@ai-matrx/media 0.14.0 → 0.15.0

package/dist/voices.d.cts

@@ -1,3 +1,5 @@
+export { A as ASSISTANT_VOICE_ID, C as CARTESIA_API_VERSION, a as COMMON_ABBREVIATION_EXPANSIONS, L as LEGACY_DEFAULT_VOICE_ID, P as ParseMarkdownToTextOptions, R as READING_VOICE_ID, S as SPEECH_BLANK_WORD, b as SpeechPronunciation, T as TTS_DEFAULT_SPEED, c as TTS_DEFAULT_VOLUME, d as TTS_MODEL_ID, e as TTS_PLAYBACK_BUFFER_SEC, f as TTS_SPEED_MAX, g as TTS_SPEED_MIN, h as TTS_STREAMING_BUFFER_SEC, V as VoicePurpose, n as normalizeSpeechAbbreviations, i as normalizeSpeechBlanks, p as parseMarkdownToText, r as resolveSpeed, j as resolveVoiceId } from './tts-config-JbF86KIt.cjs';
+
 declare const availableVoices: {
     id: string;
     name: string;
@@ -10,52 +12,6 @@ declare const allVoices: {
     description: string;
 }[];
 
-/**
- * Central Cartesia TTS configuration — the single source of truth for the model,
- * API version, system default voices, speed/volume baselines, and playback
- * buffering used by every in-app TTS surface (chat read-aloud, studio
- * read-aloud, voice playgrounds, the admin tester).
- *
- * Moved from matrx-frontend lib/cartesia/config.ts (P16 voices). The SDK-typed
- * `buildGenerationConfig` stays with the host's Cartesia client.
- */
-/** Current model + API version for all in-app TTS. */
-declare const TTS_MODEL_ID = "sonic-3.5";
-declare const CARTESIA_API_VERSION = "2026-08-14";
-/**
- * System default voices, used only when a user has not chosen their own.
- *   - reading   → Skylar (primary female; document / read-aloud)
- *   - assistant → Daniel (primary male; assistant replies)
- */
-declare const READING_VOICE_ID = "db6b0ed5-d5d3-463d-ae85-518a07d3c2b4";
-declare const ASSISTANT_VOICE_ID = "47c38ca4-5f35-497b-b1a3-415245fb35e1";
-/**
- * The pre-2026 hardcoded default voice. Treated as "unset" by resolveVoiceId so
- * users who never explicitly chose a voice transition to the new defaults.
- */
-declare const LEGACY_DEFAULT_VOICE_ID = "156fb8d2-335b-4950-9cb3-a2d33befec77";
-type VoicePurpose = "reading" | "assistant";
-/** generation_config.speed range (1.0 = original). Our chosen baseline is 1.2. */
-declare const TTS_SPEED_MIN = 0.6;
-declare const TTS_SPEED_MAX = 1.5;
-declare const TTS_DEFAULT_SPEED = 1.2;
-/** generation_config.volume range (1.0 = original). */
-declare const TTS_DEFAULT_VOLUME = 1;
-/**
- * Client-side player buffer (seconds). Higher than the old 0.25s, which
- * caused stream underruns heard as choppy "pauses"; tune in one place.
- */
-declare const TTS_PLAYBACK_BUFFER_SEC = 0.7;
-/**
- * Lower buffer for token-by-token streaming (real-time LLM speech) where
- * latency matters more — still well above the old 0.1s that stuttered.
- */
-declare const TTS_STREAMING_BUFFER_SEC = 0.3;
-/** A user's explicit voice preference wins; otherwise the purpose default. */
-declare function resolveVoiceId(userVoiceId: string | null | undefined, purpose: VoicePurpose): string;
-/** Clamp a stored speed to the valid generation_config range, else the default. */
-declare function resolveSpeed(userSpeed: number | null | undefined): number;
-
 type VoiceSetId = "cartesia" | "xai";
 interface VoiceOption {
     id: string;
@@ -75,108 +31,6 @@ declare function voiceDisplayName(set: VoiceSetId, value: unknown): string;
 /** True when `id` is one of the live voice conversation (xAI Realtime) voices. */
 declare function isLiveConversationVoice(id: string): boolean;
 
-/** A term→spoken-form substitution applied before any other speech parsing. */
-interface SpeechPronunciation {
-    from: string;
-    to: string;
-}
-interface ParseMarkdownToTextOptions {
-    /**
-     * Custom Dictionary pronunciation substitutions. Applied FIRST — before the
-     * built-in abbreviation/number normalization — so a user term like "AI" can
-     * supply its own spoken form instead of the generic "A I" readout. Pairs
-     * should be sorted longest-first by the caller so multi-word terms replace
-     * before substrings.
-     */
-    pronunciations?: SpeechPronunciation[];
-}
-/**
- * Long-form meanings retained as reference data for search, display, or future
- * accessibility features. TTS must not use these expansions: repeatedly
- * reading "Application Programming Interface" for API is not natural speech.
- */
-declare const COMMON_ABBREVIATION_EXPANSIONS: {
-    readonly AI: "Artificial Intelligence";
-    readonly API: "Application Programming Interface";
-    readonly HTTP: "Hypertext Transfer Protocol";
-    readonly HTTPS: "Hypertext Transfer Protocol Secure";
-    readonly URL: "Uniform Resource Locator";
-    readonly URI: "Uniform Resource Identifier";
-    readonly JSON: "JavaScript Object Notation";
-    readonly XML: "eXtensible Markup Language";
-    readonly CSS: "Cascading Style Sheets";
-    readonly HTML: "Hypertext Markup Language";
-    readonly JS: "JavaScript";
-    readonly TS: "TypeScript";
-    readonly SQL: "Structured Query Language";
-    readonly DB: "Database";
-    readonly UI: "User Interface";
-    readonly UX: "User Experience";
-    readonly SEO: "Search Engine Optimization";
-    readonly SDK: "Software Development Kit";
-    readonly CLI: "Command Line Interface";
-    readonly IDE: "Integrated Development Environment";
-    readonly JWT: "JSON Web Token";
-    readonly OAuth: "Open Authorization";
-    readonly REST: "Representational State Transfer";
-    readonly CRUD: "Create Read Update Delete";
-    readonly MVC: "Model View Controller";
-    readonly SPA: "Single Page Application";
-    readonly SSR: "Server Side Rendering";
-    readonly CSR: "Client Side Rendering";
-    readonly PWA: "Progressive Web App";
-    readonly DOM: "Document Object Model";
-    readonly BOM: "Browser Object Model";
-    readonly CDN: "Content Delivery Network";
-    readonly CMS: "Content Management System";
-    readonly ERP: "Enterprise Resource Planning";
-    readonly CRM: "Customer Relationship Management";
-    readonly SaaS: "Software as a Service";
-    readonly PaaS: "Platform as a Service";
-    readonly IaaS: "Infrastructure as a Service";
-    readonly VPN: "Virtual Private Network";
-    readonly LAN: "Local Area Network";
-    readonly WAN: "Wide Area Network";
-    readonly TCP: "Transmission Control Protocol";
-    readonly UDP: "User Datagram Protocol";
-    readonly IP: "Internet Protocol";
-    readonly DNS: "Domain Name System";
-    readonly FTP: "File Transfer Protocol";
-    readonly SMTP: "Simple Mail Transfer Protocol";
-    readonly POP3: "Post Office Protocol version 3";
-    readonly IMAP: "Internet Message Access Protocol";
-};
-/**
- * Normalize technical abbreviations for both active TTS lanes.
- *
- * Cartesia Sonic 3.5 recommends single-space character delimiters for forced
- * readout ("A P I"), and that plain-text form also works with the catalog's
- * Groq/Orpheus engine, which has no SSML/pronunciation-dictionary input. Word
- * acronyms stay conventional so models can say "json", "oh-auth", etc.
- */
-declare function normalizeSpeechAbbreviations(text: string): string;
-/**
- * The spoken word a fill-in-the-blank run of underscores becomes. One constant
- * so every speech surface reads the same thing — change it here (e.g. "what")
- * and TTS + generated flashcard fronts follow.
- */
-declare const SPEECH_BLANK_WORD = "blank";
-/**
- * Fill-in-the-blank underscores → a spoken word. A run of two-or-more
- * underscores (any length — "___", "_____"), including the spaced style
- * ("_ _ _"), reads as "blank" instead of the literal "underscore" every TTS
- * engine would otherwise say. Generic and reusable: the centralized speech
- * cleaner AND the generated flashcard spoken-fronts both route through this, so
- * a blank sounds identical everywhere.
- *
- * Guards (so it only ever touches real blanks):
- *  - bounded by non-alphanumerics → never mangles `snake_case` or `__dunder__`;
- *  - a line of ONLY underscores is a markdown thematic break (divider), left
- *    untouched here for the horizontal-rule rule to drop.
- */
-declare function normalizeSpeechBlanks(text: string): string;
-declare function parseMarkdownToText(markdown: string, options?: ParseMarkdownToTextOptions): string;
-
 /**
  * The languages read-aloud offers, with each language's own name — the union of the website's
  * voice-language setting and the extension's quick picker (which added Persian).
@@ -190,4 +44,4 @@ declare const READ_ALOUD_LANGUAGES: readonly ReadAloudLanguage[];
 /** The language for `code`, or English when the code is not offered. */
 declare function readAloudLanguage(code: string | null | undefined): ReadAloudLanguage;
 
-export { ASSISTANT_VOICE_ID, CARTESIA_API_VERSION, COMMON_ABBREVIATION_EXPANSIONS, LEGACY_DEFAULT_VOICE_ID, LIVE_CONVERSATION_SAMPLE_MODEL, LIVE_CONVERSATION_VOICES, type ParseMarkdownToTextOptions, READING_VOICE_ID, READ_ALOUD_LANGUAGES, type ReadAloudLanguage, SPEECH_BLANK_WORD, type SpeechPronunciation, TTS_DEFAULT_SPEED, TTS_DEFAULT_VOLUME, TTS_MODEL_ID, TTS_PLAYBACK_BUFFER_SEC, TTS_SPEED_MAX, TTS_SPEED_MIN, TTS_STREAMING_BUFFER_SEC, type VoiceOption, type VoicePurpose, type VoiceSetId, allVoices, availableVoices, isLiveConversationVoice, normalizeSpeechAbbreviations, normalizeSpeechBlanks, parseMarkdownToText, readAloudLanguage, resolveSpeed, resolveVoiceId, voiceDisplayName, voiceOptions, voiceSetDefaultLabel, voiceSetOf };
+export { LIVE_CONVERSATION_SAMPLE_MODEL, LIVE_CONVERSATION_VOICES, READ_ALOUD_LANGUAGES, type ReadAloudLanguage, type VoiceOption, type VoiceSetId, allVoices, availableVoices, isLiveConversationVoice, readAloudLanguage, voiceDisplayName, voiceOptions, voiceSetDefaultLabel, voiceSetOf };

package/dist/voices.d.ts

@@ -1,3 +1,5 @@
+export { A as ASSISTANT_VOICE_ID, C as CARTESIA_API_VERSION, a as COMMON_ABBREVIATION_EXPANSIONS, L as LEGACY_DEFAULT_VOICE_ID, P as ParseMarkdownToTextOptions, R as READING_VOICE_ID, S as SPEECH_BLANK_WORD, b as SpeechPronunciation, T as TTS_DEFAULT_SPEED, c as TTS_DEFAULT_VOLUME, d as TTS_MODEL_ID, e as TTS_PLAYBACK_BUFFER_SEC, f as TTS_SPEED_MAX, g as TTS_SPEED_MIN, h as TTS_STREAMING_BUFFER_SEC, V as VoicePurpose, n as normalizeSpeechAbbreviations, i as normalizeSpeechBlanks, p as parseMarkdownToText, r as resolveSpeed, j as resolveVoiceId } from './tts-config-JbF86KIt.js';
+
 declare const availableVoices: {
     id: string;
     name: string;
@@ -10,52 +12,6 @@ declare const allVoices: {
     description: string;
 }[];
 
-/**
- * Central Cartesia TTS configuration — the single source of truth for the model,
- * API version, system default voices, speed/volume baselines, and playback
- * buffering used by every in-app TTS surface (chat read-aloud, studio
- * read-aloud, voice playgrounds, the admin tester).
- *
- * Moved from matrx-frontend lib/cartesia/config.ts (P16 voices). The SDK-typed
- * `buildGenerationConfig` stays with the host's Cartesia client.
- */
-/** Current model + API version for all in-app TTS. */
-declare const TTS_MODEL_ID = "sonic-3.5";
-declare const CARTESIA_API_VERSION = "2026-08-14";
-/**
- * System default voices, used only when a user has not chosen their own.
- *   - reading   → Skylar (primary female; document / read-aloud)
- *   - assistant → Daniel (primary male; assistant replies)
- */
-declare const READING_VOICE_ID = "db6b0ed5-d5d3-463d-ae85-518a07d3c2b4";
-declare const ASSISTANT_VOICE_ID = "47c38ca4-5f35-497b-b1a3-415245fb35e1";
-/**
- * The pre-2026 hardcoded default voice. Treated as "unset" by resolveVoiceId so
- * users who never explicitly chose a voice transition to the new defaults.
- */
-declare const LEGACY_DEFAULT_VOICE_ID = "156fb8d2-335b-4950-9cb3-a2d33befec77";
-type VoicePurpose = "reading" | "assistant";
-/** generation_config.speed range (1.0 = original). Our chosen baseline is 1.2. */
-declare const TTS_SPEED_MIN = 0.6;
-declare const TTS_SPEED_MAX = 1.5;
-declare const TTS_DEFAULT_SPEED = 1.2;
-/** generation_config.volume range (1.0 = original). */
-declare const TTS_DEFAULT_VOLUME = 1;
-/**
- * Client-side player buffer (seconds). Higher than the old 0.25s, which
- * caused stream underruns heard as choppy "pauses"; tune in one place.
- */
-declare const TTS_PLAYBACK_BUFFER_SEC = 0.7;
-/**
- * Lower buffer for token-by-token streaming (real-time LLM speech) where
- * latency matters more — still well above the old 0.1s that stuttered.
- */
-declare const TTS_STREAMING_BUFFER_SEC = 0.3;
-/** A user's explicit voice preference wins; otherwise the purpose default. */
-declare function resolveVoiceId(userVoiceId: string | null | undefined, purpose: VoicePurpose): string;
-/** Clamp a stored speed to the valid generation_config range, else the default. */
-declare function resolveSpeed(userSpeed: number | null | undefined): number;
-
 type VoiceSetId = "cartesia" | "xai";
 interface VoiceOption {
     id: string;
@@ -75,108 +31,6 @@ declare function voiceDisplayName(set: VoiceSetId, value: unknown): string;
 /** True when `id` is one of the live voice conversation (xAI Realtime) voices. */
 declare function isLiveConversationVoice(id: string): boolean;
 
-/** A term→spoken-form substitution applied before any other speech parsing. */
-interface SpeechPronunciation {
-    from: string;
-    to: string;
-}
-interface ParseMarkdownToTextOptions {
-    /**
-     * Custom Dictionary pronunciation substitutions. Applied FIRST — before the
-     * built-in abbreviation/number normalization — so a user term like "AI" can
-     * supply its own spoken form instead of the generic "A I" readout. Pairs
-     * should be sorted longest-first by the caller so multi-word terms replace
-     * before substrings.
-     */
-    pronunciations?: SpeechPronunciation[];
-}
-/**
- * Long-form meanings retained as reference data for search, display, or future
- * accessibility features. TTS must not use these expansions: repeatedly
- * reading "Application Programming Interface" for API is not natural speech.
- */
-declare const COMMON_ABBREVIATION_EXPANSIONS: {
-    readonly AI: "Artificial Intelligence";
-    readonly API: "Application Programming Interface";
-    readonly HTTP: "Hypertext Transfer Protocol";
-    readonly HTTPS: "Hypertext Transfer Protocol Secure";
-    readonly URL: "Uniform Resource Locator";
-    readonly URI: "Uniform Resource Identifier";
-    readonly JSON: "JavaScript Object Notation";
-    readonly XML: "eXtensible Markup Language";
-    readonly CSS: "Cascading Style Sheets";
-    readonly HTML: "Hypertext Markup Language";
-    readonly JS: "JavaScript";
-    readonly TS: "TypeScript";
-    readonly SQL: "Structured Query Language";
-    readonly DB: "Database";
-    readonly UI: "User Interface";
-    readonly UX: "User Experience";
-    readonly SEO: "Search Engine Optimization";
-    readonly SDK: "Software Development Kit";
-    readonly CLI: "Command Line Interface";
-    readonly IDE: "Integrated Development Environment";
-    readonly JWT: "JSON Web Token";
-    readonly OAuth: "Open Authorization";
-    readonly REST: "Representational State Transfer";
-    readonly CRUD: "Create Read Update Delete";
-    readonly MVC: "Model View Controller";
-    readonly SPA: "Single Page Application";
-    readonly SSR: "Server Side Rendering";
-    readonly CSR: "Client Side Rendering";
-    readonly PWA: "Progressive Web App";
-    readonly DOM: "Document Object Model";
-    readonly BOM: "Browser Object Model";
-    readonly CDN: "Content Delivery Network";
-    readonly CMS: "Content Management System";
-    readonly ERP: "Enterprise Resource Planning";
-    readonly CRM: "Customer Relationship Management";
-    readonly SaaS: "Software as a Service";
-    readonly PaaS: "Platform as a Service";
-    readonly IaaS: "Infrastructure as a Service";
-    readonly VPN: "Virtual Private Network";
-    readonly LAN: "Local Area Network";
-    readonly WAN: "Wide Area Network";
-    readonly TCP: "Transmission Control Protocol";
-    readonly UDP: "User Datagram Protocol";
-    readonly IP: "Internet Protocol";
-    readonly DNS: "Domain Name System";
-    readonly FTP: "File Transfer Protocol";
-    readonly SMTP: "Simple Mail Transfer Protocol";
-    readonly POP3: "Post Office Protocol version 3";
-    readonly IMAP: "Internet Message Access Protocol";
-};
-/**
- * Normalize technical abbreviations for both active TTS lanes.
- *
- * Cartesia Sonic 3.5 recommends single-space character delimiters for forced
- * readout ("A P I"), and that plain-text form also works with the catalog's
- * Groq/Orpheus engine, which has no SSML/pronunciation-dictionary input. Word
- * acronyms stay conventional so models can say "json", "oh-auth", etc.
- */
-declare function normalizeSpeechAbbreviations(text: string): string;
-/**
- * The spoken word a fill-in-the-blank run of underscores becomes. One constant
- * so every speech surface reads the same thing — change it here (e.g. "what")
- * and TTS + generated flashcard fronts follow.
- */
-declare const SPEECH_BLANK_WORD = "blank";
-/**
- * Fill-in-the-blank underscores → a spoken word. A run of two-or-more
- * underscores (any length — "___", "_____"), including the spaced style
- * ("_ _ _"), reads as "blank" instead of the literal "underscore" every TTS
- * engine would otherwise say. Generic and reusable: the centralized speech
- * cleaner AND the generated flashcard spoken-fronts both route through this, so
- * a blank sounds identical everywhere.
- *
- * Guards (so it only ever touches real blanks):
- *  - bounded by non-alphanumerics → never mangles `snake_case` or `__dunder__`;
- *  - a line of ONLY underscores is a markdown thematic break (divider), left
- *    untouched here for the horizontal-rule rule to drop.
- */
-declare function normalizeSpeechBlanks(text: string): string;
-declare function parseMarkdownToText(markdown: string, options?: ParseMarkdownToTextOptions): string;
-
 /**
  * The languages read-aloud offers, with each language's own name — the union of the website's
  * voice-language setting and the extension's quick picker (which added Persian).
@@ -190,4 +44,4 @@ declare const READ_ALOUD_LANGUAGES: readonly ReadAloudLanguage[];
 /** The language for `code`, or English when the code is not offered. */
 declare function readAloudLanguage(code: string | null | undefined): ReadAloudLanguage;
 
-export { ASSISTANT_VOICE_ID, CARTESIA_API_VERSION, COMMON_ABBREVIATION_EXPANSIONS, LEGACY_DEFAULT_VOICE_ID, LIVE_CONVERSATION_SAMPLE_MODEL, LIVE_CONVERSATION_VOICES, type ParseMarkdownToTextOptions, READING_VOICE_ID, READ_ALOUD_LANGUAGES, type ReadAloudLanguage, SPEECH_BLANK_WORD, type SpeechPronunciation, TTS_DEFAULT_SPEED, TTS_DEFAULT_VOLUME, TTS_MODEL_ID, TTS_PLAYBACK_BUFFER_SEC, TTS_SPEED_MAX, TTS_SPEED_MIN, TTS_STREAMING_BUFFER_SEC, type VoiceOption, type VoicePurpose, type VoiceSetId, allVoices, availableVoices, isLiveConversationVoice, normalizeSpeechAbbreviations, normalizeSpeechBlanks, parseMarkdownToText, readAloudLanguage, resolveSpeed, resolveVoiceId, voiceDisplayName, voiceOptions, voiceSetDefaultLabel, voiceSetOf };
+export { LIVE_CONVERSATION_SAMPLE_MODEL, LIVE_CONVERSATION_VOICES, READ_ALOUD_LANGUAGES, type ReadAloudLanguage, type VoiceOption, type VoiceSetId, allVoices, availableVoices, isLiveConversationVoice, readAloudLanguage, voiceDisplayName, voiceOptions, voiceSetDefaultLabel, voiceSetOf };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-matrx/media",
-  "version": "0.14.0",
+  "version": "0.15.0",
   "description": "The AI Matrx media components kit: durable-ref-only renderers for images, video, audio and file thumbnails plus live TTS playback that just work with the Matrx file system — headless core hooks plus DOM bindings, wired to an injected MediaClient. A signed URL is a handoff, never an identity.",
   "type": "module",
   "main": "./dist/index.cjs",
@@ -138,6 +138,16 @@
         "default": "./dist/voices.cjs"
       }
     },
+    "./speech": {
+      "import": {
+        "types": "./dist/speech.d.ts",
+        "default": "./dist/speech.js"
+      },
+      "require": {
+        "types": "./dist/speech.d.cts",
+        "default": "./dist/speech.cjs"
+      }
+    },
     "./files/engine/api/assets": {
       "import": {
         "types": "./dist/files/engine/api/assets.d.ts",
@@ -751,6 +761,7 @@
     "./package.json": "./package.json"
   },
   "dependencies": {
+    "@cartesia/cartesia-js": "^4.2.0",
     "@supabase/postgrest-js": "^2.117.2",
     "@supabase/supabase-js": "^2.117.2",
     "dexie": "^4.4.6",
@@ -758,13 +769,13 @@
     "reselect": "^5.1.0",
     "tailwind-merge": "^3.7.0",
     "@ai-matrx/agents": "latest",
-    "@ai-matrx/associations": "latest",
     "@ai-matrx/content-ir": "latest",
     "@ai-matrx/data": "latest",
     "@ai-matrx/design-system": "latest",
     "@ai-matrx/kit": "latest",
-    "@ai-matrx/realtime": "latest",
-    "@ai-matrx/tap-target": "latest"
+    "@ai-matrx/associations": "latest",
+    "@ai-matrx/tap-target": "latest",
+    "@ai-matrx/realtime": "latest"
   },
   "peerDependencies": {
     "@reduxjs/toolkit": ">=2.0.0",