@ai-matrx/desktop-protocol 0.21.84 → 0.21.86

package/dist/client.d.cts

@@ -1,8 +1,8 @@
 import { BackoffOptions } from '@ai-matrx/data/realtime';
 export { RECONNECT_ALARM_ATTEMPTS } from '@ai-matrx/data/realtime';
 import { z } from 'zod';
-import { W as Welcome, D as DesktopProtocolError, C as CloseOutcome, O as OpName, a as OpParams, b as OpResult, c as OPS, S as StreamMeta, E as EndReason, d as EVENTS, e as ClientType, M as ModelsTranscribeResult, T as TranscriptionModelId, f as MlProgress } from './models-DMJ4I-2x.cjs';
-export { i as isDesktopProtocolError } from './models-DMJ4I-2x.cjs';
+import { W as Welcome, D as DesktopProtocolError, C as CloseOutcome, O as OpName, a as OpParams, b as OpResult, c as OPS, S as StreamMeta, E as EndReason, d as EVENTS, e as ClientType, M as ModelsTranscribeResult, T as TranscriptionModelId, f as MlProgress } from './models-CIpqsu-S.cjs';
+export { i as isDesktopProtocolError } from './models-CIpqsu-S.cjs';
 import { FrameHeader } from './frame.cjs';
 
 /** The subset of the WHATWG WebSocket the client uses — browser, React Native, `ws` and Workers all fit. */

package/dist/client.d.ts

@@ -1,8 +1,8 @@
 import { BackoffOptions } from '@ai-matrx/data/realtime';
 export { RECONNECT_ALARM_ATTEMPTS } from '@ai-matrx/data/realtime';
 import { z } from 'zod';
-import { W as Welcome, D as DesktopProtocolError, C as CloseOutcome, O as OpName, a as OpParams, b as OpResult, c as OPS, S as StreamMeta, E as EndReason, d as EVENTS, e as ClientType, M as ModelsTranscribeResult, T as TranscriptionModelId, f as MlProgress } from './models-DMJ4I-2x.js';
-export { i as isDesktopProtocolError } from './models-DMJ4I-2x.js';
+import { W as Welcome, D as DesktopProtocolError, C as CloseOutcome, O as OpName, a as OpParams, b as OpResult, c as OPS, S as StreamMeta, E as EndReason, d as EVENTS, e as ClientType, M as ModelsTranscribeResult, T as TranscriptionModelId, f as MlProgress } from './models-CIpqsu-S.js';
+export { i as isDesktopProtocolError } from './models-CIpqsu-S.js';
 import { FrameHeader } from './frame.js';
 
 /** The subset of the WHATWG WebSocket the client uses — browser, React Native, `ws` and Workers all fit. */

package/dist/client.js

@@ -1,7 +1,7 @@
 import {
   closeRule,
   knows
-} from "./chunk-server-CFON4SPE.js";
+} from "./chunk-server-4UMW5S3S.js";
 import {
   CloseCode,
   CoreToClientMsg,
@@ -12,7 +12,7 @@ import {
   PROTOCOL_MAJOR,
   PROTOCOL_MINOR,
   Welcome
-} from "./chunk-server-QLS73E3X.js";
+} from "./chunk-server-VYJA3PFP.js";
 import {
   DesktopProtocolError,
   FRAME_VERSION,
@@ -28,7 +28,7 @@ var registry = null;
 var registryLoad = null;
 function loadRegistry() {
   if (registry) return Promise.resolve(registry);
-  registryLoad ??= import("./chunk-server-OSCX7DKA.js").then((m) => registry = m);
+  registryLoad ??= import("./chunk-server-MOQ6N2L2.js").then((m) => registry = m);
   return registryLoad;
 }
 var WS_OPEN = 1;

package/dist/index.cjs

@@ -585,6 +585,16 @@ __export(src_exports, {
   ModelsStatusResult: () => ModelsStatusResult,
   ModelsTranscribeParams: () => ModelsTranscribeParams,
   ModelsTranscribeResult: () => ModelsTranscribeResult,
+  ModelsTtsInstallParams: () => ModelsTtsInstallParams,
+  ModelsTtsInstallResult: () => ModelsTtsInstallResult,
+  ModelsTtsSpeakParams: () => ModelsTtsSpeakParams,
+  ModelsTtsSpeakResult: () => ModelsTtsSpeakResult,
+  ModelsTtsStatusParams: () => ModelsTtsStatusParams,
+  ModelsTtsStatusResult: () => ModelsTtsStatusResult,
+  ModelsTtsVoiceRemoveParams: () => ModelsTtsVoiceRemoveParams,
+  ModelsTtsVoiceRemoveResult: () => ModelsTtsVoiceRemoveResult,
+  ModelsTtsVoiceSaveParams: () => ModelsTtsVoiceSaveParams,
+  ModelsTtsVoiceSaveResult: () => ModelsTtsVoiceSaveResult,
   NAMED_SCHEMAS: () => NAMED_SCHEMAS,
   NULL_SINCE: () => NULL_SINCE,
   NetAddress: () => NetAddress,
@@ -835,6 +845,11 @@ __export(src_exports, {
   TranscriptSegment: () => TranscriptSegment,
   TranscriptionModelId: () => TranscriptionModelId,
   TranscriptionModelInfo: () => TranscriptionModelInfo,
+  TtsAudioChunk: () => TtsAudioChunk,
+  TtsBlendPart: () => TtsBlendPart,
+  TtsModelInfo: () => TtsModelInfo,
+  TtsVoice: () => TtsVoice,
+  TtsVoiceId: () => TtsVoiceId,
   UnusableTextModel: () => UnusableTextModel,
   VERSIONED_RESULT_OPS: () => VERSIONED_RESULT_OPS,
   VaultFieldSummary: () => VaultFieldSummary,
@@ -945,7 +960,7 @@ module.exports = __toCommonJS(src_exports);
 var import_zod = require("zod");
 var import_uuid = require("@ai-matrx/kit/uuid");
 var PROTOCOL_MAJOR = 1;
-var PROTOCOL_MINOR = 80;
+var PROTOCOL_MINOR = 81;
 var PROTOCOL_VERSION = `${PROTOCOL_MAJOR}.${PROTOCOL_MINOR}`;
 var ProtocolVersion = import_zod.z.string().regex(/^[1-9][0-9]{0,2}\.[0-9]{1,3}$/);
 var RequestId = import_zod.z.string().min(1).max(64).regex(/^[A-Za-z0-9_-]+$/);
@@ -1817,13 +1832,79 @@ var ModelsTranscribeResult = import_zod10.z.strictObject({
   segments: import_zod10.z.array(TranscriptSegment).max(2e4),
   elapsed_ms: nonNegInt
 });
+var TtsVoiceId = import_zod10.z.string().min(1).max(64).regex(/^[a-z0-9][a-z0-9_]*$/);
+var TtsVoice = import_zod10.z.strictObject({
+  id: TtsVoiceId,
+  name: import_zod10.z.string().min(1).max(80),
+  /** Null for a saved voice (a blend has no single gender). */
+  gender: import_zod10.z.enum(["female", "male"]).nullable(),
+  /** The language's name (`American English`). */
+  language: import_zod10.z.string().min(1).max(64),
+  /** Kokoro's language code (`a`, `b`, `j`, `z`, `e`, `f`, `h`, `i`, `p`). */
+  lang_code: import_zod10.z.string().min(1).max(4),
+  /** The published quality grade (`A`, `B-`, `C+`); null for a saved voice. */
+  grade: import_zod10.z.string().max(4).nullable(),
+  traits: import_zod10.z.array(import_zod10.z.string().min(1).max(32)).max(8),
+  custom: import_zod10.z.boolean()
+});
+var TtsModelInfo = import_zod10.z.strictObject({ id: ModelKey, size_bytes: nonNegInt, installed: import_zod10.z.boolean() });
+var ModelsTtsStatusParams = import_zod10.z.strictObject({});
+var ModelsTtsStatusResult = import_zod10.z.strictObject({
+  /** Null when the catalog offers no speech model this computer can run. */
+  model: TtsModelInfo.nullable(),
+  runtime: MlRuntimeState,
+  /** Every voice: builtin first (in catalog order), then saved ones. Unbounded in practice; capped for the wire. */
+  voices: import_zod10.z.array(TtsVoice).max(1e3),
+  default_voice: TtsVoiceId,
+  /** Hertz of every chunk this engine produces (24 000 for Kokoro v1.0). */
+  sample_rate: import_zod10.z.int().min(8e3).max(192e3),
+  speed_min: import_zod10.z.number().min(0),
+  speed_max: import_zod10.z.number().min(0),
+  /** The last install or load failure, one human line. */
+  error: import_zod10.z.string().max(300).nullable()
+});
+var ModelsTtsInstallParams = import_zod10.z.strictObject({});
+var ModelsTtsInstallResult = import_zod10.z.strictObject({ model: ModelKey, elapsed_ms: nonNegInt });
+var TtsBlendPart = import_zod10.z.strictObject({ voice: TtsVoiceId, weight: import_zod10.z.number().min(0).max(1) });
+var ModelsTtsSpeakParams = import_zod10.z.strictObject({
+  text: import_zod10.z.string().min(1).max(1e5),
+  voice: TtsVoiceId.optional(),
+  blend: import_zod10.z.array(TtsBlendPart).min(2).max(5).optional(),
+  speed: import_zod10.z.number().min(0.5).max(2).default(1),
+  /** Text longer than this many characters is split at sentences into chunks (Matrx Local's ttsStreamingThreshold). */
+  chunk_chars: import_zod10.z.int().min(40).max(5e3).optional()
+});
+var ModelsTtsSpeakResult = import_zod10.z.strictObject({
+  chunks: nonNegInt,
+  duration_s: import_zod10.z.number().min(0),
+  sample_rate: import_zod10.z.int().min(8e3).max(192e3),
+  voice: import_zod10.z.string().min(1).max(200),
+  elapsed_ms: nonNegInt
+});
+var TtsAudioChunk = import_zod10.z.strictObject({
+  index: nonNegInt,
+  sample_rate: import_zod10.z.int().min(8e3).max(192e3),
+  pcm16_base64: import_zod10.z.string().max(8e6),
+  /** The text this chunk speaks. */
+  text: import_zod10.z.string().max(5e3)
+});
+var ModelsTtsVoiceSaveParams = import_zod10.z.strictObject({
+  name: import_zod10.z.string().min(1).max(80),
+  blend: import_zod10.z.array(TtsBlendPart).min(2).max(5).optional(),
+  path: AbsPath.optional()
+});
+var ModelsTtsVoiceSaveResult = import_zod10.z.strictObject({ voice: TtsVoice });
+var ModelsTtsVoiceRemoveParams = import_zod10.z.strictObject({ voice: TtsVoiceId });
+var ModelsTtsVoiceRemoveResult = import_zod10.z.strictObject({ removed: import_zod10.z.boolean() });
 var MlProgress = import_zod10.z.strictObject({
-  kind: import_zod10.z.enum(["install", "transcribe"]),
+  kind: import_zod10.z.enum(["install", "transcribe", "speak"]),
   /** 0–1 when known (transcribe: audio covered so far; install: bytes so far). */
   fraction: import_zod10.z.number().min(0).max(1).nullable(),
   install: MlInstallProgress.nullable(),
   /** The segment just transcribed (transcribe only): text appears as it is recognized. */
-  segment: TranscriptSegment.nullable()
+  segment: TranscriptSegment.nullable(),
+  /** (1.81) The chunk just synthesized (speak only). */
+  audio: TtsAudioChunk.optional()
 });
 var TextModelId = ModelKey;
 var TextModelInfo = import_zod10.z.strictObject({
@@ -1948,7 +2029,9 @@ var MlCatalogSetParams = import_zod10.z.strictObject({
   /** When the catalog was last changed (catalog_updated_at), for the service's log. */
   catalog_updated_at: import_zod10.z.string().max(40),
   /** (1.69) Language models on this computer that cannot run here, each with its one-line reason: models.llm.status lists them. */
-  unusable: import_zod10.z.array(UnusableTextModel).optional()
+  unusable: import_zod10.z.array(UnusableTextModel).optional(),
+  /** (1.81) Text-to-speech models (Kokoro): what models.tts.* runs. */
+  speech: import_zod10.z.array(CatalogModelEntry).optional()
 });
 var MlCatalogSetResult = import_zod10.z.strictObject({ text: nonNegInt, transcription: nonNegInt });
 var DownloadKind = import_zod10.z.enum(["speech_model", "text_model", "runtime"]);
@@ -5860,6 +5943,11 @@ var OPS = {
   "models.downloads.cancel": { params: ModelsDownloadsCancelParams, result: ModelsDownloadsCancelResult, kind: "unary", since: "1.45", errors: ["NOT_FOUND", "INTERNAL"] },
   "models.local.list": { params: ModelsLocalListParams, result: ModelsLocalListResult, kind: "unary", since: "1.74", errors: ["INTERNAL"] },
   "models.local.remove": { params: ModelsLocalRemoveParams, result: ModelsLocalRemoveResult, kind: "unary", since: "1.74", errors: ["NOT_FOUND", "INVALID_PARAMS", "RESOURCE_BUSY", "INTERNAL"] },
+  "models.tts.status": { params: ModelsTtsStatusParams, result: ModelsTtsStatusResult, kind: "unary", since: "1.81", errors: ["RESOURCE_BUSY", "INTERNAL"] },
+  "models.tts.install": { params: ModelsTtsInstallParams, result: ModelsTtsInstallResult, kind: "unary", since: "1.81", errors: ["NOT_FOUND", "RESOURCE_BUSY", "RESOURCE_GONE", "TIMEOUT", "CANCELLED", "INTERNAL"] },
+  "models.tts.speak": { params: ModelsTtsSpeakParams, result: ModelsTtsSpeakResult, kind: "unary", since: "1.81", errors: ["NOT_FOUND", "INVALID_PARAMS", "RESOURCE_BUSY", "RESOURCE_GONE", "TIMEOUT", "CANCELLED", "INTERNAL"] },
+  "models.tts.voice_save": { params: ModelsTtsVoiceSaveParams, result: ModelsTtsVoiceSaveResult, kind: "unary", since: "1.81", errors: ["NOT_FOUND", "INVALID_PARAMS", "PAYLOAD_TOO_LARGE", "CONFLICT", "RESOURCE_BUSY", "INTERNAL"] },
+  "models.tts.voice_remove": { params: ModelsTtsVoiceRemoveParams, result: ModelsTtsVoiceRemoveResult, kind: "unary", since: "1.81", errors: ["NOT_FOUND", "INVALID_PARAMS", "INTERNAL"] },
   // Desktop apps and their profiles (SPEC §6 apps); profiles are rows in users.desktop_app_profile.
   "apps.list": { params: AppsListParams, result: AppsListResult, kind: "unary", since: "1.2", errors: ["INTERNAL"] },
   "apps.profiles.list": { params: AppsProfilesListParams, result: AppsProfilesListResult, kind: "unary", since: "1.2", errors: ["NOT_FOUND", "INVALID_PARAMS", "INTERNAL"] },
@@ -6142,6 +6230,11 @@ var SERVICES = {
       "ml.download_cancel": { params: ModelsDownloadsCancelParams, result: ModelsDownloadsCancelResult },
       "ml.local_list": { params: ModelsLocalListParams, result: ModelsLocalListResult },
       "ml.local_remove": { params: ModelsLocalRemoveParams, result: ModelsLocalRemoveResult },
+      "ml.tts_status": { params: ModelsTtsStatusParams, result: ModelsTtsStatusResult },
+      "ml.tts_install": { params: ModelsTtsInstallParams, result: ModelsTtsInstallResult },
+      "ml.tts_speak": { params: ModelsTtsSpeakParams, result: ModelsTtsSpeakResult },
+      "ml.tts_voice_save": { params: ModelsTtsVoiceSaveParams, result: ModelsTtsVoiceSaveResult },
+      "ml.tts_voice_remove": { params: ModelsTtsVoiceRemoveParams, result: ModelsTtsVoiceRemoveResult },
       "ml.catalog_set": { params: MlCatalogSetParams, result: MlCatalogSetResult }
     }
   },
@@ -6349,6 +6442,21 @@ var NAMED_SCHEMAS = {
   ModelsLocalListResult,
   ModelsLocalRemoveParams,
   ModelsLocalRemoveResult,
+  TtsVoiceId,
+  TtsVoice,
+  TtsModelInfo,
+  ModelsTtsStatusParams,
+  ModelsTtsStatusResult,
+  ModelsTtsInstallParams,
+  ModelsTtsInstallResult,
+  TtsBlendPart,
+  ModelsTtsSpeakParams,
+  ModelsTtsSpeakResult,
+  TtsAudioChunk,
+  ModelsTtsVoiceSaveParams,
+  ModelsTtsVoiceSaveResult,
+  ModelsTtsVoiceRemoveParams,
+  ModelsTtsVoiceRemoveResult,
   CodingServiceCredentials,
   CodingHistoryBlocked,
   CodingHistoryFailure,
@@ -7013,7 +7121,8 @@ var import_versions = require("@ai-matrx/kit/versions");
 var FIELDS_SINCE = {
   LlmRunning: { context_length: "1.48", gpu_layers: "1.48", threads: "1.48" },
   TextModelInfo: { title: "1.69", problem: "1.69" },
-  MlCatalogSetParams: { unusable: "1.69" },
+  MlCatalogSetParams: { unusable: "1.69", speech: "1.81" },
+  MlProgress: { audio: "1.81" },
   ModelsLlmStartParams: { gpu_layers: "1.48", threads: "1.48", load_timeout_s: "1.48" },
   FsPatchResult: { edits_applied: "1.1", edits_failed: "1.1" },
   "ProtocolErrorBody.data": { edits_failed: "1.1" },
@@ -7362,7 +7471,7 @@ function forPeer(op, part, value, version) {
 var VERSIONED_RESULT_OPS = [.../* @__PURE__ */ new Set([...RESULT_PATH.keys(), ...WINDOW_RESULTS.keys(), ...NESTED_RESULTS.keys()])];
 
 // src/contract.ts
-var CONTRACT_SHA256 = "c416872bf8877ac2f210f9a4da3c7873a49879bfe07d4c58f4f29dc91e09be27";
+var CONTRACT_SHA256 = "f88cc7b499a4e0c6f8dd49d8c67b88d009218977e625330423421d231e1cca80";
 
 // src/close-policy.ts
 var CLOSE_RULES = {