speed } };\n    }\n    default:\n      return { handled: false };\n  }\n}\n\nexport function createOpenAiCompatibleSpeechProvider(\n  options: OpenAiCompatibleSpeechProviderOptions,\n): SpeechProviderPlugin {\n  const providerConfigKey = options.id;\n  const voiceCompatibleResponseFormats = options.voiceCompatibleResponseFormats ?? ['opus'];\n\n  function readExtraConfig(raw: Record<string, unknown> | undefined): Record<string, unknown> {\n    return options.readExtraConfig?.(raw) ?? {};\n  }\n\n  function normalizeConfig(rawConfig: Record<string, unknown>): OpenAiCompatibleSpeechProviderConfig {\n    const raw = asObject(rawConfig[providerConfigKey]) ?? rawConfig;\n    return {\n      apiKey: trimToUndefined(raw.apiKey),\n      baseUrl: trimToUndefined(raw.baseUrl)\n        ? normalizeBaseUrlGeneric(String(raw.baseUrl))\n        : undefined,\n      model: trimToUndefined(raw.model ?? raw.modelId) ?? options.defaultModel,\n      voice: trimToUndefined(raw.voice ?? raw.voiceId) ?? options.defaultVoice,\n      speed: asFiniteNumber(raw.speed),\n      responseFormat: normalizeResponseFormat(options.label, options.responseFormats, raw.responseFormat),\n      ...readExtraConfig(raw),\n    };\n  }\n\n  function readProviderConfig(config: SpeechProviderConfig): OpenAiCompatibleSpeechProviderConfig {\n    return {\n      apiKey: trimToUndefined(config.apiKey),\n      baseUrl: trimToUndefined(config.baseUrl)\n        ? normalizeBaseUrlGeneric(String(config.baseUrl))\n        : undefined,\n      model: trimToUndefined(config.model ?? config.modelId) ?? options.defaultModel,\n      voice: trimToUndefined(config.voice ?? config.voiceId) ?? options.defaultVoice,\n      speed: asFiniteNumber(config.speed),\n      responseFormat:\n        normalizeResponseFormat(options.label, options.responseFormats, config.responseFormat) ??\n        options.defaultResponseFormat,\n      ...readExtraConfig(config),\n    };\n  }\n\n  function resolveApiKey(providerConfig: OpenAiCompatibleSpeechProviderConfig): string | undefined {\n    return providerConfig.apiKey ?? trimToUndefined(process.env[options.envKey]);\n  }\n\n  function resolveBaseUrl(providerConfig: OpenAiCompatibleSpeechProviderConfig): string {\n    return providerConfig.baseUrl ?? options.defaultBaseUrl.replace(/\\/+$/u, '');\n  }\n\n  return {\n    id: options.id,\n    aliases: options.aliases,\n    autoSelectOrder: options.autoSelectOrder,\n\n    resolveConfig: (ctx: SpeechProviderResolveConfigContext) => normalizeConfig(ctx.rawConfig),\n\n    isConfigured: (ctx: SpeechProviderConfiguredContext) =>\n      Boolean(resolveApiKey(readProviderConfig(ctx.providerConfig))),\n\n    parseDirectiveToken: (ctx) => parseDirectiveTokenInternal(ctx, providerConfigKey),\n\n    listVoices: async (): Promise<SpeechVoiceOption[]> =>\n      options.voices.map((voice) => ({ id: voice, name: voice })),\n\n    synthesize: async (req: SpeechSynthesisRequest): Promise<SpeechSynthesisResult> => {\n      const config = readProviderConfig(req.providerConfig);\n      const overrides = readSpeechOverrides(req.providerOverrides);\n      const apiKey = resolveApiKey(config);\n      if (!apiKey) {\n        throw new Error(options.missingApiKeyError ?? `${options.label} API key missing`);\n      }\n      const baseUrl = resolveBaseUrl(config);\n      const responseFormat = config.responseFormat ?? options.defaultResponseFormat;\n      const speed = overrides.speed ?? config.speed;\n      const model = overrides.model ?? config.model;\n      const voice = overrides.voice ?? config.voice;\n\n      const url = `${baseUrl}/audio/speech`;\n      log.debug(\n        { provider: options.id, model, voice, responseFormat, baseUrl, target: req.target },\n        `Calling ${options.label} TTS`,\n      );\n\n      // postJsonRequest already throws ProviderHttpError on non-2xx.\n      const response = await postJsonRequest(url, {\n        timeoutMs: req.timeoutMs,\n        label: options.apiErrorLabel ?? `${options.label} TTS API error`,\n        headers: {\n          Authorization: `Bearer ${apiKey}`,\n          ...options.extraHeaders,\n        },\n        body: {\n          model,\n          input: req.text,\n          voice,\n          response_format: responseFormat,\n          ...(speed === undefined ? {} : { speed }),\n          ...buildExtraJsonBodyFields(config, options.extraJsonBodyFields),\n        },\n      });\n\n      const audioBuffer = Buffer.from(await response.arrayBuffer());\n      return {\n        audioBuffer,\n        outputFormat: responseFormat,\n        fileExtension: responseFormat,\n        voiceCompatible: voiceCompatibleResponseFormats.includes(responseFormat),\n      };\n    },\n\n    /**\n     * Streaming via OpenAI's POST /audio/speech (the response body itself is\n     * the audio stream — no SSE / WebSocket needed). We do NOT call assertOk\n     * via postJsonRequest because we want to surface the body as-is for\n     * streaming consumers; instead we manually check status and convert to\n     * ProviderHttpError on failure (matches openclaw behavior).\n     */\n    synthesizeStream: async (req) => {\n      const config = readProviderConfig(req.providerConfig);\n      const overrides = readSpeechOverrides(req.providerOverrides);\n      const apiKey = resolveApiKey(config);\n      if (!apiKey) {\n        throw new Error(options.missingApiKeyError ?? `${options.label} API key missing`);\n      }\n      const baseUrl = resolveBaseUrl(config);\n      const responseFormat = config.responseFormat ?? options.defaultResponseFormat;\n      const speed = overrides.speed ?? config.speed;\n      const model = overrides.model ?? config.model;\n      const voice = overrides.voice ?? config.voice;\n\n      const response = await fetchWithTimeoutGuarded(`${baseUrl}/audio/speech`, {\n        timeoutMs: req.timeoutMs,\n        label: options.apiErrorLabel ?? `${options.label} TTS streaming`,\n        init: {\n          method: 'POST',\n          headers: {\n            Authorization: `Bearer ${apiKey}`,\n            'content-type': 'application/json',\n            accept: 'application/octet-stream',\n            ...options.extraHeaders,\n          },\n          body: JSON.stringify({\n            model,\n            input: req.text,\n            voice,\n            response_format: responseFormat,\n            stream: true,\n            ...(speed === undefined ? {} : { speed }),\n            ...buildExtraJsonBodyFields(config, options.extraJsonBodyFields),\n          }),\n        },\n      });\n\n      if (!response.ok) {\n        const detail = await response.text().catch(() => undefined);\n        throw new ProviderHttpError({\n          label: options.apiErrorLabel ?? `${options.label} TTS streaming`,\n          status: response.status,\n          detail: detail?.slice(0, 220),\n        });\n      }\n      if (!response.body) {\n        throw new Error(`${options.label} TTS streaming: empty response body`);\n      }\n\n      return {\n        audioStream: response.body,\n        outputFormat: responseFormat,\n        fileExtension: responseFormat,\n        voiceCompatible: voiceCompatibleResponseFormats.includes(responseFormat),\n      };\n    },\n  };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;aAgCqD;AAerD,MAAM,MAAM,aAAa,8BAA8B;AA4DvD,SAAS,SAAS,OAAqD;AACrE,QAAO,OAAO,UAAU,YAAY,UAAU,QAAQ,CAAC,MAAM,QAAQ,MAAM,GACtE,QACD,KAAA;;AAGN,SAAS,gBAAgB,OAAoC;AAC3D,KAAI,OAAO,UAAU,SACnB;CAEF,MAAM,UAAU,MAAM,MAAM;AAC5B,QAAO,QAAQ,SAAS,IAAI,UAAU,KAAA;;AAGxC,SAAS,eAAe,OAAoC;AAC1D,KAAI,OAAO,UAAU,YAAY,OAAO,SAAS,MAAM,CACrD,QAAO;AAET,KAAI,OAAO,UAAU,UAAU;EAC7B,MAAM,SAAS,OAAO,WAAW,MAAM;AACvC,MAAI,OAAO,SAAS,OAAO,CACzB,QAAO;;;AAMb,SAAS,WAAW,OAAuB;AACzC,QAAO,MAAM,QAAQ,iBAAiB,GAAG,CAAC,aAAa;;AAGzD,SAAS,wBAAwB,OAAe,SAA4B,OAAoC;CAC9G,MAAM,OAAO,gBAAgB,MAAM,EAAE,aAAa;AAClD,KAAI,CAAC,KACH;AAEF,KAAI,QAAQ,SAAS,KAAK,CACxB,QAAO;AAET,OAAM,IAAI,MAAM,WAAW,MAAM,0BAA0B,OAAO;;AAGpE,SAAS,yBACP,QACA,QACyB;CACzB,MAAM,OAAgC,EAAE;AACxC,MAAK,MAAM,SAAS,UAAU,EAAE,EAAE;EAChC,MAAM,QAAQ,OAAO,MAAM;AAC3B,MAAI,UAAU,KAAA,KAAa,UAAU,QAAQ,UAAU,GACrD,MAAK,MAAM,cAAc,MAAM,aAAa;;AAGhD,QAAO;;AAGT,SAAS,oBAAoB,WAI3B;AACA,KAAI,CAAC,UACH,QAAO,EAAE;AAEX,QAAO;EACL,OAAO,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC5D,OAAO,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC5D,OAAO,eAAe,UAAU,MAAM;EACvC;;AAGH,SAAS,4BACP,KACA,mBACiC;CACjC,MAAM,qBAAqB,WAAW,kBAAkB;AACxD,SAAQ,IAAI,KAAZ;EACE,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK,GAAG,kBAAkB;EAC1B,KAAK,GAAG,mBAAmB;AACzB,OAAI,CAAC,IAAI,OAAO,WACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK,GAAG,kBAAkB;EAC1B,KAAK,GAAG,mBAAmB;AACzB,OAAI,CAAC,IAAI,OAAO,aACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,KAAK,SAAS;GACZ,MAAM,QAAQ,eAAe,IAAI,MAAM;AACvC,OAAI,UAAU,KAAA,EACZ,QAAO;IAAE,SAAS;IAAM,UAAU,CAAC,wBAAwB,IAAI,QAAQ;IAAE;AAE3E,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO;IAAE;;EAEhD,QACE,QAAO,EAAE,SAAS,OAAO;;;AAI/B,SAAgB,qCACd,SACsB;CACtB,MAAM,oBAAoB,QAAQ;CAClC,MAAM,iCAAiC,QAAQ,kCAAkC,CAAC,OAAO;CAEzF,SAAS,gBAAgB,KAAmE;AAC1F,SAAO,QAAQ,kBAAkB,IAAI,IAAI,EAAE;;CAG7C,SAAS,gBAAgB,WAA0E;EACjG,MAAM,MAAM,SAAS,UAAU,mBAAmB,IAAI;AACtD,SAAO;GACL,QAAQ,gBAAgB,IAAI,OAAO;GACnC,SAAS,gBAAgB,IAAI,QAAQ,GACjCA,iBAAwB,OAAO,IAAI,QAAQ,CAAC,GAC5C,KAAA;GACJ,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI,QAAQ;GAC5D,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI,QAAQ;GAC5D,OAAO,eAAe,IAAI,MAAM;GAChC,gBAAgB,wBAAwB,QAAQ,OAAO,QAAQ,iBAAiB,IAAI,eAAe;GACnG,GAAG,gBAAgB,IAAI;GACxB;;CAGH,SAAS,mBAAmB,QAAoE;AAC9F,SAAO;GACL,QAAQ,gBAAgB,OAAO,OAAO;GACtC,SAAS,gBAAgB,OAAO,QAAQ,GACpCA,iBAAwB,OAAO,OAAO,QAAQ,CAAC,GAC/C,KAAA;GACJ,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI,QAAQ;GAClE,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI,QAAQ;GAClE,OAAO,eAAe,OAAO,MAAM;GACnC,gBACE,wBAAwB,QAAQ,OAAO,QAAQ,iBAAiB,OAAO,eAAe,IACtF,QAAQ;GACV,GAAG,gBAAgB,OAAO;GAC3B;;CAGH,SAAS,cAAc,gBAA0E;AAC/F,SAAO,eAAe,UAAU,gBAAgB,QAAQ,IAAI,QAAQ,QAAQ;;CAG9E,SAAS,eAAe,gBAA8D;AACpF,SAAO,eAAe,WAAW,QAAQ,eAAe,QAAQ,SAAS,GAAG;;AAG9E,QAAO;EACL,IAAI,QAAQ;EACZ,SAAS,QAAQ;EACjB,iBAAiB,QAAQ;EAEzB,gBAAgB,QAA4C,gBAAgB,IAAI,UAAU;EAE1F,eAAe,QACb,QAAQ,cAAc,mBAAmB,IAAI,eAAe,CAAC,CAAC;EAEhE,sBAAsB,QAAQ,4BAA4B,KAAK,kBAAkB;EAEjF,YAAY,YACV,QAAQ,OAAO,KAAK,WAAW;GAAE,IAAI;GAAO,MAAM;GAAO,EAAE;EAE7D,YAAY,OAAO,QAAgE;GACjF,MAAM,SAAS,mBAAmB,IAAI,eAAe;GACrD,MAAM,YAAY,oBAAoB,IAAI,kBAAkB;GAC5D,MAAM,SAAS,cAAc,OAAO;AACpC,OAAI,CAAC,OACH,OAAM,IAAI,MAAM,QAAQ,sBAAsB,GAAG,QAAQ,MAAM,kBAAkB;GAEnF,MAAM,UAAU,eAAe,OAAO;GACtC,MAAM,iBAAiB,OAAO,kBAAkB,QAAQ;GACxD,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GAExC,MAAM,MAAM,GAAG,QAAQ;AACvB,OAAI,MACF;IAAE,UAAU,QAAQ;IAAI;IAAO;IAAO;IAAgB;IAAS,QAAQ,IAAI;IAAQ,EACnF,WAAW,QAAQ,MAAM,MAC1B;GAGD,MAAM,WAAW,MAAM,gBAAgB,KAAK;IAC1C,WAAW,IAAI;IACf,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;IACjD,SAAS;KACP,eAAe,UAAU;KACzB,GAAG,QAAQ;KACZ;IACD,MAAM;KACJ;KACA,OAAO,IAAI;KACX;KACA,iBAAiB;KACjB,GAAI,UAAU,KAAA,IAAY,EAAE,GAAG,EAAE,OAAO;KACxC,GAAG,yBAAyB,QAAQ,QAAQ,oBAAoB;KACjE;IACF,CAAC;AAGF,UAAO;IACL,aAFkB,OAAO,KAAK,MAAM,SAAS,aAAa,CAE/C;IACX,cAAc;IACd,eAAe;IACf,iBAAiB,+BAA+B,SAAS,eAAe;IACzE;;;;;;;;;EAUH,kBAAkB,OAAO,QAAQ;GAC/B,MAAM,SAAS,mBAAmB,IAAI,eAAe;GACrD,MAAM,YAAY,oBAAoB,IAAI,kBAAkB;GAC5D,MAAM,SAAS,cAAc,OAAO;AACpC,OAAI,CAAC,OACH,OAAM,IAAI,MAAM,QAAQ,sBAAsB,GAAG,QAAQ,MAAM,kBAAkB;GAEnF,MAAM,UAAU,eAAe,OAAO;GACtC,MAAM,iBAAiB,OAAO,kBAAkB,QAAQ;GACxD,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GAExC,MAAM,WAAW,MAAM,wBAAwB,GAAG,QAAQ,gBAAgB;IACxE,WAAW,IAAI;IACf,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;IACjD,MAAM;KACJ,QAAQ;KACR,SAAS;MACP,eAAe,UAAU;MACzB,gBAAgB;MAChB,QAAQ;MACR,GAAG,QAAQ;MACZ;KACD,MAAM,KAAK,UAAU;MACnB;MACA,OAAO,IAAI;MACX;MACA,iBAAiB;MACjB,QAAQ;MACR,GAAI,UAAU,KAAA,IAAY,EAAE,GAAG,EAAE,OAAO;MACxC,GAAG,yBAAyB,QAAQ,QAAQ,oBAAoB;MACjE,CAAC;KACH;IACF,CAAC;AAEF,OAAI,CAAC,SAAS,IAAI;IAChB,MAAM,SAAS,MAAM,SAAS,MAAM,CAAC,YAAY,KAAA,EAAU;AAC3D,UAAM,IAAI,kBAAkB;KAC1B,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;KACjD,QAAQ,SAAS;KACjB,QAAQ,QAAQ,MAAM,GAAG,IAAI;KAC9B,CAAC;;AAEJ,OAAI,CAAC,SAAS,KACZ,OAAM,IAAI,MAAM,GAAG,QAAQ,MAAM,qCAAqC;AAGxE,UAAO;IACL,aAAa,SAAS;IACtB,cAAc;IACd,eAAe;IACf,iBAAiB,+BAA+B,SAAS,eAAe;IACzE;;EAEJ"}
+{"version":3,"file":"openai-compatible-speech.js","names":["normalizeBaseUrlGeneric"],"sources":["../../../../src/voice/tts/openai-compatible-speech.ts"],"sourcesContent":["/**\n * Factory: createOpenAiCompatibleSpeechProvider\n *\n * Most TTS providers (OpenAI, MiniMax, ElevenLabs, OpenRouter, ...) speak the\n * OpenAI `/audio/speech` POST shape. This factory eliminates the boilerplate so\n * a new compatible vendor is ~30 lines of declaration instead of a full provider\n * implementation.\n *\n * Ported from openclaw/src/tts/openai-compatible-speech-provider.ts (commit\n * baseline 2026-05-08), with the following INTENTIONAL OMISSIONS per\n * docs/voice-rearchitecture.md §15.3:\n *   - resolveTalkConfig / resolveTalkOverrides (talk-mode is excluded)\n *   - Persona resolution context (persona system is excluded)\n *   - Custom dispatcherPolicy (xopc uses Node 22 native fetch — see ssrf-guard.ts)\n *\n * DECISION: We use `Record<string, unknown>` for the extra-config slot rather\n * than openclaw's generic `<ExtraConfig extends Record<string, unknown>>`. The\n * generic adds significant TypeScript complexity but no runtime safety; we\n * declare per-provider config types in each provider file's own normalizer.\n *\n * DECISION: `voiceCompatibleResponseFormats` defaults to `['opus']` if omitted,\n * because Telegram voice notes require opus/ogg. Providers that natively support\n * opus output (OpenAI, ElevenLabs) can rely on this; others (MiniMax mp3-only)\n * pass an empty array and downstream ffmpeg-compresses to opus.\n */\n\nimport {\n  ProviderHttpError,\n  fetchWithTimeoutGuarded,\n  normalizeBaseUrl as normalizeBaseUrlGeneric,\n  postJsonRequest,\n} from '../../media-shared/http/index.js';\nimport { createLogger } from '../../utils/logger.js';\n\nimport type {\n  SpeechDirectiveTokenParseContext,\n  SpeechDirectiveTokenParseResult,\n  SpeechProviderConfig,\n  SpeechProviderConfiguredContext,\n  SpeechProviderOverrides,\n  SpeechProviderPlugin,\n  SpeechProviderResolveConfigContext,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechVoiceOption,\n} from './speech-provider-types.js';\n\nconst log = createLogger('SpeechProvider:OpenAiCompat');\n\ninterface OpenAiCompatibleBaseConfig {\n  apiKey?: string;\n  baseUrl?: string;\n  model: string;\n  voice: string;\n  speed?: number;\n  responseFormat?: string;\n}\n\nexport type OpenAiCompatibleSpeechProviderConfig = OpenAiCompatibleBaseConfig &\n  Record<string, unknown>;\n\nexport interface OpenAiCompatibleSpeechProviderExtraJsonBodyField {\n  /** Key to read from normalized config. */\n  configKey: string;\n  /** Key to send in the HTTP body. Defaults to `configKey`. */\n  requestKey?: string;\n}\n\nexport interface OpenAiCompatibleSpeechProviderOptions {\n  /** Canonical provider id (e.g. \"openai\", \"minimax\"). */\n  id: string;\n  /** Optional aliases for provider lookup. */\n  aliases?: readonly string[];\n  /** Human-readable label for error messages and logs (e.g. \"OpenAI\", \"MiniMax\"). */\n  label: string;\n  /** Sort key for UI auto-fallback ordering. Lower = higher priority. */\n  autoSelectOrder: number;\n  /** Allowed model ids. Used by listVoices and validation. */\n  models: readonly string[];\n  /** Allowed voice ids. */\n  voices: readonly string[];\n  /** Default model when config does not specify. */\n  defaultModel: string;\n  /** Default voice. */\n  defaultVoice: string;\n  /** Default base URL (e.g. \"https://api.openai.com/v1\"). */\n  defaultBaseUrl: string;\n  /** Env var name for the api key (e.g. \"OPENAI_API_KEY\"). */\n  envKey: string;\n  /** Allowed response formats. */\n  responseFormats: readonly string[];\n  /** Default response format (e.g. \"opus\" for OpenAI, \"mp3\" for MiniMax). */\n  defaultResponseFormat: string;\n  /** Response formats that produce a voice-note-compatible buffer (no ffmpeg needed). */\n  voiceCompatibleResponseFormats?: readonly string[];\n  /** Extra static headers (e.g. `{ \"OpenAI-Beta\": \"...\" }`). Auth header is added automatically. */\n  extraHeaders?: Record<string, string>;\n  /** Read provider-specific extra config keys (e.g. MiniMax's `groupId`). */\n  readExtraConfig?: (raw: Record<string, unknown> | undefined) => Record<string, unknown>;\n  /** Provider-specific extra body fields to copy from config into request JSON. */\n  extraJsonBodyFields?: readonly OpenAiCompatibleSpeechProviderExtraJsonBodyField[];\n  /** Override the API error label (defaults to `${label} TTS API error`). */\n  apiErrorLabel?: string;\n  /** Override the missing-key error message. */\n  missingApiKeyError?: string;\n}\n\nfunction asObject(value: unknown): Record<string, unknown> | undefined {\n  return typeof value === 'object' && value !== null && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined;\n}\n\nfunction trimToUndefined(value: unknown): string | undefined {\n  if (typeof value !== 'string') {\n    return undefined;\n  }\n  const trimmed = value.trim();\n  return trimmed.length > 0 ? trimmed : undefined;\n}\n\nfunction asFiniteNumber(value: unknown): number | undefined {\n  if (typeof value === 'number' && Number.isFinite(value)) {\n    return value;\n  }\n  if (typeof value === 'string') {\n    const parsed = Number.parseFloat(value);\n    if (Number.isFinite(parsed)) {\n      return parsed;\n    }\n  }\n  return undefined;\n}\n\nfunction compactKey(value: string): string {\n  return value.replace(/[^a-z0-9]+/giu, '').toLowerCase();\n}\n\nfunction normalizeResponseFormat(label: string, formats: readonly string[], value: unknown): string | undefined {\n  const next = trimToUndefined(value)?.toLowerCase();\n  if (!next) {\n    return undefined;\n  }\n  if (formats.includes(next)) {\n    return next;\n  }\n  throw new Error(`Invalid ${label} speech responseFormat: ${next}`);\n}\n\nfunction buildExtraJsonBodyFields(\n  config: OpenAiCompatibleSpeechProviderConfig,\n  fields: readonly OpenAiCompatibleSpeechProviderExtraJsonBodyField[] | undefined,\n): Record<string, unknown> {\n  const body: Record<string, unknown> = {};\n  for (const field of fields ?? []) {\n    const value = config[field.configKey];\n    if (value !== undefined && value !== null && value !== '') {\n      body[field.requestKey ?? field.configKey] = value;\n    }\n  }\n  return body;\n}\n\nfunction readSpeechOverrides(overrides: SpeechProviderOverrides | undefined): {\n  model?: string;\n  voice?: string;\n  speed?: number;\n} {\n  if (!overrides) {\n    return {};\n  }\n  return {\n    model: trimToUndefined(overrides.model ?? overrides.modelId),\n    voice: trimToUndefined(overrides.voice ?? overrides.voiceId),\n    speed: asFiniteNumber(overrides.speed),\n  };\n}\n\nfunction parseDirectiveTokenInternal(\n  ctx: SpeechDirectiveTokenParseContext,\n  providerConfigKey: string,\n): SpeechDirectiveTokenParseResult {\n  const compactProviderKey = compactKey(providerConfigKey);\n  switch (ctx.key) {\n    case 'voice':\n    case 'voice_id':\n    case 'voiceid':\n    case `${providerConfigKey}_voice`:\n    case `${compactProviderKey}voice`:\n      if (!ctx.policy.allowVoice) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { voice: ctx.value } };\n    case 'model':\n    case 'model_id':\n    case 'modelid':\n    case `${providerConfigKey}_model`:\n    case `${compactProviderKey}model`:\n      if (!ctx.policy.allowModelId) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { model: ctx.value } };\n    case 'speed': {\n      const speed = asFiniteNumber(ctx.value);\n      if (speed === undefined) {\n        return { handled: true, warnings: [`Invalid speed value: ${ctx.value}`] };\n      }\n      return { handled: true, overrides: { speed } };\n    }\n    default:\n      return { handled: false };\n  }\n}\n\nexport function createOpenAiCompatibleSpeechProvider(\n  options: OpenAiCompatibleSpeechProviderOptions,\n): SpeechProviderPlugin {\n  const providerConfigKey = options.id;\n  const voiceCompatibleResponseFormats = options.voiceCompatibleResponseFormats ?? ['opus'];\n\n  function readExtraConfig(raw: Record<string, unknown> | undefined): Record<string, unknown> {\n    return options.readExtraConfig?.(raw) ?? {};\n  }\n\n  function normalizeConfig(rawConfig: Record<string, unknown>): OpenAiCompatibleSpeechProviderConfig {\n    const raw = asObject(rawConfig[providerConfigKey]) ?? rawConfig;\n    return {\n      apiKey: trimToUndefined(raw.apiKey),\n      baseUrl: trimToUndefined(raw.baseUrl)\n        ? normalizeBaseUrlGeneric(String(raw.baseUrl))\n        : undefined,\n      model: trimToUndefined(raw.model ?? raw.modelId) ?? options.defaultModel,\n      voice: trimToUndefined(raw.voice ?? raw.voiceId) ?? options.defaultVoice,\n      speed: asFiniteNumber(raw.speed),\n      responseFormat: normalizeResponseFormat(options.label, options.responseFormats, raw.responseFormat),\n      ...readExtraConfig(raw),\n    };\n  }\n\n  function readProviderConfig(config: SpeechProviderConfig): OpenAiCompatibleSpeechProviderConfig {\n    return {\n      apiKey: trimToUndefined(config.apiKey),\n      baseUrl: trimToUndefined(config.baseUrl)\n        ? normalizeBaseUrlGeneric(String(config.baseUrl))\n        : undefined,\n      model: trimToUndefined(config.model ?? config.modelId) ?? options.defaultModel,\n      voice: trimToUndefined(config.voice ?? config.voiceId) ?? options.defaultVoice,\n      speed: asFiniteNumber(config.speed),\n      responseFormat:\n        normalizeResponseFormat(options.label, options.responseFormats, config.responseFormat) ??\n        options.defaultResponseFormat,\n      ...readExtraConfig(config),\n    };\n  }\n\n  function resolveApiKey(providerConfig: OpenAiCompatibleSpeechProviderConfig): string | undefined {\n    return providerConfig.apiKey ?? trimToUndefined(process.env[options.envKey]);\n  }\n\n  function resolveBaseUrl(providerConfig: OpenAiCompatibleSpeechProviderConfig): string {\n    return providerConfig.baseUrl ?? options.defaultBaseUrl.replace(/\\/+$/u, '');\n  }\n\n  return {\n    id: options.id,\n    aliases: options.aliases,\n    autoSelectOrder: options.autoSelectOrder,\n\n    resolveConfig: (ctx: SpeechProviderResolveConfigContext) => normalizeConfig(ctx.rawConfig),\n\n    isConfigured: (ctx: SpeechProviderConfiguredContext) =>\n      Boolean(resolveApiKey(readProviderConfig(ctx.providerConfig))),\n\n    parseDirectiveToken: (ctx) => parseDirectiveTokenInternal(ctx, providerConfigKey),\n\n    listVoices: async (): Promise<SpeechVoiceOption[]> =>\n      options.voices.map((voice) => ({ id: voice, name: voice })),\n\n    synthesize: async (req: SpeechSynthesisRequest): Promise<SpeechSynthesisResult> => {\n      const config = readProviderConfig(req.providerConfig);\n      const overrides = readSpeechOverrides(req.providerOverrides);\n      const apiKey = resolveApiKey(config);\n      if (!apiKey) {\n        throw new Error(options.missingApiKeyError ?? `${options.label} API key missing`);\n      }\n      const baseUrl = resolveBaseUrl(config);\n      const responseFormat = config.responseFormat ?? options.defaultResponseFormat;\n      const speed = overrides.speed ?? config.speed;\n      const model = overrides.model ?? config.model;\n      const voice = overrides.voice ?? config.voice;\n\n      const url = `${baseUrl}/audio/speech`;\n      log.debug(\n        { provider: options.id, model, voice, responseFormat, baseUrl, target: req.target },\n        `Calling ${options.label} TTS`,\n      );\n\n      // postJsonRequest already throws ProviderHttpError on non-2xx.\n      const response = await postJsonRequest(url, {\n        timeoutMs: req.timeoutMs,\n        label: options.apiErrorLabel ?? `${options.label} TTS API error`,\n        headers: {\n          Authorization: `Bearer ${apiKey}`,\n          ...options.extraHeaders,\n        },\n        body: {\n          model,\n          input: req.text,\n          voice,\n          response_format: responseFormat,\n          ...(speed === undefined ? {} : { speed }),\n          ...buildExtraJsonBodyFields(config, options.extraJsonBodyFields),\n        },\n      });\n\n      const audioBuffer = Buffer.from(await response.arrayBuffer());\n      return {\n        audioBuffer,\n        outputFormat: responseFormat,\n        fileExtension: responseFormat,\n        voiceCompatible: voiceCompatibleResponseFormats.includes(responseFormat),\n      };\n    },\n\n    /**\n     * Streaming via OpenAI's POST /audio/speech (the response body itself is\n     * the audio stream — no SSE / WebSocket needed). We do NOT call assertOk\n     * via postJsonRequest because we want to surface the body as-is for\n     * streaming consumers; instead we manually check status and convert to\n     * ProviderHttpError on failure (matches openclaw behavior).\n     */\n    synthesizeStream: async (req) => {\n      const config = readProviderConfig(req.providerConfig);\n      const overrides = readSpeechOverrides(req.providerOverrides);\n      const apiKey = resolveApiKey(config);\n      if (!apiKey) {\n        throw new Error(options.missingApiKeyError ?? `${options.label} API key missing`);\n      }\n      const baseUrl = resolveBaseUrl(config);\n      const responseFormat = config.responseFormat ?? options.defaultResponseFormat;\n      const speed = overrides.speed ?? config.speed;\n      const model = overrides.model ?? config.model;\n      const voice = overrides.voice ?? config.voice;\n\n      const response = await fetchWithTimeoutGuarded(`${baseUrl}/audio/speech`, {\n        timeoutMs: req.timeoutMs,\n        label: options.apiErrorLabel ?? `${options.label} TTS streaming`,\n        init: {\n          method: 'POST',\n          headers: {\n            Authorization: `Bearer ${apiKey}`,\n            'content-type': 'application/json',\n            accept: 'application/octet-stream',\n            ...options.extraHeaders,\n          },\n          body: JSON.stringify({\n            model,\n            input: req.text,\n            voice,\n            response_format: responseFormat,\n            stream: true,\n            ...(speed === undefined ? {} : { speed }),\n            ...buildExtraJsonBodyFields(config, options.extraJsonBodyFields),\n          }),\n        },\n      });\n\n      if (!response.ok) {\n        const detail = await response.text().catch(() => undefined);\n        throw new ProviderHttpError({\n          label: options.apiErrorLabel ?? `${options.label} TTS streaming`,\n          status: response.status,\n          detail: detail?.slice(0, 220),\n        });\n      }\n      if (!response.body) {\n        throw new Error(`${options.label} TTS streaming: empty response body`);\n      }\n\n      return {\n        audioStream: response.body,\n        outputFormat: responseFormat,\n        fileExtension: responseFormat,\n        voiceCompatible: voiceCompatibleResponseFormats.includes(responseFormat),\n      };\n    },\n  };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;aAgCqD;AAerD,MAAM,MAAM,aAAa,8BAA8B;AA4DvD,SAAS,SAAS,OAAqD;AACrE,QAAO,OAAO,UAAU,YAAY,UAAU,QAAQ,CAAC,MAAM,QAAQ,MAAM,GACtE,QACD,KAAA;;AAGN,SAAS,gBAAgB,OAAoC;AAC3D,KAAI,OAAO,UAAU,SACnB;CAEF,MAAM,UAAU,MAAM,MAAM;AAC5B,QAAO,QAAQ,SAAS,IAAI,UAAU,KAAA;;AAGxC,SAAS,eAAe,OAAoC;AAC1D,KAAI,OAAO,UAAU,YAAY,OAAO,SAAS,MAAM,CACrD,QAAO;AAET,KAAI,OAAO,UAAU,UAAU;EAC7B,MAAM,SAAS,OAAO,WAAW,MAAM;AACvC,MAAI,OAAO,SAAS,OAAO,CACzB,QAAO;;;AAMb,SAAS,WAAW,OAAuB;AACzC,QAAO,MAAM,QAAQ,iBAAiB,GAAG,CAAC,aAAa;;AAGzD,SAAS,wBAAwB,OAAe,SAA4B,OAAoC;CAC9G,MAAM,OAAO,gBAAgB,MAAM,EAAE,aAAa;AAClD,KAAI,CAAC,KACH;AAEF,KAAI,QAAQ,SAAS,KAAK,CACxB,QAAO;AAET,OAAM,IAAI,MAAM,WAAW,MAAM,0BAA0B,OAAO;;AAGpE,SAAS,yBACP,QACA,QACyB;CACzB,MAAM,OAAgC,EAAE;AACxC,MAAK,MAAM,SAAS,UAAU,EAAE,EAAE;EAChC,MAAM,QAAQ,OAAO,MAAM;AAC3B,MAAI,UAAU,KAAA,KAAa,UAAU,QAAQ,UAAU,GACrD,MAAK,MAAM,cAAc,MAAM,aAAa;;AAGhD,QAAO;;AAGT,SAAS,oBAAoB,WAI3B;AACA,KAAI,CAAC,UACH,QAAO,EAAE;AAEX,QAAO;EACL,OAAO,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC5D,OAAO,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC5D,OAAO,eAAe,UAAU,MAAM;EACvC;;AAGH,SAAS,4BACP,KACA,mBACiC;CACjC,MAAM,qBAAqB,WAAW,kBAAkB;AACxD,SAAQ,IAAI,KAAZ;EACE,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK,GAAG,kBAAkB;EAC1B,KAAK,GAAG,mBAAmB;AACzB,OAAI,CAAC,IAAI,OAAO,WACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK,GAAG,kBAAkB;EAC1B,KAAK,GAAG,mBAAmB;AACzB,OAAI,CAAC,IAAI,OAAO,aACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,KAAK,SAAS;GACZ,MAAM,QAAQ,eAAe,IAAI,MAAM;AACvC,OAAI,UAAU,KAAA,EACZ,QAAO;IAAE,SAAS;IAAM,UAAU,CAAC,wBAAwB,IAAI,QAAQ;IAAE;AAE3E,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO;IAAE;;EAEhD,QACE,QAAO,EAAE,SAAS,OAAO;;;AAI/B,SAAgB,qCACd,SACsB;CACtB,MAAM,oBAAoB,QAAQ;CAClC,MAAM,iCAAiC,QAAQ,kCAAkC,CAAC,OAAO;CAEzF,SAAS,gBAAgB,KAAmE;AAC1F,SAAO,QAAQ,kBAAkB,IAAI,IAAI,EAAE;;CAG7C,SAAS,gBAAgB,WAA0E;EACjG,MAAM,MAAM,SAAS,UAAU,mBAAmB,IAAI;AACtD,SAAO;GACL,QAAQ,gBAAgB,IAAI,OAAO;GACnC,SAAS,gBAAgB,IAAI,QAAQ,GACjCA,iBAAwB,OAAO,IAAI,QAAQ,CAAC,GAC5C,KAAA;GACJ,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI,QAAQ;GAC5D,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI,QAAQ;GAC5D,OAAO,eAAe,IAAI,MAAM;GAChC,gBAAgB,wBAAwB,QAAQ,OAAO,QAAQ,iBAAiB,IAAI,eAAe;GACnG,GAAG,gBAAgB,IAAI;GACxB;;CAGH,SAAS,mBAAmB,QAAoE;AAC9F,SAAO;GACL,QAAQ,gBAAgB,OAAO,OAAO;GACtC,SAAS,gBAAgB,OAAO,QAAQ,GACpCA,iBAAwB,OAAO,OAAO,QAAQ,CAAC,GAC/C,KAAA;GACJ,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI,QAAQ;GAClE,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI,QAAQ;GAClE,OAAO,eAAe,OAAO,MAAM;GACnC,gBACE,wBAAwB,QAAQ,OAAO,QAAQ,iBAAiB,OAAO,eAAe,IACtF,QAAQ;GACV,GAAG,gBAAgB,OAAO;GAC3B;;CAGH,SAAS,cAAc,gBAA0E;AAC/F,SAAO,eAAe,UAAU,gBAAgB,QAAQ,IAAI,QAAQ,QAAQ;;CAG9E,SAAS,eAAe,gBAA8D;AACpF,SAAO,eAAe,WAAW,QAAQ,eAAe,QAAQ,SAAS,GAAG;;AAG9E,QAAO;EACL,IAAI,QAAQ;EACZ,SAAS,QAAQ;EACjB,iBAAiB,QAAQ;EAEzB,gBAAgB,QAA4C,gBAAgB,IAAI,UAAU;EAE1F,eAAe,QACb,QAAQ,cAAc,mBAAmB,IAAI,eAAe,CAAC,CAAC;EAEhE,sBAAsB,QAAQ,4BAA4B,KAAK,kBAAkB;EAEjF,YAAY,YACV,QAAQ,OAAO,KAAK,WAAW;GAAE,IAAI;GAAO,MAAM;GAAO,EAAE;EAE7D,YAAY,OAAO,QAAgE;GACjF,MAAM,SAAS,mBAAmB,IAAI,eAAe;GACrD,MAAM,YAAY,oBAAoB,IAAI,kBAAkB;GAC5D,MAAM,SAAS,cAAc,OAAO;AACpC,OAAI,CAAC,OACH,OAAM,IAAI,MAAM,QAAQ,sBAAsB,GAAG,QAAQ,MAAM,kBAAkB;GAEnF,MAAM,UAAU,eAAe,OAAO;GACtC,MAAM,iBAAiB,OAAO,kBAAkB,QAAQ;GACxD,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GAExC,MAAM,MAAM,GAAG,QAAQ;AACvB,OAAI,MACF;IAAE,UAAU,QAAQ;IAAI;IAAO;IAAO;IAAgB;IAAS,QAAQ,IAAI;IAAQ,EACnF,WAAW,QAAQ,MAAM,MAC1B;GAGD,MAAM,WAAW,MAAM,gBAAgB,KAAK;IAC1C,WAAW,IAAI;IACf,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;IACjD,SAAS;KACP,eAAe,UAAU;KACzB,GAAG,QAAQ;KACZ;IACD,MAAM;KACJ;KACA,OAAO,IAAI;KACX;KACA,iBAAiB;KACjB,GAAI,UAAU,KAAA,IAAY,EAAE,GAAG,EAAE,OAAO;KACxC,GAAG,yBAAyB,QAAQ,QAAQ,oBAAoB;KACjE;IACF,CAAC;AAGF,UAAO;IACL,aAFkB,OAAO,KAAK,MAAM,SAAS,aAAa,CAE/C;IACX,cAAc;IACd,eAAe;IACf,iBAAiB,+BAA+B,SAAS,eAAe;IACzE;;;;;;;;;EAUH,kBAAkB,OAAO,QAAQ;GAC/B,MAAM,SAAS,mBAAmB,IAAI,eAAe;GACrD,MAAM,YAAY,oBAAoB,IAAI,kBAAkB;GAC5D,MAAM,SAAS,cAAc,OAAO;AACpC,OAAI,CAAC,OACH,OAAM,IAAI,MAAM,QAAQ,sBAAsB,GAAG,QAAQ,MAAM,kBAAkB;GAEnF,MAAM,UAAU,eAAe,OAAO;GACtC,MAAM,iBAAiB,OAAO,kBAAkB,QAAQ;GACxD,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GACxC,MAAM,QAAQ,UAAU,SAAS,OAAO;GAExC,MAAM,WAAW,MAAM,wBAAwB,GAAG,QAAQ,gBAAgB;IACxE,WAAW,IAAI;IACf,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;IACjD,MAAM;KACJ,QAAQ;KACR,SAAS;MACP,eAAe,UAAU;MACzB,gBAAgB;MAChB,QAAQ;MACR,GAAG,QAAQ;MACZ;KACD,MAAM,KAAK,UAAU;MACnB;MACA,OAAO,IAAI;MACX;MACA,iBAAiB;MACjB,QAAQ;MACR,GAAI,UAAU,KAAA,IAAY,EAAE,GAAG,EAAE,OAAO;MACxC,GAAG,yBAAyB,QAAQ,QAAQ,oBAAoB;MACjE,CAAC;KACH;IACF,CAAC;AAEF,OAAI,CAAC,SAAS,IAAI;IAChB,MAAM,SAAS,MAAM,SAAS,MAAM,CAAC,YAAY,KAAA,EAAU;AAC3D,UAAM,IAAI,kBAAkB;KAC1B,OAAO,QAAQ,iBAAiB,GAAG,QAAQ,MAAM;KACjD,QAAQ,SAAS;KACjB,QAAQ,QAAQ,MAAM,GAAG,IAAI;KAC9B,CAAC;;AAEJ,OAAI,CAAC,SAAS,KACZ,OAAM,IAAI,MAAM,GAAG,QAAQ,MAAM,qCAAqC;AAGxE,UAAO;IACL,aAAa,SAAS;IACtB,cAAc;IACd,eAAe;IACf,iBAAiB,+BAA+B,SAAS,eAAe;IACzE;;EAEJ"}

package/dist/src/voice/tts/providers/edge-speech.js

@@ -1,9 +1,9 @@
 import { createLogger } from "../../../utils/logger/index.js";
 import { init_logger } from "../../../utils/logger.js";
 import { registerSpeechProvider } from "../speech-registry.js";
-import { tmpdir } from "node:os";
+import { mkdtempSync, readFileSync, rmSync, statSync } from "node:fs";
 import path from "node:path";
-import { mkdtempSync, readFileSync, rmSync } from "node:fs";
+import { tmpdir } from "node:os";
 //#region src/voice/tts/providers/edge-speech.ts
 /**
 * Edge TTS provider — wraps the `node-edge-tts` package. Requires no API key
@@ -94,6 +94,15 @@ function parseDirectiveTokenInternal(ctx) {
 		default: return { handled: false };
 	}
 }
+async function waitForNonEmptyFile(filePath, timeoutMs) {
+	const deadline = Date.now() + Math.min(Math.max(timeoutMs, 1e3), 5e3);
+	while (Date.now() < deadline) {
+		try {
+			if (statSync(filePath).size > 0) return;
+		} catch {}
+		await new Promise((resolve) => setTimeout(resolve, 50));
+	}
+}
 async function synthesizeToBuffer(text, config, voiceOverride, timeoutMs) {
 	const { EdgeTTS } = await import("node-edge-tts");
 	const voice = voiceOverride ?? config.voice;
@@ -116,8 +125,11 @@ async function synthesizeToBuffer(text, config, voiceOverride, timeoutMs) {
 			volume: config.volume,
 			timeout: timeoutMs
 		}).ttsPromise(text, outputPath);
+		await waitForNonEmptyFile(outputPath, timeoutMs);
+		const buffer = readFileSync(outputPath);
+		if (buffer.length === 0) throw new Error(`Edge TTS produced an empty ${format} file for voice ${voice}`);
 		return {
-			buffer: readFileSync(outputPath),
+			buffer,
 			outputFormat: format,
 			ext
 		};

package/dist/src/voice/tts/providers/edge-speech.js.map

@@ -1 +1 @@
-{"version":3,"file":"edge-speech.js","names":[],"sources":["../../../../../src/voice/tts/providers/edge-speech.ts"],"sourcesContent":["/**\n * Edge TTS provider — wraps the `node-edge-tts` package. Requires no API key\n * (Microsoft's Edge browser endpoint is unauthenticated).\n *\n * Implementation notes:\n *   - Temp-file dance (mkdtemp → ttsPromise → readFileSync → rm) is required\n *     because `node-edge-tts` does not expose a stream-to-buffer API.\n *   - `synthesizeStream` is implemented as \"synthesize + wrap as single-chunk\n *     stream\" so callers see uniform behavior. This is not a true network\n *     stream; it's buffered-then-wrapped.\n *   - `inferEdgeExtension` derives the file extension from Edge's outputFormat\n *     string.\n */\n\nimport { mkdtempSync, readFileSync, rmSync } from 'node:fs';\nimport { tmpdir } from 'node:os';\nimport path from 'node:path';\n\nimport { createLogger } from '../../../utils/logger.js';\nimport { registerSpeechProvider } from '../speech-registry.js';\nimport type {\n  SpeechDirectiveTokenParseContext,\n  SpeechDirectiveTokenParseResult,\n  SpeechProviderConfig,\n  SpeechProviderConfiguredContext,\n  SpeechProviderPlugin,\n  SpeechProviderResolveConfigContext,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechSynthesisStreamRequest,\n  SpeechSynthesisStreamResult,\n} from '../speech-provider-types.js';\n\nconst log = createLogger('SpeechProvider:Edge');\n\nconst DEFAULT_VOICE = 'en-US-MichelleNeural';\nconst DEFAULT_LANG = 'en-US';\nconst DEFAULT_OUTPUT_FORMAT = 'audio-24khz-48kbitrate-mono-mp3';\n\ninterface EdgeTtsConfig extends Record<string, unknown> {\n  voice: string;\n  lang: string;\n  outputFormat: string;\n  pitch?: string;\n  rate?: string;\n  volume?: string;\n  proxy?: string;\n  enabled: boolean;\n}\n\nfunction trimToUndefined(value: unknown): string | undefined {\n  if (typeof value !== 'string') {\n    return undefined;\n  }\n  const trimmed = value.trim();\n  return trimmed.length > 0 ? trimmed : undefined;\n}\n\nfunction asObject(value: unknown): Record<string, unknown> | undefined {\n  return typeof value === 'object' && value !== null && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined;\n}\n\nfunction inferEdgeExtension(outputFormat: string): { ext: string; format: string } {\n  const normalized = outputFormat.toLowerCase();\n  if (normalized.includes('webm')) return { ext: 'webm', format: 'webm' };\n  if (normalized.includes('opus')) return { ext: 'opus', format: 'opus' };\n  if (normalized.includes('ogg')) return { ext: 'ogg', format: 'ogg' };\n  if (normalized.includes('wav') || normalized.includes('riff') || normalized.includes('pcm')) {\n    return { ext: 'wav', format: 'wav' };\n  }\n  return { ext: 'mp3', format: 'mp3' };\n}\n\nfunction normalizeConfig(rawConfig: Record<string, unknown>): EdgeTtsConfig {\n  const raw = asObject(rawConfig.edge) ?? rawConfig;\n  return {\n    voice: trimToUndefined(raw.voice ?? raw.voiceId) ?? DEFAULT_VOICE,\n    lang: trimToUndefined(raw.lang) ?? DEFAULT_LANG,\n    outputFormat: trimToUndefined(raw.outputFormat) ?? DEFAULT_OUTPUT_FORMAT,\n    pitch: trimToUndefined(raw.pitch),\n    rate: trimToUndefined(raw.rate),\n    volume: trimToUndefined(raw.volume),\n    proxy: trimToUndefined(raw.proxy),\n    enabled: raw.enabled !== false, // default-on\n  };\n}\n\nfunction readProviderConfig(config: SpeechProviderConfig): EdgeTtsConfig {\n  return {\n    voice: trimToUndefined(config.voice ?? config.voiceId) ?? DEFAULT_VOICE,\n    lang: trimToUndefined(config.lang) ?? DEFAULT_LANG,\n    outputFormat: trimToUndefined(config.outputFormat) ?? DEFAULT_OUTPUT_FORMAT,\n    pitch: trimToUndefined(config.pitch),\n    rate: trimToUndefined(config.rate),\n    volume: trimToUndefined(config.volume),\n    proxy: trimToUndefined(config.proxy),\n    enabled: config.enabled !== false,\n  };\n}\n\nfunction parseDirectiveTokenInternal(\n  ctx: SpeechDirectiveTokenParseContext,\n): SpeechDirectiveTokenParseResult {\n  switch (ctx.key) {\n    case 'voice':\n    case 'voice_id':\n    case 'voiceid':\n    case 'edge_voice':\n    case 'edgevoice':\n      if (!ctx.policy.allowVoice) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { voice: ctx.value } };\n    default:\n      return { handled: false };\n  }\n}\n\nasync function synthesizeToBuffer(\n  text: string,\n  config: EdgeTtsConfig,\n  voiceOverride: string | undefined,\n  timeoutMs: number,\n): Promise<{ buffer: Buffer; outputFormat: string; ext: string }> {\n  // Lazy import — node-edge-tts loads ws transitively; skip cost for users that\n  // never enable the edge provider.\n  const { EdgeTTS } = await import('node-edge-tts');\n\n  const voice = voiceOverride ?? config.voice;\n  const tempDir = mkdtempSync(path.join(tmpdir(), 'tts-edge-'));\n  const { ext, format } = inferEdgeExtension(config.outputFormat);\n  const outputPath = path.join(tempDir, `speech-${Date.now()}.${ext}`);\n\n  try {\n    log.debug(\n      { voice, outputFormat: config.outputFormat, textLength: text.length },\n      'Calling Edge TTS',\n    );\n    const tts = new EdgeTTS({\n      voice,\n      lang: config.lang,\n      outputFormat: config.outputFormat,\n      proxy: config.proxy,\n      rate: config.rate,\n      pitch: config.pitch,\n      volume: config.volume,\n      timeout: timeoutMs,\n    });\n    await tts.ttsPromise(text, outputPath);\n    const buffer = readFileSync(outputPath);\n    return { buffer, outputFormat: format, ext };\n  } finally {\n    try {\n      rmSync(tempDir, { recursive: true, force: true });\n    } catch (cleanupError) {\n      log.warn({ err: cleanupError, tempDir }, 'Failed to cleanup Edge TTS temp directory');\n    }\n  }\n}\n\nexport const edgeSpeechProvider: SpeechProviderPlugin = {\n  id: 'edge',\n  autoSelectOrder: 100,\n\n  resolveConfig: (ctx: SpeechProviderResolveConfigContext) => normalizeConfig(ctx.rawConfig),\n\n  isConfigured: (ctx: SpeechProviderConfiguredContext) => readProviderConfig(ctx.providerConfig).enabled,\n\n  parseDirectiveToken: parseDirectiveTokenInternal,\n\n  // Edge has hundreds of voices; we don't enumerate them statically. UI should\n  // call upstream Microsoft endpoint for the full catalog.\n  listVoices: async () => [{ id: DEFAULT_VOICE, name: DEFAULT_VOICE, locale: DEFAULT_LANG }],\n\n  synthesize: async (req: SpeechSynthesisRequest): Promise<SpeechSynthesisResult> => {\n    const config = readProviderConfig(req.providerConfig);\n    if (!config.enabled) {\n      throw new Error('Edge TTS is disabled in config (messages.tts.providers.edge.enabled = false)');\n    }\n    const overrides = req.providerOverrides ?? {};\n    const voiceOverride = trimToUndefined(overrides.voice ?? overrides.voiceId);\n    const { buffer, outputFormat, ext } = await synthesizeToBuffer(\n      req.text,\n      config,\n      voiceOverride,\n      req.timeoutMs,\n    );\n    return {\n      audioBuffer: buffer,\n      outputFormat,\n      fileExtension: ext,\n      voiceCompatible: outputFormat === 'opus' || outputFormat === 'ogg',\n    };\n  },\n\n  // True streaming would require a node-edge-tts upstream change. We expose a\n  // \"fake stream\" (single chunk) so callers don't fall through to speak-core's\n  // wrapBufferAsStream branch and can treat all providers uniformly.\n  synthesizeStream: async (req: SpeechSynthesisStreamRequest): Promise<SpeechSynthesisStreamResult> => {\n    const config = readProviderConfig(req.providerConfig);\n    if (!config.enabled) {\n      throw new Error('Edge TTS is disabled in config (messages.tts.providers.edge.enabled = false)');\n    }\n    const overrides = req.providerOverrides ?? {};\n    const voiceOverride = trimToUndefined(overrides.voice ?? overrides.voiceId);\n    const { buffer, outputFormat, ext } = await synthesizeToBuffer(\n      req.text,\n      config,\n      voiceOverride,\n      req.timeoutMs,\n    );\n    const audioStream = new ReadableStream<Uint8Array>({\n      start(controller) {\n        controller.enqueue(new Uint8Array(buffer));\n        controller.close();\n      },\n    });\n    return {\n      audioStream,\n      outputFormat,\n      fileExtension: ext,\n      voiceCompatible: outputFormat === 'opus' || outputFormat === 'ogg',\n    };\n  },\n};\n\nregisterSpeechProvider(edgeSpeechProvider);\n"],"mappings":";;;;;;;;;;;;;;;;;;;;aAkBwD;AAexD,MAAM,MAAM,aAAa,sBAAsB;AAE/C,MAAM,gBAAgB;AACtB,MAAM,eAAe;AACrB,MAAM,wBAAwB;AAa9B,SAAS,gBAAgB,OAAoC;AAC3D,KAAI,OAAO,UAAU,SACnB;CAEF,MAAM,UAAU,MAAM,MAAM;AAC5B,QAAO,QAAQ,SAAS,IAAI,UAAU,KAAA;;AAGxC,SAAS,SAAS,OAAqD;AACrE,QAAO,OAAO,UAAU,YAAY,UAAU,QAAQ,CAAC,MAAM,QAAQ,MAAM,GACtE,QACD,KAAA;;AAGN,SAAS,mBAAmB,cAAuD;CACjF,MAAM,aAAa,aAAa,aAAa;AAC7C,KAAI,WAAW,SAAS,OAAO,CAAE,QAAO;EAAE,KAAK;EAAQ,QAAQ;EAAQ;AACvE,KAAI,WAAW,SAAS,OAAO,CAAE,QAAO;EAAE,KAAK;EAAQ,QAAQ;EAAQ;AACvE,KAAI,WAAW,SAAS,MAAM,CAAE,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;AACpE,KAAI,WAAW,SAAS,MAAM,IAAI,WAAW,SAAS,OAAO,IAAI,WAAW,SAAS,MAAM,CACzF,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;AAEtC,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;;AAGtC,SAAS,gBAAgB,WAAmD;CAC1E,MAAM,MAAM,SAAS,UAAU,KAAK,IAAI;AACxC,QAAO;EACL,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI;EACpD,MAAM,gBAAgB,IAAI,KAAK,IAAI;EACnC,cAAc,gBAAgB,IAAI,aAAa,IAAI;EACnD,OAAO,gBAAgB,IAAI,MAAM;EACjC,MAAM,gBAAgB,IAAI,KAAK;EAC/B,QAAQ,gBAAgB,IAAI,OAAO;EACnC,OAAO,gBAAgB,IAAI,MAAM;EACjC,SAAS,IAAI,YAAY;EAC1B;;AAGH,SAAS,mBAAmB,QAA6C;AACvE,QAAO;EACL,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI;EAC1D,MAAM,gBAAgB,OAAO,KAAK,IAAI;EACtC,cAAc,gBAAgB,OAAO,aAAa,IAAI;EACtD,OAAO,gBAAgB,OAAO,MAAM;EACpC,MAAM,gBAAgB,OAAO,KAAK;EAClC,QAAQ,gBAAgB,OAAO,OAAO;EACtC,OAAO,gBAAgB,OAAO,MAAM;EACpC,SAAS,OAAO,YAAY;EAC7B;;AAGH,SAAS,4BACP,KACiC;AACjC,SAAQ,IAAI,KAAZ;EACE,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK;AACH,OAAI,CAAC,IAAI,OAAO,WACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,QACE,QAAO,EAAE,SAAS,OAAO;;;AAI/B,eAAe,mBACb,MACA,QACA,eACA,WACgE;CAGhE,MAAM,EAAE,YAAY,MAAM,OAAO;CAEjC,MAAM,QAAQ,iBAAiB,OAAO;CACtC,MAAM,UAAU,YAAY,KAAK,KAAK,QAAQ,EAAE,YAAY,CAAC;CAC7D,MAAM,EAAE,KAAK,WAAW,mBAAmB,OAAO,aAAa;CAC/D,MAAM,aAAa,KAAK,KAAK,SAAS,UAAU,KAAK,KAAK,CAAC,GAAG,MAAM;AAEpE,KAAI;AACF,MAAI,MACF;GAAE;GAAO,cAAc,OAAO;GAAc,YAAY,KAAK;GAAQ,EACrE,mBACD;AAWD,QAAM,IAVU,QAAQ;GACtB;GACA,MAAM,OAAO;GACb,cAAc,OAAO;GACrB,OAAO,OAAO;GACd,MAAM,OAAO;GACb,OAAO,OAAO;GACd,QAAQ,OAAO;GACf,SAAS;GACV,CACQ,CAAC,WAAW,MAAM,WAAW;AAEtC,SAAO;GAAE,QADM,aAAa,WACb;GAAE,cAAc;GAAQ;GAAK;WACpC;AACR,MAAI;AACF,UAAO,SAAS;IAAE,WAAW;IAAM,OAAO;IAAM,CAAC;WAC1C,cAAc;AACrB,OAAI,KAAK;IAAE,KAAK;IAAc;IAAS,EAAE,4CAA4C;;;;AAK3F,MAAa,qBAA2C;CACtD,IAAI;CACJ,iBAAiB;CAEjB,gBAAgB,QAA4C,gBAAgB,IAAI,UAAU;CAE1F,eAAe,QAAyC,mBAAmB,IAAI,eAAe,CAAC;CAE/F,qBAAqB;CAIrB,YAAY,YAAY,CAAC;EAAE,IAAI;EAAe,MAAM;EAAe,QAAQ;EAAc,CAAC;CAE1F,YAAY,OAAO,QAAgE;EACjF,MAAM,SAAS,mBAAmB,IAAI,eAAe;AACrD,MAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,+EAA+E;EAEjG,MAAM,YAAY,IAAI,qBAAqB,EAAE;EAC7C,MAAM,gBAAgB,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC3E,MAAM,EAAE,QAAQ,cAAc,QAAQ,MAAM,mBAC1C,IAAI,MACJ,QACA,eACA,IAAI,UACL;AACD,SAAO;GACL,aAAa;GACb;GACA,eAAe;GACf,iBAAiB,iBAAiB,UAAU,iBAAiB;GAC9D;;CAMH,kBAAkB,OAAO,QAA4E;EACnG,MAAM,SAAS,mBAAmB,IAAI,eAAe;AACrD,MAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,+EAA+E;EAEjG,MAAM,YAAY,IAAI,qBAAqB,EAAE;EAC7C,MAAM,gBAAgB,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC3E,MAAM,EAAE,QAAQ,cAAc,QAAQ,MAAM,mBAC1C,IAAI,MACJ,QACA,eACA,IAAI,UACL;AAOD,SAAO;GACL,aAAA,IAPsB,eAA2B,EACjD,MAAM,YAAY;AAChB,eAAW,QAAQ,IAAI,WAAW,OAAO,CAAC;AAC1C,eAAW,OAAO;MAErB,CAEY;GACX;GACA,eAAe;GACf,iBAAiB,iBAAiB,UAAU,iBAAiB;GAC9D;;CAEJ;AAED,uBAAuB,mBAAmB"}
+{"version":3,"file":"edge-speech.js","names":[],"sources":["../../../../../src/voice/tts/providers/edge-speech.ts"],"sourcesContent":["/**\n * Edge TTS provider — wraps the `node-edge-tts` package. Requires no API key\n * (Microsoft's Edge browser endpoint is unauthenticated).\n *\n * Implementation notes:\n *   - Temp-file dance (mkdtemp → ttsPromise → readFileSync → rm) is required\n *     because `node-edge-tts` does not expose a stream-to-buffer API.\n *   - `synthesizeStream` is implemented as \"synthesize + wrap as single-chunk\n *     stream\" so callers see uniform behavior. This is not a true network\n *     stream; it's buffered-then-wrapped.\n *   - `inferEdgeExtension` derives the file extension from Edge's outputFormat\n *     string.\n */\n\nimport { mkdtempSync, readFileSync, rmSync, statSync } from 'node:fs';\nimport { tmpdir } from 'node:os';\nimport path from 'node:path';\n\nimport { createLogger } from '../../../utils/logger.js';\nimport { registerSpeechProvider } from '../speech-registry.js';\nimport type {\n  SpeechDirectiveTokenParseContext,\n  SpeechDirectiveTokenParseResult,\n  SpeechProviderConfig,\n  SpeechProviderConfiguredContext,\n  SpeechProviderPlugin,\n  SpeechProviderResolveConfigContext,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechSynthesisStreamRequest,\n  SpeechSynthesisStreamResult,\n} from '../speech-provider-types.js';\n\nconst log = createLogger('SpeechProvider:Edge');\n\nconst DEFAULT_VOICE = 'en-US-MichelleNeural';\nconst DEFAULT_LANG = 'en-US';\nconst DEFAULT_OUTPUT_FORMAT = 'audio-24khz-48kbitrate-mono-mp3';\n\ninterface EdgeTtsConfig extends Record<string, unknown> {\n  voice: string;\n  lang: string;\n  outputFormat: string;\n  pitch?: string;\n  rate?: string;\n  volume?: string;\n  proxy?: string;\n  enabled: boolean;\n}\n\nfunction trimToUndefined(value: unknown): string | undefined {\n  if (typeof value !== 'string') {\n    return undefined;\n  }\n  const trimmed = value.trim();\n  return trimmed.length > 0 ? trimmed : undefined;\n}\n\nfunction asObject(value: unknown): Record<string, unknown> | undefined {\n  return typeof value === 'object' && value !== null && !Array.isArray(value)\n    ? (value as Record<string, unknown>)\n    : undefined;\n}\n\nfunction inferEdgeExtension(outputFormat: string): { ext: string; format: string } {\n  const normalized = outputFormat.toLowerCase();\n  if (normalized.includes('webm')) return { ext: 'webm', format: 'webm' };\n  if (normalized.includes('opus')) return { ext: 'opus', format: 'opus' };\n  if (normalized.includes('ogg')) return { ext: 'ogg', format: 'ogg' };\n  if (normalized.includes('wav') || normalized.includes('riff') || normalized.includes('pcm')) {\n    return { ext: 'wav', format: 'wav' };\n  }\n  return { ext: 'mp3', format: 'mp3' };\n}\n\nfunction normalizeConfig(rawConfig: Record<string, unknown>): EdgeTtsConfig {\n  const raw = asObject(rawConfig.edge) ?? rawConfig;\n  return {\n    voice: trimToUndefined(raw.voice ?? raw.voiceId) ?? DEFAULT_VOICE,\n    lang: trimToUndefined(raw.lang) ?? DEFAULT_LANG,\n    outputFormat: trimToUndefined(raw.outputFormat) ?? DEFAULT_OUTPUT_FORMAT,\n    pitch: trimToUndefined(raw.pitch),\n    rate: trimToUndefined(raw.rate),\n    volume: trimToUndefined(raw.volume),\n    proxy: trimToUndefined(raw.proxy),\n    enabled: raw.enabled !== false, // default-on\n  };\n}\n\nfunction readProviderConfig(config: SpeechProviderConfig): EdgeTtsConfig {\n  return {\n    voice: trimToUndefined(config.voice ?? config.voiceId) ?? DEFAULT_VOICE,\n    lang: trimToUndefined(config.lang) ?? DEFAULT_LANG,\n    outputFormat: trimToUndefined(config.outputFormat) ?? DEFAULT_OUTPUT_FORMAT,\n    pitch: trimToUndefined(config.pitch),\n    rate: trimToUndefined(config.rate),\n    volume: trimToUndefined(config.volume),\n    proxy: trimToUndefined(config.proxy),\n    enabled: config.enabled !== false,\n  };\n}\n\nfunction parseDirectiveTokenInternal(\n  ctx: SpeechDirectiveTokenParseContext,\n): SpeechDirectiveTokenParseResult {\n  switch (ctx.key) {\n    case 'voice':\n    case 'voice_id':\n    case 'voiceid':\n    case 'edge_voice':\n    case 'edgevoice':\n      if (!ctx.policy.allowVoice) {\n        return { handled: true };\n      }\n      return { handled: true, overrides: { voice: ctx.value } };\n    default:\n      return { handled: false };\n  }\n}\n\nasync function waitForNonEmptyFile(filePath: string, timeoutMs: number): Promise<void> {\n  const deadline = Date.now() + Math.min(Math.max(timeoutMs, 1000), 5000);\n  while (Date.now() < deadline) {\n    try {\n      if (statSync(filePath).size > 0) return;\n    } catch {\n      // The Edge package may resolve before the file is visible on slower filesystems.\n    }\n    await new Promise((resolve) => setTimeout(resolve, 50));\n  }\n}\n\nasync function synthesizeToBuffer(\n  text: string,\n  config: EdgeTtsConfig,\n  voiceOverride: string | undefined,\n  timeoutMs: number,\n): Promise<{ buffer: Buffer; outputFormat: string; ext: string }> {\n  // Lazy import — node-edge-tts loads ws transitively; skip cost for users that\n  // never enable the edge provider.\n  const { EdgeTTS } = await import('node-edge-tts');\n\n  const voice = voiceOverride ?? config.voice;\n  const tempDir = mkdtempSync(path.join(tmpdir(), 'tts-edge-'));\n  const { ext, format } = inferEdgeExtension(config.outputFormat);\n  const outputPath = path.join(tempDir, `speech-${Date.now()}.${ext}`);\n\n  try {\n    log.debug(\n      { voice, outputFormat: config.outputFormat, textLength: text.length },\n      'Calling Edge TTS',\n    );\n    const tts = new EdgeTTS({\n      voice,\n      lang: config.lang,\n      outputFormat: config.outputFormat,\n      proxy: config.proxy,\n      rate: config.rate,\n      pitch: config.pitch,\n      volume: config.volume,\n      timeout: timeoutMs,\n    });\n    await tts.ttsPromise(text, outputPath);\n    await waitForNonEmptyFile(outputPath, timeoutMs);\n    const buffer = readFileSync(outputPath);\n    if (buffer.length === 0) {\n      throw new Error(`Edge TTS produced an empty ${format} file for voice ${voice}`);\n    }\n    return { buffer, outputFormat: format, ext };\n  } finally {\n    try {\n      rmSync(tempDir, { recursive: true, force: true });\n    } catch (cleanupError) {\n      log.warn({ err: cleanupError, tempDir }, 'Failed to cleanup Edge TTS temp directory');\n    }\n  }\n}\n\nexport const edgeSpeechProvider: SpeechProviderPlugin = {\n  id: 'edge',\n  autoSelectOrder: 100,\n\n  resolveConfig: (ctx: SpeechProviderResolveConfigContext) => normalizeConfig(ctx.rawConfig),\n\n  isConfigured: (ctx: SpeechProviderConfiguredContext) => readProviderConfig(ctx.providerConfig).enabled,\n\n  parseDirectiveToken: parseDirectiveTokenInternal,\n\n  // Edge has hundreds of voices; we don't enumerate them statically. UI should\n  // call upstream Microsoft endpoint for the full catalog.\n  listVoices: async () => [{ id: DEFAULT_VOICE, name: DEFAULT_VOICE, locale: DEFAULT_LANG }],\n\n  synthesize: async (req: SpeechSynthesisRequest): Promise<SpeechSynthesisResult> => {\n    const config = readProviderConfig(req.providerConfig);\n    if (!config.enabled) {\n      throw new Error('Edge TTS is disabled in config (messages.tts.providers.edge.enabled = false)');\n    }\n    const overrides = req.providerOverrides ?? {};\n    const voiceOverride = trimToUndefined(overrides.voice ?? overrides.voiceId);\n    const { buffer, outputFormat, ext } = await synthesizeToBuffer(\n      req.text,\n      config,\n      voiceOverride,\n      req.timeoutMs,\n    );\n    return {\n      audioBuffer: buffer,\n      outputFormat,\n      fileExtension: ext,\n      voiceCompatible: outputFormat === 'opus' || outputFormat === 'ogg',\n    };\n  },\n\n  // True streaming would require a node-edge-tts upstream change. We expose a\n  // \"fake stream\" (single chunk) so callers don't fall through to speak-core's\n  // wrapBufferAsStream branch and can treat all providers uniformly.\n  synthesizeStream: async (req: SpeechSynthesisStreamRequest): Promise<SpeechSynthesisStreamResult> => {\n    const config = readProviderConfig(req.providerConfig);\n    if (!config.enabled) {\n      throw new Error('Edge TTS is disabled in config (messages.tts.providers.edge.enabled = false)');\n    }\n    const overrides = req.providerOverrides ?? {};\n    const voiceOverride = trimToUndefined(overrides.voice ?? overrides.voiceId);\n    const { buffer, outputFormat, ext } = await synthesizeToBuffer(\n      req.text,\n      config,\n      voiceOverride,\n      req.timeoutMs,\n    );\n    const audioStream = new ReadableStream<Uint8Array>({\n      start(controller) {\n        controller.enqueue(new Uint8Array(buffer));\n        controller.close();\n      },\n    });\n    return {\n      audioStream,\n      outputFormat,\n      fileExtension: ext,\n      voiceCompatible: outputFormat === 'opus' || outputFormat === 'ogg',\n    };\n  },\n};\n\nregisterSpeechProvider(edgeSpeechProvider);\n"],"mappings":";;;;;;;;;;;;;;;;;;;;aAkBwD;AAexD,MAAM,MAAM,aAAa,sBAAsB;AAE/C,MAAM,gBAAgB;AACtB,MAAM,eAAe;AACrB,MAAM,wBAAwB;AAa9B,SAAS,gBAAgB,OAAoC;AAC3D,KAAI,OAAO,UAAU,SACnB;CAEF,MAAM,UAAU,MAAM,MAAM;AAC5B,QAAO,QAAQ,SAAS,IAAI,UAAU,KAAA;;AAGxC,SAAS,SAAS,OAAqD;AACrE,QAAO,OAAO,UAAU,YAAY,UAAU,QAAQ,CAAC,MAAM,QAAQ,MAAM,GACtE,QACD,KAAA;;AAGN,SAAS,mBAAmB,cAAuD;CACjF,MAAM,aAAa,aAAa,aAAa;AAC7C,KAAI,WAAW,SAAS,OAAO,CAAE,QAAO;EAAE,KAAK;EAAQ,QAAQ;EAAQ;AACvE,KAAI,WAAW,SAAS,OAAO,CAAE,QAAO;EAAE,KAAK;EAAQ,QAAQ;EAAQ;AACvE,KAAI,WAAW,SAAS,MAAM,CAAE,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;AACpE,KAAI,WAAW,SAAS,MAAM,IAAI,WAAW,SAAS,OAAO,IAAI,WAAW,SAAS,MAAM,CACzF,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;AAEtC,QAAO;EAAE,KAAK;EAAO,QAAQ;EAAO;;AAGtC,SAAS,gBAAgB,WAAmD;CAC1E,MAAM,MAAM,SAAS,UAAU,KAAK,IAAI;AACxC,QAAO;EACL,OAAO,gBAAgB,IAAI,SAAS,IAAI,QAAQ,IAAI;EACpD,MAAM,gBAAgB,IAAI,KAAK,IAAI;EACnC,cAAc,gBAAgB,IAAI,aAAa,IAAI;EACnD,OAAO,gBAAgB,IAAI,MAAM;EACjC,MAAM,gBAAgB,IAAI,KAAK;EAC/B,QAAQ,gBAAgB,IAAI,OAAO;EACnC,OAAO,gBAAgB,IAAI,MAAM;EACjC,SAAS,IAAI,YAAY;EAC1B;;AAGH,SAAS,mBAAmB,QAA6C;AACvE,QAAO;EACL,OAAO,gBAAgB,OAAO,SAAS,OAAO,QAAQ,IAAI;EAC1D,MAAM,gBAAgB,OAAO,KAAK,IAAI;EACtC,cAAc,gBAAgB,OAAO,aAAa,IAAI;EACtD,OAAO,gBAAgB,OAAO,MAAM;EACpC,MAAM,gBAAgB,OAAO,KAAK;EAClC,QAAQ,gBAAgB,OAAO,OAAO;EACtC,OAAO,gBAAgB,OAAO,MAAM;EACpC,SAAS,OAAO,YAAY;EAC7B;;AAGH,SAAS,4BACP,KACiC;AACjC,SAAQ,IAAI,KAAZ;EACE,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK;EACL,KAAK;AACH,OAAI,CAAC,IAAI,OAAO,WACd,QAAO,EAAE,SAAS,MAAM;AAE1B,UAAO;IAAE,SAAS;IAAM,WAAW,EAAE,OAAO,IAAI,OAAO;IAAE;EAC3D,QACE,QAAO,EAAE,SAAS,OAAO;;;AAI/B,eAAe,oBAAoB,UAAkB,WAAkC;CACrF,MAAM,WAAW,KAAK,KAAK,GAAG,KAAK,IAAI,KAAK,IAAI,WAAW,IAAK,EAAE,IAAK;AACvE,QAAO,KAAK,KAAK,GAAG,UAAU;AAC5B,MAAI;AACF,OAAI,SAAS,SAAS,CAAC,OAAO,EAAG;UAC3B;AAGR,QAAM,IAAI,SAAS,YAAY,WAAW,SAAS,GAAG,CAAC;;;AAI3D,eAAe,mBACb,MACA,QACA,eACA,WACgE;CAGhE,MAAM,EAAE,YAAY,MAAM,OAAO;CAEjC,MAAM,QAAQ,iBAAiB,OAAO;CACtC,MAAM,UAAU,YAAY,KAAK,KAAK,QAAQ,EAAE,YAAY,CAAC;CAC7D,MAAM,EAAE,KAAK,WAAW,mBAAmB,OAAO,aAAa;CAC/D,MAAM,aAAa,KAAK,KAAK,SAAS,UAAU,KAAK,KAAK,CAAC,GAAG,MAAM;AAEpE,KAAI;AACF,MAAI,MACF;GAAE;GAAO,cAAc,OAAO;GAAc,YAAY,KAAK;GAAQ,EACrE,mBACD;AAWD,QAAM,IAVU,QAAQ;GACtB;GACA,MAAM,OAAO;GACb,cAAc,OAAO;GACrB,OAAO,OAAO;GACd,MAAM,OAAO;GACb,OAAO,OAAO;GACd,QAAQ,OAAO;GACf,SAAS;GACV,CACQ,CAAC,WAAW,MAAM,WAAW;AACtC,QAAM,oBAAoB,YAAY,UAAU;EAChD,MAAM,SAAS,aAAa,WAAW;AACvC,MAAI,OAAO,WAAW,EACpB,OAAM,IAAI,MAAM,8BAA8B,OAAO,kBAAkB,QAAQ;AAEjF,SAAO;GAAE;GAAQ,cAAc;GAAQ;GAAK;WACpC;AACR,MAAI;AACF,UAAO,SAAS;IAAE,WAAW;IAAM,OAAO;IAAM,CAAC;WAC1C,cAAc;AACrB,OAAI,KAAK;IAAE,KAAK;IAAc;IAAS,EAAE,4CAA4C;;;;AAK3F,MAAa,qBAA2C;CACtD,IAAI;CACJ,iBAAiB;CAEjB,gBAAgB,QAA4C,gBAAgB,IAAI,UAAU;CAE1F,eAAe,QAAyC,mBAAmB,IAAI,eAAe,CAAC;CAE/F,qBAAqB;CAIrB,YAAY,YAAY,CAAC;EAAE,IAAI;EAAe,MAAM;EAAe,QAAQ;EAAc,CAAC;CAE1F,YAAY,OAAO,QAAgE;EACjF,MAAM,SAAS,mBAAmB,IAAI,eAAe;AACrD,MAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,+EAA+E;EAEjG,MAAM,YAAY,IAAI,qBAAqB,EAAE;EAC7C,MAAM,gBAAgB,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC3E,MAAM,EAAE,QAAQ,cAAc,QAAQ,MAAM,mBAC1C,IAAI,MACJ,QACA,eACA,IAAI,UACL;AACD,SAAO;GACL,aAAa;GACb;GACA,eAAe;GACf,iBAAiB,iBAAiB,UAAU,iBAAiB;GAC9D;;CAMH,kBAAkB,OAAO,QAA4E;EACnG,MAAM,SAAS,mBAAmB,IAAI,eAAe;AACrD,MAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,+EAA+E;EAEjG,MAAM,YAAY,IAAI,qBAAqB,EAAE;EAC7C,MAAM,gBAAgB,gBAAgB,UAAU,SAAS,UAAU,QAAQ;EAC3E,MAAM,EAAE,QAAQ,cAAc,QAAQ,MAAM,mBAC1C,IAAI,MACJ,QACA,eACA,IAAI,UACL;AAOD,SAAO;GACL,aAAA,IAPsB,eAA2B,EACjD,MAAM,YAAY;AAChB,eAAW,QAAQ,IAAI,WAAW,OAAO,CAAC;AAC1C,eAAW,OAAO;MAErB,CAEY;GACX;GACA,eAAe;GACf,iBAAiB,iBAAiB,UAAU,iBAAiB;GAC9D;;CAEJ;AAED,uBAAuB,mBAAmB"}

package/dist/src/voice/tts/speak-core.js

@@ -100,7 +100,7 @@ function buildSynthesisRequest(resolved, preparedText, ttsOptions, directiveOver
 	};
 }
 /** Convert SpeechSynthesisResult → TTSResult shape (audio/format/provider). */
-function toLegacyResult(resolved, result, durationSeconds) {
+function toTtsResult(resolved, result, durationSeconds) {
 	return {
 		audio: result.audioBuffer,
 		format: result.outputFormat,
@@ -167,7 +167,7 @@ async function speak(text, config, options) {
 				size: result.audioBuffer.length
 			}, "TTS succeeded");
 			return {
-				...toLegacyResult(resolved, result, durationSeconds),
+				...toTtsResult(resolved, result, durationSeconds),
 				attempts,
 				...fallbackFrom ? { fallbackFrom } : {},
 				attemptedProviders,
@@ -211,7 +211,7 @@ async function speakWithProvider(text, config, providerName, options) {
 	const startTime = Date.now();
 	const request = buildSynthesisRequest(resolved, prepared.text, prepared.ttsOptions, prepared.directiveOverrides, options?.appConfig, "audio-file");
 	return {
-		...toLegacyResult(resolved, await resolved.plugin.synthesize(request), (Date.now() - startTime) / 1e3),
+		...toTtsResult(resolved, await resolved.plugin.synthesize(request), (Date.now() - startTime) / 1e3),
 		attempts: [{
 			provider: providerName,
 			outcome: "success",

package/dist/src/voice/tts/speak-core.js.map

@@ -1 +1 @@
-{"version":3,"file":"speak-core.js","names":[],"sources":["../../../../src/voice/tts/speak-core.ts"],"sourcesContent":["/**\n * speak-core — orchestrates per-call TTS synthesis through the SpeechProviderPlugin chain.\n *\n * Provider invocation goes through `plugin.synthesize` / `plugin.synthesizeStream`.\n * Streaming consumers (Telegram draft etc.) use `speakStream`; providers without\n * `synthesizeStream` automatically fall back to `synthesize` wrapped as a\n * single-chunk ReadableStream (see wrapBufferAsStream).\n *\n * Directive overrides are routed via per-provider buckets (`overrides.openai`,\n * `overrides.minimax`, …). The bucket whose key matches the currently active\n * provider id wins; cross-provider keys are ignored.\n *\n * `cfg: Config` is threaded through `SpeakOptions.appConfig`. The orchestrator\n * does NOT auto-summarize / preprocess for `speakStream` beyond the same path\n * as `speak`; streaming is a transport optimization, not a different semantic.\n */\n\nimport type { Config } from '../../config/schema.js';\nimport { createLogger } from '../../utils/logger.js';\n\nimport { parseTtsDirectives } from './directives.js';\nimport { resolveSpeechProviderChain, resolveSpeechProvider, type ResolvedSpeechProvider } from './factory.js';\nimport { preprocessText, type PreprocessOptions, type PreprocessResult } from './preprocess.js';\nimport type {\n  SpeechProviderOverrides,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechSynthesisStreamRequest,\n  SpeechSynthesisStreamResult,\n} from './speech-provider-types.js';\nimport { summarizeForTts } from './summarize.js';\nimport type {\n  ProviderAttempt,\n  ProviderFailureReason,\n  TTSConfig,\n  TTSModelOverrideConfig,\n  TTSOptions,\n  TTSProvider,\n  TTSResultWithTracking,\n  TtsDirectiveOverrides,\n} from './types.js';\n\nconst log = createLogger('TTS');\n\nexport interface SpeakOptions {\n  tts?: TTSOptions;\n  preprocess?: PreprocessOptions;\n  parseDirectives?: boolean;\n  modelOverrides?: TTSModelOverrideConfig;\n  appConfig?: Config;\n  /** Optional caller-supplied AbortSignal (currently passed through to providers via timeoutMs guard only). */\n  signal?: AbortSignal;\n}\n\n/** Pulls the override bucket for a given provider id from the directive overrides. */\nfunction pickProviderOverrides(\n  overrides: TtsDirectiveOverrides | undefined,\n  providerId: TTSProvider,\n): SpeechProviderOverrides | undefined {\n  if (!overrides) return undefined;\n  const bucket = (overrides as Record<string, unknown>)[providerId];\n  if (bucket && typeof bucket === 'object') {\n    return bucket as SpeechProviderOverrides;\n  }\n  return undefined;\n}\n\n/** Merges TTSOptions (caller-supplied) with directive bucket. Directive wins on overlap. */\nfunction mergeOverrides(\n  ttsOptions: TTSOptions | undefined,\n  directiveBucket: SpeechProviderOverrides | undefined,\n): SpeechProviderOverrides | undefined {\n  const merged: Record<string, unknown> = {};\n  if (ttsOptions?.voice) merged.voice = ttsOptions.voice;\n  if (ttsOptions?.model) merged.model = ttsOptions.model;\n  if (ttsOptions?.speed) merged.speed = ttsOptions.speed;\n  if (directiveBucket) {\n    Object.assign(merged, directiveBucket);\n  }\n  return Object.keys(merged).length > 0 ? merged : undefined;\n}\n\nfunction classifyError(error: unknown): ProviderFailureReason {\n  if (error instanceof Error) {\n    const message = error.message.toLowerCase();\n    if (message.includes('not configured') || message.includes('api key')) {\n      return 'not_configured';\n    }\n    if (message.includes('timeout') || message.includes('timed out') || message.includes('aborted')) {\n      return 'timeout';\n    }\n    if (message.includes('too long') || message.includes('text length') || message.includes('exceeds')) {\n      return 'text_too_long';\n    }\n    return 'provider_error';\n  }\n  return 'unknown';\n}\n\n/** Common pre-flight: directive parse → preprocess → optional summarize. */\nasync function prepareTextForSynthesis(\n  text: string,\n  config: TTSConfig,\n  options: SpeakOptions | undefined,\n): Promise<{\n  text: string;\n  wasPreprocessed: boolean;\n  wasSummarized: boolean;\n  ttsOptions: TTSOptions | undefined;\n  directiveOverrides: TtsDirectiveOverrides | undefined;\n  selectedProvider: TTSProvider;\n}> {\n  let ttsText = text;\n  let ttsOptions = options?.tts;\n  let directiveOverrides: TtsDirectiveOverrides | undefined;\n  let selectedProvider: TTSProvider = config.provider;\n\n  if (options?.parseDirectives !== false && config.modelOverrides?.enabled) {\n    const directiveResult = parseTtsDirectives(text, config.modelOverrides);\n    ttsText = directiveResult.ttsText || directiveResult.cleanedText;\n    directiveOverrides = directiveResult.overrides;\n    if (directiveResult.overrides?.provider) {\n      selectedProvider = directiveResult.overrides.provider;\n    }\n  }\n\n  const maxLength = config.maxTextLength || 4096;\n  const preprocessOptions: PreprocessOptions = {\n    maxLength,\n    stripMarkdown: true,\n    normalizeWhitespace: true,\n    ...options?.preprocess,\n  };\n\n  let preprocessResult: PreprocessResult = preprocessText(ttsText, preprocessOptions);\n  const wasPreprocessed =\n    preprocessResult.wasTruncated || preprocessResult.originalLength !== preprocessResult.finalLength;\n\n  let wasSummarized = false;\n  const sumCfg = config.summarization;\n  const threshold = sumCfg?.threshold ?? maxLength;\n  if (sumCfg?.enabled !== false && preprocessResult.text.length > threshold) {\n    const summarizeResult = await summarizeForTts({\n      text: preprocessResult.text,\n      targetLength: sumCfg?.targetLength ?? maxLength,\n      config: options?.appConfig,\n      modelRef: sumCfg?.model,\n    });\n    if (summarizeResult.wasSummarized) {\n      preprocessResult = {\n        ...preprocessResult,\n        text: summarizeResult.summary,\n        finalLength: summarizeResult.summaryLength,\n        wasTruncated: false,\n      };\n      wasSummarized = true;\n      log.info(\n        {\n          originalLength: summarizeResult.originalLength,\n          summaryLength: summarizeResult.summaryLength,\n        },\n        'Text summarized before TTS',\n      );\n    }\n  }\n\n  return {\n    text: preprocessResult.text,\n    wasPreprocessed,\n    wasSummarized,\n    ttsOptions,\n    directiveOverrides,\n    selectedProvider,\n  };\n}\n\n/** Build the SpeechSynthesisRequest from a resolved provider + prepared text. */\nfunction buildSynthesisRequest(\n  resolved: ResolvedSpeechProvider,\n  preparedText: string,\n  ttsOptions: TTSOptions | undefined,\n  directiveOverrides: TtsDirectiveOverrides | undefined,\n  appConfig: Config | undefined,\n  target: 'audio-file' | 'voice-note',\n): SpeechSynthesisRequest {\n  const directiveBucket = pickProviderOverrides(directiveOverrides, resolved.providerId);\n  const providerOverrides = mergeOverrides(ttsOptions, directiveBucket);\n  const callTimeoutMs = ttsOptions?.timeoutMs ?? resolved.timeoutMs;\n  return {\n    text: preparedText,\n    cfg: appConfig as Config,\n    providerConfig: resolved.providerConfig,\n    target,\n    timeoutMs: callTimeoutMs,\n    ...(providerOverrides ? { providerOverrides } : {}),\n  };\n}\n\n/** Convert SpeechSynthesisResult → TTSResult shape (audio/format/provider). */\nfunction toLegacyResult(\n  resolved: ResolvedSpeechProvider,\n  result: SpeechSynthesisResult,\n  durationSeconds: number | undefined,\n): { audio: Buffer; format: string; provider: string; duration?: number } {\n  return {\n    audio: result.audioBuffer,\n    format: result.outputFormat,\n    provider: resolved.providerId,\n    ...(durationSeconds !== undefined ? { duration: durationSeconds } : {}),\n  };\n}\n\n// ---- Public API ---------------------------------------------------------\n\n/**\n * Run the TTS chain (primary → fallback) and return the first successful result.\n */\nexport async function speak(\n  text: string,\n  config: TTSConfig,\n  options?: SpeakOptions,\n): Promise<TTSResultWithTracking> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  if (config.trigger === 'tagged' && !/\\[\\[tts/.test(text)) {\n    throw new Error('TTS trigger is tagged but no [[tts]] directive found');\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const chain = resolveSpeechProviderChain(config);\n\n  // Reorder so the directive-selected provider (if registered + configured) goes first.\n  const orderedChain = (() => {\n    const idx = chain.findIndex((c) => c.providerId === prepared.selectedProvider);\n    if (idx <= 0) return chain;\n    const reordered = [...chain];\n    const [picked] = reordered.splice(idx, 1);\n    if (picked) {\n      reordered.unshift(picked);\n    }\n    return reordered;\n  })();\n\n  const attempts: ProviderAttempt[] = [];\n  const attemptedProviders: string[] = [];\n  let lastError: Error | undefined;\n\n  for (const resolved of orderedChain) {\n    const startTime = Date.now();\n    attemptedProviders.push(resolved.providerId);\n\n    if (typeof resolved.plugin.synthesize !== 'function') {\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'skipped',\n        reasonCode: 'not_configured',\n        latencyMs: Date.now() - startTime,\n        error: `Provider \"${resolved.providerId}\" does not implement synthesize()`,\n      });\n      continue;\n    }\n\n    try {\n      log.debug(\n        { textLength: prepared.text.length, provider: resolved.providerId },\n        'Converting text to speech',\n      );\n      const request = buildSynthesisRequest(\n        resolved,\n        prepared.text,\n        prepared.ttsOptions,\n        prepared.directiveOverrides,\n        options?.appConfig,\n        'audio-file',\n      );\n      const result = await resolved.plugin.synthesize(request);\n      const durationSeconds = (Date.now() - startTime) / 1000;\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'success',\n        reasonCode: 'success',\n        latencyMs: Date.now() - startTime,\n      });\n\n      const primaryProvider = orderedChain[0]!.providerId;\n      const fallbackFrom = resolved.providerId !== primaryProvider ? primaryProvider : undefined;\n      if (fallbackFrom) {\n        log.info(\n          { primaryProvider: fallbackFrom, actualProvider: resolved.providerId, attempts },\n          'TTS used fallback provider',\n        );\n      } else {\n        log.info(\n          { provider: resolved.providerId, format: result.outputFormat, size: result.audioBuffer.length },\n          'TTS succeeded',\n        );\n      }\n\n      return {\n        ...toLegacyResult(resolved, result, durationSeconds),\n        attempts,\n        ...(fallbackFrom ? { fallbackFrom } : {}),\n        attemptedProviders,\n        wasPreprocessed: prepared.wasPreprocessed,\n        ttsText: prepared.text,\n        wasSummarized: prepared.wasSummarized,\n      };\n    } catch (error) {\n      const latencyMs = Date.now() - startTime;\n      const reasonCode = classifyError(error);\n      const errorMsg = error instanceof Error ? error.message : String(error);\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'failed',\n        reasonCode,\n        latencyMs,\n        error: errorMsg,\n      });\n      lastError = error instanceof Error ? error : new Error(String(error));\n      log.warn(\n        { provider: resolved.providerId, errorMessage: errorMsg, reasonCode, latencyMs },\n        'TTS provider failed, trying next',\n      );\n    }\n  }\n\n  log.error({ attempts, attemptedProviders }, 'All TTS providers failed');\n  throw (\n    lastError ?? new Error(`All TTS providers failed: ${attempts.map((a) => a.error).join('; ')}`)\n  );\n}\n\n/** Single-provider variant; bypasses the fallback chain. */\nexport async function speakWithProvider(\n  text: string,\n  config: TTSConfig,\n  providerName: TTSProvider,\n  options?: SpeakOptions,\n): Promise<TTSResultWithTracking> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  const resolved = resolveSpeechProvider(providerName, config);\n  if (!resolved) {\n    throw new Error(`Provider '${providerName}' is not available`);\n  }\n  if (typeof resolved.plugin.synthesize !== 'function') {\n    throw new Error(`Provider '${providerName}' does not implement synthesize()`);\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const startTime = Date.now();\n  const request = buildSynthesisRequest(\n    resolved,\n    prepared.text,\n    prepared.ttsOptions,\n    prepared.directiveOverrides,\n    options?.appConfig,\n    'audio-file',\n  );\n  const result = await resolved.plugin.synthesize(request);\n  const durationSeconds = (Date.now() - startTime) / 1000;\n\n  return {\n    ...toLegacyResult(resolved, result, durationSeconds),\n    attempts: [\n      {\n        provider: providerName,\n        outcome: 'success',\n        reasonCode: 'success',\n        latencyMs: Date.now() - startTime,\n      },\n    ],\n    attemptedProviders: [providerName],\n    wasPreprocessed: prepared.wasPreprocessed,\n    ttsText: prepared.text,\n    wasSummarized: prepared.wasSummarized,\n  };\n}\n\n// ---- Streaming surface --------------------------------------------------\n\nexport interface SpeakStreamResult {\n  /** Audio chunks. Caller is responsible for consuming the stream. */\n  audioStream: ReadableStream<Uint8Array>;\n  outputFormat: string;\n  fileExtension: string;\n  voiceCompatible: boolean;\n  /** Provider id that produced this stream. */\n  provider: TTSProvider;\n  /** Cleanup hook (close sockets etc.). Always defined; no-op when nothing to release. */\n  release: () => Promise<void>;\n  /** Pre-flight metadata (after directive parse / preprocess / summarize). */\n  ttsText: string;\n  wasPreprocessed: boolean;\n  wasSummarized: boolean;\n}\n\n/**\n * Streaming TTS — tries the chain and returns the first provider's stream.\n * Providers without `synthesizeStream` are wrapped via wrapBufferAsStream.\n */\nexport async function speakStream(\n  text: string,\n  config: TTSConfig,\n  options?: SpeakOptions,\n): Promise<SpeakStreamResult> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  if (config.trigger === 'tagged' && !/\\[\\[tts/.test(text)) {\n    throw new Error('TTS trigger is tagged but no [[tts]] directive found');\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const chain = resolveSpeechProviderChain(config);\n\n  let lastError: Error | undefined;\n  for (const resolved of chain) {\n    try {\n      const baseRequest: SpeechSynthesisStreamRequest = buildSynthesisRequest(\n        resolved,\n        prepared.text,\n        prepared.ttsOptions,\n        prepared.directiveOverrides,\n        options?.appConfig,\n        'voice-note',\n      );\n\n      let streamResult: SpeechSynthesisStreamResult;\n      if (typeof resolved.plugin.synthesizeStream === 'function') {\n        streamResult = await resolved.plugin.synthesizeStream(baseRequest);\n      } else if (typeof resolved.plugin.synthesize === 'function') {\n        // Fallback: buffer-then-wrap. Documented behavior: callers see a\n        // single-chunk stream with the same outputFormat.\n        const bufferResult = await resolved.plugin.synthesize(baseRequest);\n        streamResult = wrapBufferAsStream(bufferResult);\n      } else {\n        throw new Error(\n          `Provider \"${resolved.providerId}\" implements neither synthesize nor synthesizeStream`,\n        );\n      }\n\n      log.info(\n        {\n          provider: resolved.providerId,\n          outputFormat: streamResult.outputFormat,\n          voiceCompatible: streamResult.voiceCompatible,\n          streamed: typeof resolved.plugin.synthesizeStream === 'function',\n        },\n        'TTS stream ready',\n      );\n\n      const release = streamResult.release ?? (async () => {});\n      return {\n        audioStream: streamResult.audioStream,\n        outputFormat: streamResult.outputFormat,\n        fileExtension: streamResult.fileExtension,\n        voiceCompatible: streamResult.voiceCompatible,\n        provider: resolved.providerId,\n        release,\n        ttsText: prepared.text,\n        wasPreprocessed: prepared.wasPreprocessed,\n        wasSummarized: prepared.wasSummarized,\n      };\n    } catch (error) {\n      lastError = error instanceof Error ? error : new Error(String(error));\n      log.warn(\n        { err: lastError, provider: resolved.providerId },\n        `TTS stream provider \"${resolved.providerId}\" failed, trying next`,\n      );\n    }\n  }\n  throw lastError ?? new Error('All TTS providers failed (streaming)');\n}\n\n/** Wrap a one-shot synthesis result into a single-chunk stream (no-op release). */\nfunction wrapBufferAsStream(result: SpeechSynthesisResult): SpeechSynthesisStreamResult {\n  const { audioBuffer } = result;\n  const audioStream = new ReadableStream<Uint8Array>({\n    start(controller) {\n      controller.enqueue(new Uint8Array(audioBuffer));\n      controller.close();\n    },\n  });\n  return {\n    audioStream,\n    outputFormat: result.outputFormat,\n    fileExtension: result.fileExtension,\n    voiceCompatible: result.voiceCompatible,\n  };\n}\n"],"mappings":";;;;;;;aAkBqD;AAwBrD,MAAM,MAAM,aAAa,MAAM;;AAa/B,SAAS,sBACP,WACA,YACqC;AACrC,KAAI,CAAC,UAAW,QAAO,KAAA;CACvB,MAAM,SAAU,UAAsC;AACtD,KAAI,UAAU,OAAO,WAAW,SAC9B,QAAO;;;AAMX,SAAS,eACP,YACA,iBACqC;CACrC,MAAM,SAAkC,EAAE;AAC1C,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,gBACF,QAAO,OAAO,QAAQ,gBAAgB;AAExC,QAAO,OAAO,KAAK,OAAO,CAAC,SAAS,IAAI,SAAS,KAAA;;AAGnD,SAAS,cAAc,OAAuC;AAC5D,KAAI,iBAAiB,OAAO;EAC1B,MAAM,UAAU,MAAM,QAAQ,aAAa;AAC3C,MAAI,QAAQ,SAAS,iBAAiB,IAAI,QAAQ,SAAS,UAAU,CACnE,QAAO;AAET,MAAI,QAAQ,SAAS,UAAU,IAAI,QAAQ,SAAS,YAAY,IAAI,QAAQ,SAAS,UAAU,CAC7F,QAAO;AAET,MAAI,QAAQ,SAAS,WAAW,IAAI,QAAQ,SAAS,cAAc,IAAI,QAAQ,SAAS,UAAU,CAChG,QAAO;AAET,SAAO;;AAET,QAAO;;;AAIT,eAAe,wBACb,MACA,QACA,SAQC;CACD,IAAI,UAAU;CACd,IAAI,aAAa,SAAS;CAC1B,IAAI;CACJ,IAAI,mBAAgC,OAAO;AAE3C,KAAI,SAAS,oBAAoB,SAAS,OAAO,gBAAgB,SAAS;EACxE,MAAM,kBAAkB,mBAAmB,MAAM,OAAO,eAAe;AACvE,YAAU,gBAAgB,WAAW,gBAAgB;AACrD,uBAAqB,gBAAgB;AACrC,MAAI,gBAAgB,WAAW,SAC7B,oBAAmB,gBAAgB,UAAU;;CAIjD,MAAM,YAAY,OAAO,iBAAiB;CAC1C,MAAM,oBAAuC;EAC3C;EACA,eAAe;EACf,qBAAqB;EACrB,GAAG,SAAS;EACb;CAED,IAAI,mBAAqC,eAAe,SAAS,kBAAkB;CACnF,MAAM,kBACJ,iBAAiB,gBAAgB,iBAAiB,mBAAmB,iBAAiB;CAExF,IAAI,gBAAgB;CACpB,MAAM,SAAS,OAAO;CACtB,MAAM,YAAY,QAAQ,aAAa;AACvC,KAAI,QAAQ,YAAY,SAAS,iBAAiB,KAAK,SAAS,WAAW;EACzE,MAAM,kBAAkB,MAAM,gBAAgB;GAC5C,MAAM,iBAAiB;GACvB,cAAc,QAAQ,gBAAgB;GACtC,QAAQ,SAAS;GACjB,UAAU,QAAQ;GACnB,CAAC;AACF,MAAI,gBAAgB,eAAe;AACjC,sBAAmB;IACjB,GAAG;IACH,MAAM,gBAAgB;IACtB,aAAa,gBAAgB;IAC7B,cAAc;IACf;AACD,mBAAgB;AAChB,OAAI,KACF;IACE,gBAAgB,gBAAgB;IAChC,eAAe,gBAAgB;IAChC,EACD,6BACD;;;AAIL,QAAO;EACL,MAAM,iBAAiB;EACvB;EACA;EACA;EACA;EACA;EACD;;;AAIH,SAAS,sBACP,UACA,cACA,YACA,oBACA,WACA,QACwB;CAExB,MAAM,oBAAoB,eAAe,YADjB,sBAAsB,oBAAoB,SAAS,WACP,CAAC;CACrE,MAAM,gBAAgB,YAAY,aAAa,SAAS;AACxD,QAAO;EACL,MAAM;EACN,KAAK;EACL,gBAAgB,SAAS;EACzB;EACA,WAAW;EACX,GAAI,oBAAoB,EAAE,mBAAmB,GAAG,EAAE;EACnD;;;AAIH,SAAS,eACP,UACA,QACA,iBACwE;AACxE,QAAO;EACL,OAAO,OAAO;EACd,QAAQ,OAAO;EACf,UAAU,SAAS;EACnB,GAAI,oBAAoB,KAAA,IAAY,EAAE,UAAU,iBAAiB,GAAG,EAAE;EACvE;;;;;AAQH,eAAsB,MACpB,MACA,QACA,SACgC;AAChC,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;AAEvC,KAAI,OAAO,YAAY,YAAY,CAAC,UAAU,KAAK,KAAK,CACtD,OAAM,IAAI,MAAM,uDAAuD;CAGzE,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,QAAQ,2BAA2B,OAAO;CAGhD,MAAM,sBAAsB;EAC1B,MAAM,MAAM,MAAM,WAAW,MAAM,EAAE,eAAe,SAAS,iBAAiB;AAC9E,MAAI,OAAO,EAAG,QAAO;EACrB,MAAM,YAAY,CAAC,GAAG,MAAM;EAC5B,MAAM,CAAC,UAAU,UAAU,OAAO,KAAK,EAAE;AACzC,MAAI,OACF,WAAU,QAAQ,OAAO;AAE3B,SAAO;KACL;CAEJ,MAAM,WAA8B,EAAE;CACtC,MAAM,qBAA+B,EAAE;CACvC,IAAI;AAEJ,MAAK,MAAM,YAAY,cAAc;EACnC,MAAM,YAAY,KAAK,KAAK;AAC5B,qBAAmB,KAAK,SAAS,WAAW;AAE5C,MAAI,OAAO,SAAS,OAAO,eAAe,YAAY;AACpD,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT,YAAY;IACZ,WAAW,KAAK,KAAK,GAAG;IACxB,OAAO,aAAa,SAAS,WAAW;IACzC,CAAC;AACF;;AAGF,MAAI;AACF,OAAI,MACF;IAAE,YAAY,SAAS,KAAK;IAAQ,UAAU,SAAS;IAAY,EACnE,4BACD;GACD,MAAM,UAAU,sBACd,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;GACD,MAAM,SAAS,MAAM,SAAS,OAAO,WAAW,QAAQ;GACxD,MAAM,mBAAmB,KAAK,KAAK,GAAG,aAAa;AACnD,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT,YAAY;IACZ,WAAW,KAAK,KAAK,GAAG;IACzB,CAAC;GAEF,MAAM,kBAAkB,aAAa,GAAI;GACzC,MAAM,eAAe,SAAS,eAAe,kBAAkB,kBAAkB,KAAA;AACjF,OAAI,aACF,KAAI,KACF;IAAE,iBAAiB;IAAc,gBAAgB,SAAS;IAAY;IAAU,EAChF,6BACD;OAED,KAAI,KACF;IAAE,UAAU,SAAS;IAAY,QAAQ,OAAO;IAAc,MAAM,OAAO,YAAY;IAAQ,EAC/F,gBACD;AAGH,UAAO;IACL,GAAG,eAAe,UAAU,QAAQ,gBAAgB;IACpD;IACA,GAAI,eAAe,EAAE,cAAc,GAAG,EAAE;IACxC;IACA,iBAAiB,SAAS;IAC1B,SAAS,SAAS;IAClB,eAAe,SAAS;IACzB;WACM,OAAO;GACd,MAAM,YAAY,KAAK,KAAK,GAAG;GAC/B,MAAM,aAAa,cAAc,MAAM;GACvC,MAAM,WAAW,iBAAiB,QAAQ,MAAM,UAAU,OAAO,MAAM;AACvE,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT;IACA;IACA,OAAO;IACR,CAAC;AACF,eAAY,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,MAAM,CAAC;AACrE,OAAI,KACF;IAAE,UAAU,SAAS;IAAY,cAAc;IAAU;IAAY;IAAW,EAChF,mCACD;;;AAIL,KAAI,MAAM;EAAE;EAAU;EAAoB,EAAE,2BAA2B;AACvE,OACE,6BAAa,IAAI,MAAM,6BAA6B,SAAS,KAAK,MAAM,EAAE,MAAM,CAAC,KAAK,KAAK,GAAG;;;AAKlG,eAAsB,kBACpB,MACA,QACA,cACA,SACgC;AAChC,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;CAEvC,MAAM,WAAW,sBAAsB,cAAc,OAAO;AAC5D,KAAI,CAAC,SACH,OAAM,IAAI,MAAM,aAAa,aAAa,oBAAoB;AAEhE,KAAI,OAAO,SAAS,OAAO,eAAe,WACxC,OAAM,IAAI,MAAM,aAAa,aAAa,mCAAmC;CAG/E,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,YAAY,KAAK,KAAK;CAC5B,MAAM,UAAU,sBACd,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;AAID,QAAO;EACL,GAAG,eAAe,UAAU,MAJT,SAAS,OAAO,WAAW,QAAQ,GAC/B,KAAK,KAAK,GAAG,aAAa,IAGG;EACpD,UAAU,CACR;GACE,UAAU;GACV,SAAS;GACT,YAAY;GACZ,WAAW,KAAK,KAAK,GAAG;GACzB,CACF;EACD,oBAAoB,CAAC,aAAa;EAClC,iBAAiB,SAAS;EAC1B,SAAS,SAAS;EAClB,eAAe,SAAS;EACzB;;;;;;AAyBH,eAAsB,YACpB,MACA,QACA,SAC4B;AAC5B,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;AAEvC,KAAI,OAAO,YAAY,YAAY,CAAC,UAAU,KAAK,KAAK,CACtD,OAAM,IAAI,MAAM,uDAAuD;CAGzE,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,QAAQ,2BAA2B,OAAO;CAEhD,IAAI;AACJ,MAAK,MAAM,YAAY,MACrB,KAAI;EACF,MAAM,cAA4C,sBAChD,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;EAED,IAAI;AACJ,MAAI,OAAO,SAAS,OAAO,qBAAqB,WAC9C,gBAAe,MAAM,SAAS,OAAO,iBAAiB,YAAY;WACzD,OAAO,SAAS,OAAO,eAAe,WAI/C,gBAAe,mBAAmB,MADP,SAAS,OAAO,WAAW,YAAY,CACnB;MAE/C,OAAM,IAAI,MACR,aAAa,SAAS,WAAW,sDAClC;AAGH,MAAI,KACF;GACE,UAAU,SAAS;GACnB,cAAc,aAAa;GAC3B,iBAAiB,aAAa;GAC9B,UAAU,OAAO,SAAS,OAAO,qBAAqB;GACvD,EACD,mBACD;EAED,MAAM,UAAU,aAAa,YAAY,YAAY;AACrD,SAAO;GACL,aAAa,aAAa;GAC1B,cAAc,aAAa;GAC3B,eAAe,aAAa;GAC5B,iBAAiB,aAAa;GAC9B,UAAU,SAAS;GACnB;GACA,SAAS,SAAS;GAClB,iBAAiB,SAAS;GAC1B,eAAe,SAAS;GACzB;UACM,OAAO;AACd,cAAY,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,MAAM,CAAC;AACrE,MAAI,KACF;GAAE,KAAK;GAAW,UAAU,SAAS;GAAY,EACjD,wBAAwB,SAAS,WAAW,uBAC7C;;AAGL,OAAM,6BAAa,IAAI,MAAM,uCAAuC;;;AAItE,SAAS,mBAAmB,QAA4D;CACtF,MAAM,EAAE,gBAAgB;AAOxB,QAAO;EACL,aAAA,IAPsB,eAA2B,EACjD,MAAM,YAAY;AAChB,cAAW,QAAQ,IAAI,WAAW,YAAY,CAAC;AAC/C,cAAW,OAAO;KAErB,CAEY;EACX,cAAc,OAAO;EACrB,eAAe,OAAO;EACtB,iBAAiB,OAAO;EACzB"}
+{"version":3,"file":"speak-core.js","names":[],"sources":["../../../../src/voice/tts/speak-core.ts"],"sourcesContent":["/**\n * speak-core — orchestrates per-call TTS synthesis through the SpeechProviderPlugin chain.\n *\n * Provider invocation goes through `plugin.synthesize` / `plugin.synthesizeStream`.\n * Streaming consumers (Telegram draft etc.) use `speakStream`; providers without\n * `synthesizeStream` automatically fall back to `synthesize` wrapped as a\n * single-chunk ReadableStream (see wrapBufferAsStream).\n *\n * Directive overrides are routed via per-provider buckets (`overrides.openai`,\n * `overrides.minimax`, …). The bucket whose key matches the currently active\n * provider id wins; cross-provider keys are ignored.\n *\n * `cfg: Config` is threaded through `SpeakOptions.appConfig`. The orchestrator\n * does NOT auto-summarize / preprocess for `speakStream` beyond the same path\n * as `speak`; streaming is a transport optimization, not a different semantic.\n */\n\nimport type { Config } from '../../config/schema.js';\nimport { createLogger } from '../../utils/logger.js';\n\nimport { parseTtsDirectives } from './directives.js';\nimport { resolveSpeechProviderChain, resolveSpeechProvider, type ResolvedSpeechProvider } from './factory.js';\nimport { preprocessText, type PreprocessOptions, type PreprocessResult } from './preprocess.js';\nimport type {\n  SpeechProviderOverrides,\n  SpeechSynthesisRequest,\n  SpeechSynthesisResult,\n  SpeechSynthesisStreamRequest,\n  SpeechSynthesisStreamResult,\n} from './speech-provider-types.js';\nimport { summarizeForTts } from './summarize.js';\nimport type {\n  ProviderAttempt,\n  ProviderFailureReason,\n  TTSConfig,\n  TTSModelOverrideConfig,\n  TTSOptions,\n  TTSProvider,\n  TTSResultWithTracking,\n  TtsDirectiveOverrides,\n} from './types.js';\n\nconst log = createLogger('TTS');\n\nexport interface SpeakOptions {\n  tts?: TTSOptions;\n  preprocess?: PreprocessOptions;\n  parseDirectives?: boolean;\n  modelOverrides?: TTSModelOverrideConfig;\n  appConfig?: Config;\n  /** Optional caller-supplied AbortSignal (currently passed through to providers via timeoutMs guard only). */\n  signal?: AbortSignal;\n}\n\n/** Pulls the override bucket for a given provider id from the directive overrides. */\nfunction pickProviderOverrides(\n  overrides: TtsDirectiveOverrides | undefined,\n  providerId: TTSProvider,\n): SpeechProviderOverrides | undefined {\n  if (!overrides) return undefined;\n  const bucket = (overrides as Record<string, unknown>)[providerId];\n  if (bucket && typeof bucket === 'object') {\n    return bucket as SpeechProviderOverrides;\n  }\n  return undefined;\n}\n\n/** Merges TTSOptions (caller-supplied) with directive bucket. Directive wins on overlap. */\nfunction mergeOverrides(\n  ttsOptions: TTSOptions | undefined,\n  directiveBucket: SpeechProviderOverrides | undefined,\n): SpeechProviderOverrides | undefined {\n  const merged: Record<string, unknown> = {};\n  if (ttsOptions?.voice) merged.voice = ttsOptions.voice;\n  if (ttsOptions?.model) merged.model = ttsOptions.model;\n  if (ttsOptions?.speed) merged.speed = ttsOptions.speed;\n  if (directiveBucket) {\n    Object.assign(merged, directiveBucket);\n  }\n  return Object.keys(merged).length > 0 ? merged : undefined;\n}\n\nfunction classifyError(error: unknown): ProviderFailureReason {\n  if (error instanceof Error) {\n    const message = error.message.toLowerCase();\n    if (message.includes('not configured') || message.includes('api key')) {\n      return 'not_configured';\n    }\n    if (message.includes('timeout') || message.includes('timed out') || message.includes('aborted')) {\n      return 'timeout';\n    }\n    if (message.includes('too long') || message.includes('text length') || message.includes('exceeds')) {\n      return 'text_too_long';\n    }\n    return 'provider_error';\n  }\n  return 'unknown';\n}\n\n/** Common pre-flight: directive parse → preprocess → optional summarize. */\nasync function prepareTextForSynthesis(\n  text: string,\n  config: TTSConfig,\n  options: SpeakOptions | undefined,\n): Promise<{\n  text: string;\n  wasPreprocessed: boolean;\n  wasSummarized: boolean;\n  ttsOptions: TTSOptions | undefined;\n  directiveOverrides: TtsDirectiveOverrides | undefined;\n  selectedProvider: TTSProvider;\n}> {\n  let ttsText = text;\n  let ttsOptions = options?.tts;\n  let directiveOverrides: TtsDirectiveOverrides | undefined;\n  let selectedProvider: TTSProvider = config.provider;\n\n  if (options?.parseDirectives !== false && config.modelOverrides?.enabled) {\n    const directiveResult = parseTtsDirectives(text, config.modelOverrides);\n    ttsText = directiveResult.ttsText || directiveResult.cleanedText;\n    directiveOverrides = directiveResult.overrides;\n    if (directiveResult.overrides?.provider) {\n      selectedProvider = directiveResult.overrides.provider;\n    }\n  }\n\n  const maxLength = config.maxTextLength || 4096;\n  const preprocessOptions: PreprocessOptions = {\n    maxLength,\n    stripMarkdown: true,\n    normalizeWhitespace: true,\n    ...options?.preprocess,\n  };\n\n  let preprocessResult: PreprocessResult = preprocessText(ttsText, preprocessOptions);\n  const wasPreprocessed =\n    preprocessResult.wasTruncated || preprocessResult.originalLength !== preprocessResult.finalLength;\n\n  let wasSummarized = false;\n  const sumCfg = config.summarization;\n  const threshold = sumCfg?.threshold ?? maxLength;\n  if (sumCfg?.enabled !== false && preprocessResult.text.length > threshold) {\n    const summarizeResult = await summarizeForTts({\n      text: preprocessResult.text,\n      targetLength: sumCfg?.targetLength ?? maxLength,\n      config: options?.appConfig,\n      modelRef: sumCfg?.model,\n    });\n    if (summarizeResult.wasSummarized) {\n      preprocessResult = {\n        ...preprocessResult,\n        text: summarizeResult.summary,\n        finalLength: summarizeResult.summaryLength,\n        wasTruncated: false,\n      };\n      wasSummarized = true;\n      log.info(\n        {\n          originalLength: summarizeResult.originalLength,\n          summaryLength: summarizeResult.summaryLength,\n        },\n        'Text summarized before TTS',\n      );\n    }\n  }\n\n  return {\n    text: preprocessResult.text,\n    wasPreprocessed,\n    wasSummarized,\n    ttsOptions,\n    directiveOverrides,\n    selectedProvider,\n  };\n}\n\n/** Build the SpeechSynthesisRequest from a resolved provider + prepared text. */\nfunction buildSynthesisRequest(\n  resolved: ResolvedSpeechProvider,\n  preparedText: string,\n  ttsOptions: TTSOptions | undefined,\n  directiveOverrides: TtsDirectiveOverrides | undefined,\n  appConfig: Config | undefined,\n  target: 'audio-file' | 'voice-note',\n): SpeechSynthesisRequest {\n  const directiveBucket = pickProviderOverrides(directiveOverrides, resolved.providerId);\n  const providerOverrides = mergeOverrides(ttsOptions, directiveBucket);\n  const callTimeoutMs = ttsOptions?.timeoutMs ?? resolved.timeoutMs;\n  return {\n    text: preparedText,\n    cfg: appConfig as Config,\n    providerConfig: resolved.providerConfig,\n    target,\n    timeoutMs: callTimeoutMs,\n    ...(providerOverrides ? { providerOverrides } : {}),\n  };\n}\n\n/** Convert SpeechSynthesisResult → TTSResult shape (audio/format/provider). */\nfunction toTtsResult(\n  resolved: ResolvedSpeechProvider,\n  result: SpeechSynthesisResult,\n  durationSeconds: number | undefined,\n): { audio: Buffer; format: string; provider: string; duration?: number } {\n  return {\n    audio: result.audioBuffer,\n    format: result.outputFormat,\n    provider: resolved.providerId,\n    ...(durationSeconds !== undefined ? { duration: durationSeconds } : {}),\n  };\n}\n\n// ---- Public API ---------------------------------------------------------\n\n/**\n * Run the TTS chain (primary → fallback) and return the first successful result.\n */\nexport async function speak(\n  text: string,\n  config: TTSConfig,\n  options?: SpeakOptions,\n): Promise<TTSResultWithTracking> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  if (config.trigger === 'tagged' && !/\\[\\[tts/.test(text)) {\n    throw new Error('TTS trigger is tagged but no [[tts]] directive found');\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const chain = resolveSpeechProviderChain(config);\n\n  // Reorder so the directive-selected provider (if registered + configured) goes first.\n  const orderedChain = (() => {\n    const idx = chain.findIndex((c) => c.providerId === prepared.selectedProvider);\n    if (idx <= 0) return chain;\n    const reordered = [...chain];\n    const [picked] = reordered.splice(idx, 1);\n    if (picked) {\n      reordered.unshift(picked);\n    }\n    return reordered;\n  })();\n\n  const attempts: ProviderAttempt[] = [];\n  const attemptedProviders: string[] = [];\n  let lastError: Error | undefined;\n\n  for (const resolved of orderedChain) {\n    const startTime = Date.now();\n    attemptedProviders.push(resolved.providerId);\n\n    if (typeof resolved.plugin.synthesize !== 'function') {\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'skipped',\n        reasonCode: 'not_configured',\n        latencyMs: Date.now() - startTime,\n        error: `Provider \"${resolved.providerId}\" does not implement synthesize()`,\n      });\n      continue;\n    }\n\n    try {\n      log.debug(\n        { textLength: prepared.text.length, provider: resolved.providerId },\n        'Converting text to speech',\n      );\n      const request = buildSynthesisRequest(\n        resolved,\n        prepared.text,\n        prepared.ttsOptions,\n        prepared.directiveOverrides,\n        options?.appConfig,\n        'audio-file',\n      );\n      const result = await resolved.plugin.synthesize(request);\n      const durationSeconds = (Date.now() - startTime) / 1000;\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'success',\n        reasonCode: 'success',\n        latencyMs: Date.now() - startTime,\n      });\n\n      const primaryProvider = orderedChain[0]!.providerId;\n      const fallbackFrom = resolved.providerId !== primaryProvider ? primaryProvider : undefined;\n      if (fallbackFrom) {\n        log.info(\n          { primaryProvider: fallbackFrom, actualProvider: resolved.providerId, attempts },\n          'TTS used fallback provider',\n        );\n      } else {\n        log.info(\n          { provider: resolved.providerId, format: result.outputFormat, size: result.audioBuffer.length },\n          'TTS succeeded',\n        );\n      }\n\n      return {\n        ...toTtsResult(resolved, result, durationSeconds),\n        attempts,\n        ...(fallbackFrom ? { fallbackFrom } : {}),\n        attemptedProviders,\n        wasPreprocessed: prepared.wasPreprocessed,\n        ttsText: prepared.text,\n        wasSummarized: prepared.wasSummarized,\n      };\n    } catch (error) {\n      const latencyMs = Date.now() - startTime;\n      const reasonCode = classifyError(error);\n      const errorMsg = error instanceof Error ? error.message : String(error);\n      attempts.push({\n        provider: resolved.providerId,\n        outcome: 'failed',\n        reasonCode,\n        latencyMs,\n        error: errorMsg,\n      });\n      lastError = error instanceof Error ? error : new Error(String(error));\n      log.warn(\n        { provider: resolved.providerId, errorMessage: errorMsg, reasonCode, latencyMs },\n        'TTS provider failed, trying next',\n      );\n    }\n  }\n\n  log.error({ attempts, attemptedProviders }, 'All TTS providers failed');\n  throw (\n    lastError ?? new Error(`All TTS providers failed: ${attempts.map((a) => a.error).join('; ')}`)\n  );\n}\n\n/** Single-provider variant; bypasses the fallback chain. */\nexport async function speakWithProvider(\n  text: string,\n  config: TTSConfig,\n  providerName: TTSProvider,\n  options?: SpeakOptions,\n): Promise<TTSResultWithTracking> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  const resolved = resolveSpeechProvider(providerName, config);\n  if (!resolved) {\n    throw new Error(`Provider '${providerName}' is not available`);\n  }\n  if (typeof resolved.plugin.synthesize !== 'function') {\n    throw new Error(`Provider '${providerName}' does not implement synthesize()`);\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const startTime = Date.now();\n  const request = buildSynthesisRequest(\n    resolved,\n    prepared.text,\n    prepared.ttsOptions,\n    prepared.directiveOverrides,\n    options?.appConfig,\n    'audio-file',\n  );\n  const result = await resolved.plugin.synthesize(request);\n  const durationSeconds = (Date.now() - startTime) / 1000;\n\n  return {\n    ...toTtsResult(resolved, result, durationSeconds),\n    attempts: [\n      {\n        provider: providerName,\n        outcome: 'success',\n        reasonCode: 'success',\n        latencyMs: Date.now() - startTime,\n      },\n    ],\n    attemptedProviders: [providerName],\n    wasPreprocessed: prepared.wasPreprocessed,\n    ttsText: prepared.text,\n    wasSummarized: prepared.wasSummarized,\n  };\n}\n\n// ---- Streaming surface --------------------------------------------------\n\nexport interface SpeakStreamResult {\n  /** Audio chunks. Caller is responsible for consuming the stream. */\n  audioStream: ReadableStream<Uint8Array>;\n  outputFormat: string;\n  fileExtension: string;\n  voiceCompatible: boolean;\n  /** Provider id that produced this stream. */\n  provider: TTSProvider;\n  /** Cleanup hook (close sockets etc.). Always defined; no-op when nothing to release. */\n  release: () => Promise<void>;\n  /** Pre-flight metadata (after directive parse / preprocess / summarize). */\n  ttsText: string;\n  wasPreprocessed: boolean;\n  wasSummarized: boolean;\n}\n\n/**\n * Streaming TTS — tries the chain and returns the first provider's stream.\n * Providers without `synthesizeStream` are wrapped via wrapBufferAsStream.\n */\nexport async function speakStream(\n  text: string,\n  config: TTSConfig,\n  options?: SpeakOptions,\n): Promise<SpeakStreamResult> {\n  if (!config.enabled) {\n    throw new Error('TTS is not enabled');\n  }\n  if (config.trigger === 'tagged' && !/\\[\\[tts/.test(text)) {\n    throw new Error('TTS trigger is tagged but no [[tts]] directive found');\n  }\n\n  const prepared = await prepareTextForSynthesis(text, config, options);\n  const chain = resolveSpeechProviderChain(config);\n\n  let lastError: Error | undefined;\n  for (const resolved of chain) {\n    try {\n      const baseRequest: SpeechSynthesisStreamRequest = buildSynthesisRequest(\n        resolved,\n        prepared.text,\n        prepared.ttsOptions,\n        prepared.directiveOverrides,\n        options?.appConfig,\n        'voice-note',\n      );\n\n      let streamResult: SpeechSynthesisStreamResult;\n      if (typeof resolved.plugin.synthesizeStream === 'function') {\n        streamResult = await resolved.plugin.synthesizeStream(baseRequest);\n      } else if (typeof resolved.plugin.synthesize === 'function') {\n        // Fallback: buffer-then-wrap. Documented behavior: callers see a\n        // single-chunk stream with the same outputFormat.\n        const bufferResult = await resolved.plugin.synthesize(baseRequest);\n        streamResult = wrapBufferAsStream(bufferResult);\n      } else {\n        throw new Error(\n          `Provider \"${resolved.providerId}\" implements neither synthesize nor synthesizeStream`,\n        );\n      }\n\n      log.info(\n        {\n          provider: resolved.providerId,\n          outputFormat: streamResult.outputFormat,\n          voiceCompatible: streamResult.voiceCompatible,\n          streamed: typeof resolved.plugin.synthesizeStream === 'function',\n        },\n        'TTS stream ready',\n      );\n\n      const release = streamResult.release ?? (async () => {});\n      return {\n        audioStream: streamResult.audioStream,\n        outputFormat: streamResult.outputFormat,\n        fileExtension: streamResult.fileExtension,\n        voiceCompatible: streamResult.voiceCompatible,\n        provider: resolved.providerId,\n        release,\n        ttsText: prepared.text,\n        wasPreprocessed: prepared.wasPreprocessed,\n        wasSummarized: prepared.wasSummarized,\n      };\n    } catch (error) {\n      lastError = error instanceof Error ? error : new Error(String(error));\n      log.warn(\n        { err: lastError, provider: resolved.providerId },\n        `TTS stream provider \"${resolved.providerId}\" failed, trying next`,\n      );\n    }\n  }\n  throw lastError ?? new Error('All TTS providers failed (streaming)');\n}\n\n/** Wrap a one-shot synthesis result into a single-chunk stream (no-op release). */\nfunction wrapBufferAsStream(result: SpeechSynthesisResult): SpeechSynthesisStreamResult {\n  const { audioBuffer } = result;\n  const audioStream = new ReadableStream<Uint8Array>({\n    start(controller) {\n      controller.enqueue(new Uint8Array(audioBuffer));\n      controller.close();\n    },\n  });\n  return {\n    audioStream,\n    outputFormat: result.outputFormat,\n    fileExtension: result.fileExtension,\n    voiceCompatible: result.voiceCompatible,\n  };\n}\n"],"mappings":";;;;;;;aAkBqD;AAwBrD,MAAM,MAAM,aAAa,MAAM;;AAa/B,SAAS,sBACP,WACA,YACqC;AACrC,KAAI,CAAC,UAAW,QAAO,KAAA;CACvB,MAAM,SAAU,UAAsC;AACtD,KAAI,UAAU,OAAO,WAAW,SAC9B,QAAO;;;AAMX,SAAS,eACP,YACA,iBACqC;CACrC,MAAM,SAAkC,EAAE;AAC1C,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,YAAY,MAAO,QAAO,QAAQ,WAAW;AACjD,KAAI,gBACF,QAAO,OAAO,QAAQ,gBAAgB;AAExC,QAAO,OAAO,KAAK,OAAO,CAAC,SAAS,IAAI,SAAS,KAAA;;AAGnD,SAAS,cAAc,OAAuC;AAC5D,KAAI,iBAAiB,OAAO;EAC1B,MAAM,UAAU,MAAM,QAAQ,aAAa;AAC3C,MAAI,QAAQ,SAAS,iBAAiB,IAAI,QAAQ,SAAS,UAAU,CACnE,QAAO;AAET,MAAI,QAAQ,SAAS,UAAU,IAAI,QAAQ,SAAS,YAAY,IAAI,QAAQ,SAAS,UAAU,CAC7F,QAAO;AAET,MAAI,QAAQ,SAAS,WAAW,IAAI,QAAQ,SAAS,cAAc,IAAI,QAAQ,SAAS,UAAU,CAChG,QAAO;AAET,SAAO;;AAET,QAAO;;;AAIT,eAAe,wBACb,MACA,QACA,SAQC;CACD,IAAI,UAAU;CACd,IAAI,aAAa,SAAS;CAC1B,IAAI;CACJ,IAAI,mBAAgC,OAAO;AAE3C,KAAI,SAAS,oBAAoB,SAAS,OAAO,gBAAgB,SAAS;EACxE,MAAM,kBAAkB,mBAAmB,MAAM,OAAO,eAAe;AACvE,YAAU,gBAAgB,WAAW,gBAAgB;AACrD,uBAAqB,gBAAgB;AACrC,MAAI,gBAAgB,WAAW,SAC7B,oBAAmB,gBAAgB,UAAU;;CAIjD,MAAM,YAAY,OAAO,iBAAiB;CAC1C,MAAM,oBAAuC;EAC3C;EACA,eAAe;EACf,qBAAqB;EACrB,GAAG,SAAS;EACb;CAED,IAAI,mBAAqC,eAAe,SAAS,kBAAkB;CACnF,MAAM,kBACJ,iBAAiB,gBAAgB,iBAAiB,mBAAmB,iBAAiB;CAExF,IAAI,gBAAgB;CACpB,MAAM,SAAS,OAAO;CACtB,MAAM,YAAY,QAAQ,aAAa;AACvC,KAAI,QAAQ,YAAY,SAAS,iBAAiB,KAAK,SAAS,WAAW;EACzE,MAAM,kBAAkB,MAAM,gBAAgB;GAC5C,MAAM,iBAAiB;GACvB,cAAc,QAAQ,gBAAgB;GACtC,QAAQ,SAAS;GACjB,UAAU,QAAQ;GACnB,CAAC;AACF,MAAI,gBAAgB,eAAe;AACjC,sBAAmB;IACjB,GAAG;IACH,MAAM,gBAAgB;IACtB,aAAa,gBAAgB;IAC7B,cAAc;IACf;AACD,mBAAgB;AAChB,OAAI,KACF;IACE,gBAAgB,gBAAgB;IAChC,eAAe,gBAAgB;IAChC,EACD,6BACD;;;AAIL,QAAO;EACL,MAAM,iBAAiB;EACvB;EACA;EACA;EACA;EACA;EACD;;;AAIH,SAAS,sBACP,UACA,cACA,YACA,oBACA,WACA,QACwB;CAExB,MAAM,oBAAoB,eAAe,YADjB,sBAAsB,oBAAoB,SAAS,WACP,CAAC;CACrE,MAAM,gBAAgB,YAAY,aAAa,SAAS;AACxD,QAAO;EACL,MAAM;EACN,KAAK;EACL,gBAAgB,SAAS;EACzB;EACA,WAAW;EACX,GAAI,oBAAoB,EAAE,mBAAmB,GAAG,EAAE;EACnD;;;AAIH,SAAS,YACP,UACA,QACA,iBACwE;AACxE,QAAO;EACL,OAAO,OAAO;EACd,QAAQ,OAAO;EACf,UAAU,SAAS;EACnB,GAAI,oBAAoB,KAAA,IAAY,EAAE,UAAU,iBAAiB,GAAG,EAAE;EACvE;;;;;AAQH,eAAsB,MACpB,MACA,QACA,SACgC;AAChC,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;AAEvC,KAAI,OAAO,YAAY,YAAY,CAAC,UAAU,KAAK,KAAK,CACtD,OAAM,IAAI,MAAM,uDAAuD;CAGzE,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,QAAQ,2BAA2B,OAAO;CAGhD,MAAM,sBAAsB;EAC1B,MAAM,MAAM,MAAM,WAAW,MAAM,EAAE,eAAe,SAAS,iBAAiB;AAC9E,MAAI,OAAO,EAAG,QAAO;EACrB,MAAM,YAAY,CAAC,GAAG,MAAM;EAC5B,MAAM,CAAC,UAAU,UAAU,OAAO,KAAK,EAAE;AACzC,MAAI,OACF,WAAU,QAAQ,OAAO;AAE3B,SAAO;KACL;CAEJ,MAAM,WAA8B,EAAE;CACtC,MAAM,qBAA+B,EAAE;CACvC,IAAI;AAEJ,MAAK,MAAM,YAAY,cAAc;EACnC,MAAM,YAAY,KAAK,KAAK;AAC5B,qBAAmB,KAAK,SAAS,WAAW;AAE5C,MAAI,OAAO,SAAS,OAAO,eAAe,YAAY;AACpD,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT,YAAY;IACZ,WAAW,KAAK,KAAK,GAAG;IACxB,OAAO,aAAa,SAAS,WAAW;IACzC,CAAC;AACF;;AAGF,MAAI;AACF,OAAI,MACF;IAAE,YAAY,SAAS,KAAK;IAAQ,UAAU,SAAS;IAAY,EACnE,4BACD;GACD,MAAM,UAAU,sBACd,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;GACD,MAAM,SAAS,MAAM,SAAS,OAAO,WAAW,QAAQ;GACxD,MAAM,mBAAmB,KAAK,KAAK,GAAG,aAAa;AACnD,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT,YAAY;IACZ,WAAW,KAAK,KAAK,GAAG;IACzB,CAAC;GAEF,MAAM,kBAAkB,aAAa,GAAI;GACzC,MAAM,eAAe,SAAS,eAAe,kBAAkB,kBAAkB,KAAA;AACjF,OAAI,aACF,KAAI,KACF;IAAE,iBAAiB;IAAc,gBAAgB,SAAS;IAAY;IAAU,EAChF,6BACD;OAED,KAAI,KACF;IAAE,UAAU,SAAS;IAAY,QAAQ,OAAO;IAAc,MAAM,OAAO,YAAY;IAAQ,EAC/F,gBACD;AAGH,UAAO;IACL,GAAG,YAAY,UAAU,QAAQ,gBAAgB;IACjD;IACA,GAAI,eAAe,EAAE,cAAc,GAAG,EAAE;IACxC;IACA,iBAAiB,SAAS;IAC1B,SAAS,SAAS;IAClB,eAAe,SAAS;IACzB;WACM,OAAO;GACd,MAAM,YAAY,KAAK,KAAK,GAAG;GAC/B,MAAM,aAAa,cAAc,MAAM;GACvC,MAAM,WAAW,iBAAiB,QAAQ,MAAM,UAAU,OAAO,MAAM;AACvE,YAAS,KAAK;IACZ,UAAU,SAAS;IACnB,SAAS;IACT;IACA;IACA,OAAO;IACR,CAAC;AACF,eAAY,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,MAAM,CAAC;AACrE,OAAI,KACF;IAAE,UAAU,SAAS;IAAY,cAAc;IAAU;IAAY;IAAW,EAChF,mCACD;;;AAIL,KAAI,MAAM;EAAE;EAAU;EAAoB,EAAE,2BAA2B;AACvE,OACE,6BAAa,IAAI,MAAM,6BAA6B,SAAS,KAAK,MAAM,EAAE,MAAM,CAAC,KAAK,KAAK,GAAG;;;AAKlG,eAAsB,kBACpB,MACA,QACA,cACA,SACgC;AAChC,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;CAEvC,MAAM,WAAW,sBAAsB,cAAc,OAAO;AAC5D,KAAI,CAAC,SACH,OAAM,IAAI,MAAM,aAAa,aAAa,oBAAoB;AAEhE,KAAI,OAAO,SAAS,OAAO,eAAe,WACxC,OAAM,IAAI,MAAM,aAAa,aAAa,mCAAmC;CAG/E,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,YAAY,KAAK,KAAK;CAC5B,MAAM,UAAU,sBACd,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;AAID,QAAO;EACL,GAAG,YAAY,UAAU,MAJN,SAAS,OAAO,WAAW,QAAQ,GAC/B,KAAK,KAAK,GAAG,aAAa,IAGA;EACjD,UAAU,CACR;GACE,UAAU;GACV,SAAS;GACT,YAAY;GACZ,WAAW,KAAK,KAAK,GAAG;GACzB,CACF;EACD,oBAAoB,CAAC,aAAa;EAClC,iBAAiB,SAAS;EAC1B,SAAS,SAAS;EAClB,eAAe,SAAS;EACzB;;;;;;AAyBH,eAAsB,YACpB,MACA,QACA,SAC4B;AAC5B,KAAI,CAAC,OAAO,QACV,OAAM,IAAI,MAAM,qBAAqB;AAEvC,KAAI,OAAO,YAAY,YAAY,CAAC,UAAU,KAAK,KAAK,CACtD,OAAM,IAAI,MAAM,uDAAuD;CAGzE,MAAM,WAAW,MAAM,wBAAwB,MAAM,QAAQ,QAAQ;CACrE,MAAM,QAAQ,2BAA2B,OAAO;CAEhD,IAAI;AACJ,MAAK,MAAM,YAAY,MACrB,KAAI;EACF,MAAM,cAA4C,sBAChD,UACA,SAAS,MACT,SAAS,YACT,SAAS,oBACT,SAAS,WACT,aACD;EAED,IAAI;AACJ,MAAI,OAAO,SAAS,OAAO,qBAAqB,WAC9C,gBAAe,MAAM,SAAS,OAAO,iBAAiB,YAAY;WACzD,OAAO,SAAS,OAAO,eAAe,WAI/C,gBAAe,mBAAmB,MADP,SAAS,OAAO,WAAW,YAAY,CACnB;MAE/C,OAAM,IAAI,MACR,aAAa,SAAS,WAAW,sDAClC;AAGH,MAAI,KACF;GACE,UAAU,SAAS;GACnB,cAAc,aAAa;GAC3B,iBAAiB,aAAa;GAC9B,UAAU,OAAO,SAAS,OAAO,qBAAqB;GACvD,EACD,mBACD;EAED,MAAM,UAAU,aAAa,YAAY,YAAY;AACrD,SAAO;GACL,aAAa,aAAa;GAC1B,cAAc,aAAa;GAC3B,eAAe,aAAa;GAC5B,iBAAiB,aAAa;GAC9B,UAAU,SAAS;GACnB;GACA,SAAS,SAAS;GAClB,iBAAiB,SAAS;GAC1B,eAAe,SAAS;GACzB;UACM,OAAO;AACd,cAAY,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,MAAM,CAAC;AACrE,MAAI,KACF;GAAE,KAAK;GAAW,UAAU,SAAS;GAAY,EACjD,wBAAwB,SAAS,WAAW,uBAC7C;;AAGL,OAAM,6BAAa,IAAI,MAAM,uCAAuC;;;AAItE,SAAS,mBAAmB,QAA4D;CACtF,MAAM,EAAE,gBAAgB;AAOxB,QAAO;EACL,aAAA,IAPsB,eAA2B,EACjD,MAAM,YAAY;AAChB,cAAW,QAAQ,IAAI,WAAW,YAAY,CAAC;AAC/C,cAAW,OAAO;KAErB,CAEY;EACX,cAAc,OAAO;EACrB,eAAe,OAAO;EACtB,iBAAiB,OAAO;EACzB"}

package/dist/src/voice/tts/speech-provider-types.d.ts

@@ -128,8 +128,8 @@ export interface SpeechProviderPlugin {
     /** Canonical provider id. Must be unique within the registry. */
     id: SpeechProviderId;
     /**
-     * Optional aliases (e.g. `["openai-tts"]` for back-compat with old config keys).
-     * Aliases participate in lookup but `id` is the canonical persisted name.
+     * Optional aliases for provider lookup. Aliases participate in lookup but
+     * `id` is the canonical persisted name.
      */
     aliases?: readonly string[];
     /**

package/dist/src/voice/tts/types.js

@@ -1,7 +1,7 @@
 //#region src/voice/tts/types.ts
 const DEFAULT_TTS_CONFIG = {
 	enabled: false,
-	provider: "openai",
+	provider: "edge",
 	trigger: "always",
 	fallback: {
 		enabled: true,

package/dist/src/voice/tts/types.js.map

@@ -1 +1 @@
-{"version":3,"file":"types.js","names":[],"sources":["../../../../src/voice/tts/types.ts"],"sourcesContent":["/**\n * TTS (Text-to-Speech) Types\n */\n\n/** Built-in provider ids shipped in core. Extensions add more at runtime. */\nexport type BuiltinTTSProvider = 'openai' | 'alibaba' | 'edge' | 'minimax' | 'tts-local-cli';\n\n/** Any registered SpeechProviderPlugin id (built-in or extension). */\nexport type TTSProvider = string;\n\nexport type TTSAutoMode = 'off' | 'always' | 'inbound' | 'tagged';\n\nexport interface TTSOptions {\n  /** Voice ID */\n  voice?: string;\n  /** Speed (0.25 - 4.0) */\n  speed?: number;\n  /** Output format */\n  format?: 'opus' | 'mp3' | 'wav';\n  /** Model ID (for provider-specific model selection) */\n  model?: string;\n  /** Request timeout in milliseconds */\n  timeoutMs?: number;\n}\n\nexport interface TTSResult {\n  /** Audio buffer */\n  audio: Buffer;\n  /** Format */\n  format: string;\n  /** Duration in seconds (if available) */\n  duration?: number;\n  /** Provider that generated the speech */\n  provider: string;\n}\n\nexport interface TTSProviderInterface {\n  /** Provider name */\n  name: string;\n  /** Convert text to speech */\n  speak(text: string, options?: TTSOptions): Promise<TTSResult>;\n  /** Check if provider is configured */\n  isConfigured(): boolean;\n}\n\nexport interface TTSModelOverrideConfig {\n  /** Enable model-provided overrides for TTS */\n  enabled?: boolean;\n  /** Allow model-provided TTS text blocks */\n  allowText?: boolean;\n  /** Allow model-provided provider override (default: false) */\n  allowProvider?: boolean;\n  /** Allow model-provided voice/voiceId override */\n  allowVoice?: boolean;\n  /** Allow model-provided modelId override */\n  allowModelId?: boolean;\n  /** Allow model-provided voice settings override */\n  allowVoiceSettings?: boolean;\n  /** Allow model-provided normalization or language overrides */\n  allowNormalization?: boolean;\n  /** Allow model-provided seed override */\n  allowSeed?: boolean;\n}\n\nexport interface TTSConfig {\n  enabled: boolean;\n  provider: TTSProvider;\n  /** Trigger mode: auto = reply with voice when user sends voice */\n  trigger: TTSAutoMode;\n  /** Fallback configuration */\n  fallback?: {\n    enabled: boolean;\n    order: TTSProvider[];\n  };\n  /** Maximum text length for TTS */\n  maxTextLength?: number;\n  /** API request timeout (ms) */\n  timeoutMs?: number;\n  /** Long-text summarization before TTS */\n  summarization?: TTSSummarizationConfig;\n  /** Allow model to override TTS parameters */\n  modelOverrides?: TTSModelOverrideConfig;\n  /** Provider settings map (`messages.tts.providers.<id>`). */\n  providers?: Record<string, Record<string, unknown>>;\n}\n\nexport interface TTSSummarizationConfig {\n  /** When true (default), long text is summarized via LLM before TTS */\n  enabled?: boolean;\n  targetLength?: number;\n  threshold?: number;\n  model?: string;\n}\n\nexport const DEFAULT_TTS_CONFIG: TTSConfig = {\n  enabled: false,\n  provider: 'openai',\n  trigger: 'always',\n  fallback: {\n    enabled: true,\n    order: ['openai', 'alibaba', 'minimax', 'edge'],\n  },\n  maxTextLength: 512, // Conservative default to accommodate all providers (Alibaba limit is 512)\n  timeoutMs: 30000,\n  summarization: {\n    enabled: true,\n  },\n  modelOverrides: {\n    enabled: true,\n    allowText: true,\n    allowProvider: false,\n    allowVoice: true,\n    allowModelId: true,\n    allowVoiceSettings: false,\n    allowNormalization: false,\n    allowSeed: false,\n  },\n  providers: {\n    alibaba: { model: 'qwen-tts', voice: 'Cherry' },\n    openai: { model: 'tts-1', voice: 'alloy' },\n    edge: {\n      enabled: true,\n      voice: 'en-US-MichelleNeural',\n      lang: 'en-US',\n      outputFormat: 'audio-24khz-48kbitrate-mono-mp3',\n    },\n    minimax: { model: 'speech-2.8-hd', voice: 'male-qn-qingse' },\n  },\n};\n\n/** TTS directive parse result */\nexport interface TtsDirectiveParseResult {\n  cleanedText: string;\n  ttsText?: string;\n  hasDirective: boolean;\n  overrides: TtsDirectiveOverrides;\n  warnings: string[];\n}\n\n/** TTS directive overrides */\nexport interface TtsDirectiveOverrides {\n  ttsText?: string;\n  provider?: TTSProvider;\n  openai?: {\n    voice?: string;\n    model?: string;\n  };\n  alibaba?: {\n    voice?: string;\n    model?: string;\n  };\n  edge?: {\n    voice?: string;\n  };\n  minimax?: {\n    voice?: string;\n    model?: string;\n  };\n}\n\n/** Provider attempt result (TTS fallback chain) */\nexport type ProviderAttemptOutcome = 'success' | 'skipped' | 'failed';\n\n/** Provider failure reason classification */\nexport type ProviderFailureReason =\n  | 'success'\n  | 'not_configured'\n  | 'timeout'\n  | 'provider_error'\n  | 'text_too_long'\n  | 'unknown';\n\n/** Single provider attempt in a TTS fallback chain */\nexport interface ProviderAttempt {\n  provider: string;\n  outcome: ProviderAttemptOutcome;\n  reasonCode: ProviderFailureReason;\n  latencyMs: number;\n  error?: string;\n}\n\n/** TTS result including fallback / preprocessing metadata */\nexport interface TTSResultWithTracking extends TTSResult {\n  attempts: ProviderAttempt[];\n  fallbackFrom?: string;\n  attemptedProviders: string[];\n  wasPreprocessed?: boolean;\n  ttsText?: string;\n  wasSummarized?: boolean;\n}\n\n/** Last TTS call diagnostics (memory-only) */\nexport interface TtsStatusEntry {\n  timestamp: number;\n  success: boolean;\n  provider?: string;\n  latencyMs?: number;\n  error?: string;\n  textLength?: number;\n  audioSize?: number;\n  audioFormat?: string;\n  usedFallback?: boolean;\n  wasSummarized?: boolean;\n}\n\nexport interface TtsRuntimeStatus {\n  lastAttempt?: TtsStatusEntry;\n  recentSuccessRate?: number;\n  totalCalls: number;\n  totalSuccesses: number;\n  totalFailures: number;\n}\n"],"mappings":";AA8FA,MAAa,qBAAgC;CAC3C,SAAS;CACT,UAAU;CACV,SAAS;CACT,UAAU;EACR,SAAS;EACT,OAAO;GAAC;GAAU;GAAW;GAAW;GAAO;EAChD;CACD,eAAe;CACf,WAAW;CACX,eAAe,EACb,SAAS,MACV;CACD,gBAAgB;EACd,SAAS;EACT,WAAW;EACX,eAAe;EACf,YAAY;EACZ,cAAc;EACd,oBAAoB;EACpB,oBAAoB;EACpB,WAAW;EACZ;CACD,WAAW;EACT,SAAS;GAAE,OAAO;GAAY,OAAO;GAAU;EAC/C,QAAQ;GAAE,OAAO;GAAS,OAAO;GAAS;EAC1C,MAAM;GACJ,SAAS;GACT,OAAO;GACP,MAAM;GACN,cAAc;GACf;EACD,SAAS;GAAE,OAAO;GAAiB,OAAO;GAAkB;EAC7D;CACF"}
+{"version":3,"file":"types.js","names":[],"sources":["../../../../src/voice/tts/types.ts"],"sourcesContent":["/**\n * TTS (Text-to-Speech) Types\n */\n\n/** Built-in provider ids shipped in core. Extensions add more at runtime. */\nexport type BuiltinTTSProvider = 'openai' | 'alibaba' | 'edge' | 'minimax' | 'tts-local-cli';\n\n/** Any registered SpeechProviderPlugin id (built-in or extension). */\nexport type TTSProvider = string;\n\nexport type TTSAutoMode = 'off' | 'always' | 'inbound' | 'tagged';\n\nexport interface TTSOptions {\n  /** Voice ID */\n  voice?: string;\n  /** Speed (0.25 - 4.0) */\n  speed?: number;\n  /** Output format */\n  format?: 'opus' | 'mp3' | 'wav';\n  /** Model ID (for provider-specific model selection) */\n  model?: string;\n  /** Request timeout in milliseconds */\n  timeoutMs?: number;\n}\n\nexport interface TTSResult {\n  /** Audio buffer */\n  audio: Buffer;\n  /** Format */\n  format: string;\n  /** Duration in seconds (if available) */\n  duration?: number;\n  /** Provider that generated the speech */\n  provider: string;\n}\n\nexport interface TTSProviderInterface {\n  /** Provider name */\n  name: string;\n  /** Convert text to speech */\n  speak(text: string, options?: TTSOptions): Promise<TTSResult>;\n  /** Check if provider is configured */\n  isConfigured(): boolean;\n}\n\nexport interface TTSModelOverrideConfig {\n  /** Enable model-provided overrides for TTS */\n  enabled?: boolean;\n  /** Allow model-provided TTS text blocks */\n  allowText?: boolean;\n  /** Allow model-provided provider override (default: false) */\n  allowProvider?: boolean;\n  /** Allow model-provided voice/voiceId override */\n  allowVoice?: boolean;\n  /** Allow model-provided modelId override */\n  allowModelId?: boolean;\n  /** Allow model-provided voice settings override */\n  allowVoiceSettings?: boolean;\n  /** Allow model-provided normalization or language overrides */\n  allowNormalization?: boolean;\n  /** Allow model-provided seed override */\n  allowSeed?: boolean;\n}\n\nexport interface TTSConfig {\n  enabled: boolean;\n  provider: TTSProvider;\n  /** Trigger mode: auto = reply with voice when user sends voice */\n  trigger: TTSAutoMode;\n  /** Fallback configuration */\n  fallback?: {\n    enabled: boolean;\n    order: TTSProvider[];\n  };\n  /** Maximum text length for TTS */\n  maxTextLength?: number;\n  /** API request timeout (ms) */\n  timeoutMs?: number;\n  /** Long-text summarization before TTS */\n  summarization?: TTSSummarizationConfig;\n  /** Allow model to override TTS parameters */\n  modelOverrides?: TTSModelOverrideConfig;\n  /** Provider settings map (`messages.tts.providers.<id>`). */\n  providers?: Record<string, Record<string, unknown>>;\n}\n\nexport interface TTSSummarizationConfig {\n  /** When true (default), long text is summarized via LLM before TTS */\n  enabled?: boolean;\n  targetLength?: number;\n  threshold?: number;\n  model?: string;\n}\n\nexport const DEFAULT_TTS_CONFIG: TTSConfig = {\n  enabled: false,\n  provider: 'edge',\n  trigger: 'always',\n  fallback: {\n    enabled: true,\n    order: ['openai', 'alibaba', 'minimax', 'edge'],\n  },\n  maxTextLength: 512, // Conservative default to accommodate all providers (Alibaba limit is 512)\n  timeoutMs: 30000,\n  summarization: {\n    enabled: true,\n  },\n  modelOverrides: {\n    enabled: true,\n    allowText: true,\n    allowProvider: false,\n    allowVoice: true,\n    allowModelId: true,\n    allowVoiceSettings: false,\n    allowNormalization: false,\n    allowSeed: false,\n  },\n  providers: {\n    alibaba: { model: 'qwen-tts', voice: 'Cherry' },\n    openai: { model: 'tts-1', voice: 'alloy' },\n    edge: {\n      enabled: true,\n      voice: 'en-US-MichelleNeural',\n      lang: 'en-US',\n      outputFormat: 'audio-24khz-48kbitrate-mono-mp3',\n    },\n    minimax: { model: 'speech-2.8-hd', voice: 'male-qn-qingse' },\n  },\n};\n\n/** TTS directive parse result */\nexport interface TtsDirectiveParseResult {\n  cleanedText: string;\n  ttsText?: string;\n  hasDirective: boolean;\n  overrides: TtsDirectiveOverrides;\n  warnings: string[];\n}\n\n/** TTS directive overrides */\nexport interface TtsDirectiveOverrides {\n  ttsText?: string;\n  provider?: TTSProvider;\n  openai?: {\n    voice?: string;\n    model?: string;\n  };\n  alibaba?: {\n    voice?: string;\n    model?: string;\n  };\n  edge?: {\n    voice?: string;\n  };\n  minimax?: {\n    voice?: string;\n    model?: string;\n  };\n}\n\n/** Provider attempt result (TTS fallback chain) */\nexport type ProviderAttemptOutcome = 'success' | 'skipped' | 'failed';\n\n/** Provider failure reason classification */\nexport type ProviderFailureReason =\n  | 'success'\n  | 'not_configured'\n  | 'timeout'\n  | 'provider_error'\n  | 'text_too_long'\n  | 'unknown';\n\n/** Single provider attempt in a TTS fallback chain */\nexport interface ProviderAttempt {\n  provider: string;\n  outcome: ProviderAttemptOutcome;\n  reasonCode: ProviderFailureReason;\n  latencyMs: number;\n  error?: string;\n}\n\n/** TTS result including fallback / preprocessing metadata */\nexport interface TTSResultWithTracking extends TTSResult {\n  attempts: ProviderAttempt[];\n  fallbackFrom?: string;\n  attemptedProviders: string[];\n  wasPreprocessed?: boolean;\n  ttsText?: string;\n  wasSummarized?: boolean;\n}\n\n/** Last TTS call diagnostics (memory-only) */\nexport interface TtsStatusEntry {\n  timestamp: number;\n  success: boolean;\n  provider?: string;\n  latencyMs?: number;\n  error?: string;\n  textLength?: number;\n  audioSize?: number;\n  audioFormat?: string;\n  usedFallback?: boolean;\n  wasSummarized?: boolean;\n}\n\nexport interface TtsRuntimeStatus {\n  lastAttempt?: TtsStatusEntry;\n  recentSuccessRate?: number;\n  totalCalls: number;\n  totalSuccesses: number;\n  totalFailures: number;\n}\n"],"mappings":";AA8FA,MAAa,qBAAgC;CAC3C,SAAS;CACT,UAAU;CACV,SAAS;CACT,UAAU;EACR,SAAS;EACT,OAAO;GAAC;GAAU;GAAW;GAAW;GAAO;EAChD;CACD,eAAe;CACf,WAAW;CACX,eAAe,EACb,SAAS,MACV;CACD,gBAAgB;EACd,SAAS;EACT,WAAW;EACX,eAAe;EACf,YAAY;EACZ,cAAc;EACd,oBAAoB;EACpB,oBAAoB;EACpB,WAAW;EACZ;CACD,WAAW;EACT,SAAS;GAAE,OAAO;GAAY,OAAO;GAAU;EAC/C,QAAQ;GAAE,OAAO;GAAS,OAAO;GAAS;EAC1C,MAAM;GACJ,SAAS;GACT,OAAO;GACP,MAAM;GACN,cAAc;GACf;EACD,SAAS;GAAE,OAAO;GAAiB,OAAO;GAAkB;EAC7D;CACF"}

package/dist/src/workflows/domain/result.d.ts

@@ -6,7 +6,6 @@ export interface WorkflowResultEnvelope {
     artifacts?: WorkflowArtifactRef[];
     followUps?: WorkflowFollowUp[];
     structuredOutput?: unknown;
-    raw?: unknown;
 }
 /** Product-level outcome contract for reusable workflow deliverables. */
 export interface WorkflowOutcome {

package/dist/src/workflows/draft/workflow-draft-service.js

@@ -1,6 +1,6 @@
-import { getAgentDefaultModelRef, init_schema } from "../../config/schema.js";
 import { createLogger } from "../../utils/logger/index.js";
 import { init_logger } from "../../utils/logger.js";
+import { getAgentDefaultModelRef, init_schema } from "../../config/schema.js";
 import { getDefaultModelSync, init_providers, resolveModel } from "../../providers/index.js";
 import { resolveModelRef } from "../../config/agent-typed-models.js";
 import { buildWorkflowDraftPrompt, buildWorkflowDraftRepairPrompt } from "./workflow-draft-prompt.js";

package/dist/src/workflows/engine/workflow-engine.js

@@ -156,8 +156,9 @@ var WorkflowEngine = class {
 			}), options.timeoutSec ?? definition.defaults.timeoutSec);
 			await eventQueue;
 			if (currentPhaseId) await appendEvent("phase_completed", { phaseId: currentPhaseId });
-			assertWorkflowOutput(definition, runtimeResult.result);
-			await appendEvent("run_completed", { result: toWorkflowResultEnvelope(runtimeResult.result) });
+			const result = requireWorkflowResultEnvelope(runtimeResult.result);
+			assertWorkflowOutput(definition, result);
+			await appendEvent("run_completed", { result });
 			await this.callHooks((hook) => hook.afterRun?.({
 				runId,
 				status: "succeeded"
@@ -659,27 +660,6 @@ function truncate(value, maxLength) {
 	if (trimmed.length <= maxLength) return trimmed;
 	return `${trimmed.slice(0, maxLength - 1)}…`;
 }
-function toWorkflowResultEnvelope(value) {
-	if (isWorkflowResultEnvelope(value)) return value;
-	if (typeof value === "string") return {
-		summary: truncate(value, 800),
-		sections: [{
-			kind: "text",
-			title: "Result",
-			content: value
-		}],
-		raw: value
-	};
-	return {
-		summary: "Workflow completed.",
-		sections: [{
-			kind: "json",
-			title: "Result",
-			value
-		}],
-		raw: value
-	};
-}
 function assertWorkflowOutput(definition, value) {
 	const validation = validateWorkflowJsonSchema(definition.outputSchema, value);
 	if (validation.ok) return;
@@ -690,6 +670,10 @@ function isWorkflowResultEnvelope(value) {
 	const record = value;
 	return typeof record.summary === "string" && Array.isArray(record.sections);
 }
+function requireWorkflowResultEnvelope(value) {
+	if (isWorkflowResultEnvelope(value)) return value;
+	throw new WorkflowEngineRunError("result_validation_failed", "Workflow scripts must return a WorkflowResultEnvelope with summary and sections.", false);
+}
 function toWorkflowRunError(err, wasAborted, abortReason) {
 	if (err instanceof WorkflowEngineRunError) return {
 		code: err.code,