@ai-matrx/desktop-protocol 0.21.93 → 0.21.94

package/dist/client.d.cts

@@ -1,9 +1,9 @@
 import { BackoffOptions } from '@ai-matrx/data/realtime';
 export { RECONNECT_ALARM_ATTEMPTS } from '@ai-matrx/data/realtime';
 import { z } from 'zod';
-import { D as DesktopProtocolError, O as OpName, a as OpParams, b as OpResult } from './models-BpWoHvNe.cjs';
+import { D as DesktopProtocolError, O as OpName, a as OpParams, b as OpResult } from './models-D2apXuiZ.cjs';
 import { Welcome, CloseOutcome, OPS, StreamMeta, EndReason, EVENTS, ClientType, ModelsTranscribeResult, TranscriptionModelId, MlProgress } from './index.cjs';
-export { i as isDesktopProtocolError } from './models-BpWoHvNe.cjs';
+export { i as isDesktopProtocolError } from './models-D2apXuiZ.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

@@ -1,9 +1,9 @@
 import { BackoffOptions } from '@ai-matrx/data/realtime';
 export { RECONNECT_ALARM_ATTEMPTS } from '@ai-matrx/data/realtime';
 import { z } from 'zod';
-import { D as DesktopProtocolError, O as OpName, a as OpParams, b as OpResult } from './models-BpWoHvNe.js';
+import { D as DesktopProtocolError, O as OpName, a as OpParams, b as OpResult } from './models-D2apXuiZ.js';
 import { Welcome, CloseOutcome, OPS, StreamMeta, EndReason, EVENTS, ClientType, ModelsTranscribeResult, TranscriptionModelId, MlProgress } from './index.js';
-export { i as isDesktopProtocolError } from './models-BpWoHvNe.js';
+export { i as isDesktopProtocolError } from './models-D2apXuiZ.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-MCBYY6QK.js";
+} from "./chunk-server-Q7F4IE6F.js";
 import {
   CloseCode,
   CoreToClientMsg,
@@ -12,7 +12,7 @@ import {
   PROTOCOL_MAJOR,
   PROTOCOL_MINOR,
   Welcome
-} from "./chunk-server-HBZSBBGI.js";
+} from "./chunk-server-5MKSD7FQ.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-FJWACRBP.js").then((m) => registry = m);
+  registryLoad ??= import("./chunk-server-PXHL7V3H.js").then((m) => registry = m);
   return registryLoad;
 }
 var WS_OPEN = 1;

package/dist/index.cjs

@@ -449,6 +449,7 @@ __export(src_exports, {
   EVENTS: () => EVENTS,
   EXCLUDED_TOOLS: () => EXCLUDED_TOOLS,
   ElsewhereInstance: () => ElsewhereInstance,
+  EmbeddingVector: () => EmbeddingVector,
   EndMsg: () => EndMsg,
   EndReason: () => EndReason,
   EntitiesNerParams: () => EntitiesNerParams,
@@ -584,8 +585,10 @@ __export(src_exports, {
   LEGACY_DISPATCHERS: () => LEGACY_DISPATCHERS,
   Limits: () => Limits,
   LlamaServerBuild: () => LlamaServerBuild,
+  LlmEmbeddingSlot: () => LlmEmbeddingSlot,
   LlmRunning: () => LlmRunning,
   LlmServerState: () => LlmServerState,
+  LlmSlot: () => LlmSlot,
   LocalModelKind: () => LocalModelKind,
   LocalModelRow: () => LocalModelRow,
   LocationGetParams: () => LocationGetParams,
@@ -624,6 +627,8 @@ __export(src_exports, {
   ModelsDownloadsCancelResult: () => ModelsDownloadsCancelResult,
   ModelsDownloadsListParams: () => ModelsDownloadsListParams,
   ModelsDownloadsListResult: () => ModelsDownloadsListResult,
+  ModelsEmbedParams: () => ModelsEmbedParams,
+  ModelsEmbedResult: () => ModelsEmbedResult,
   ModelsInstallParams: () => ModelsInstallParams,
   ModelsInstallResult: () => ModelsInstallResult,
   ModelsLlmChatParams: () => ModelsLlmChatParams,
@@ -1030,7 +1035,7 @@ 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 = 87;
+var PROTOCOL_MINOR = 88;
 var PROTOCOL_VERSION = `${PROTOCOL_MAJOR}.${PROTOCOL_MINOR}`;
 var ProtocolVersion = import_zod.z.string().regex(/^[1-9][0-9]{0,2}\.[0-9]{1,3}$/);
 var RequestId = import_zod.z.string().min(1).max(64).regex(/^[A-Za-z0-9_-]+$/);
@@ -1119,7 +1124,7 @@ var ErrorReason = import_zod2.z.enum([
   // 1.34 (REASONS_SINCE): the per-device tool policy (src/tools/policy.ts)
   "tool_disabled_on_device",
   // AUTH_FORBIDDEN: the person turned this tool off for agents on this computer (Agents › Tools)
-  // 1.76 (REASONS_SINCE): models.llm.chat
+  // 1.76 (REASONS_SINCE): models.llm.chat (1.88: models.embed names the same two)
   "model_not_running",
   // NOT_FOUND (nothing runs) or RESOURCE_BUSY (another model runs): start this model first
   "model_server_unreachable"
@@ -2060,6 +2065,14 @@ var LlmRunning = import_zod11.z.strictObject({
   gpu_layers: import_zod11.z.int().min(-1).max(999).optional(),
   threads: import_zod11.z.int().min(1).max(512).optional()
 });
+var LlmSlot = import_zod11.z.enum(["chat", "embed"]);
+var LlmEmbeddingSlot = import_zod11.z.strictObject({
+  state: LlmServerState,
+  models: import_zod11.z.array(TextModelInfo),
+  running: LlmRunning.nullable(),
+  starting: TextModelId.nullable(),
+  error: import_zod11.z.string().max(300).nullable()
+});
 var ModelsLlmStatusParams = import_zod11.z.strictObject({});
 var ModelsLlmStatusResult = import_zod11.z.strictObject({
   state: LlmServerState,
@@ -2075,7 +2088,9 @@ var ModelsLlmStatusResult = import_zod11.z.strictObject({
   /** Non-null while the server program or a model's weights are being downloaded. */
   install: MlInstallProgress.nullable(),
   /** Why the last start or download failed, one human line; null otherwise. */
-  error: import_zod11.z.string().max(300).nullable()
+  error: import_zod11.z.string().max(300).nullable(),
+  /** (1.88) The embeddings slot: the same model service's second llama-server, in embeddings mode, beside the chat model. */
+  embedding: LlmEmbeddingSlot.optional()
 });
 var ModelsLlmInstallParams = import_zod11.z.strictObject({
   model: TextModelId,
@@ -2097,10 +2112,15 @@ var ModelsLlmStartParams = import_zod11.z.strictObject({
   /** (1.48) CPU threads, computed by the app from the knobs. */
   threads: import_zod11.z.int().min(1).max(512).optional(),
   /** (1.48) Seconds to wait for the model to load, computed by the app from the knobs. */
-  load_timeout_s: import_zod11.z.int().min(10).max(3600).optional()
+  load_timeout_s: import_zod11.z.int().min(10).max(3600).optional(),
+  /** (1.88) Which slot: `chat` (default) or `embed` (the embedding model, beside the chat model; neither stops the other). */
+  slot: LlmSlot.optional()
 });
 var ModelsLlmStartResult = import_zod11.z.strictObject({ running: LlmRunning, elapsed_ms: nonNegInt });
-var ModelsLlmStopParams = import_zod11.z.strictObject({});
+var ModelsLlmStopParams = import_zod11.z.strictObject({
+  /** (1.88) Which slot to stop: `chat` (default) or `embed`. */
+  slot: LlmSlot.optional()
+});
 var ModelsLlmStopResult = import_zod11.z.strictObject({ stopped: import_zod11.z.boolean() });
 var ModelsLlmChatParams = import_zod11.z.strictObject({
   model: TextModelId,
@@ -2110,6 +2130,23 @@ var ModelsLlmChatParams = import_zod11.z.strictObject({
   window_bytes: import_zod11.z.int().min(4096).max(16777216).optional()
 });
 var ModelsLlmChatResult = import_zod11.z.strictObject({ model: TextModelId, content_type: import_zod11.z.string().max(128) });
+var ModelsEmbedParams = import_zod11.z.strictObject({
+  model: TextModelId,
+  input: import_zod11.z.union([import_zod11.z.string().min(1).max(262144), import_zod11.z.array(import_zod11.z.string().min(1).max(262144)).min(1).max(256)]),
+  /** Start the model in the embeddings slot first when nothing runs there (a different embedding model running is never stopped). */
+  start_if_stopped: import_zod11.z.boolean().optional()
+});
+var EmbeddingVector = import_zod11.z.strictObject({
+  index: nonNegInt,
+  embedding: import_zod11.z.array(import_zod11.z.number()).min(1).max(16384)
+});
+var ModelsEmbedResult = import_zod11.z.strictObject({
+  model: TextModelId,
+  /** The vector length (every row has it). */
+  dimensions: import_zod11.z.int().min(1).max(16384),
+  data: import_zod11.z.array(EmbeddingVector).min(1).max(256),
+  usage: import_zod11.z.strictObject({ prompt_tokens: nonNegInt, total_tokens: nonNegInt })
+});
 var Sha256Hex = import_zod11.z.string().regex(/^[0-9a-f]{64}$/);
 var CatalogFile = import_zod11.z.strictObject({
   /** Path inside the model's folder (the artifact's path_in_repo). */
@@ -2155,7 +2192,9 @@ var MlCatalogSetParams = import_zod11.z.strictObject({
   /** (1.69) Language models on this computer that cannot run here, each with its one-line reason: models.llm.status lists them. */
   unusable: import_zod11.z.array(UnusableTextModel).optional(),
   /** (1.81) Text-to-speech models (Kokoro): what models.tts.* runs. */
-  speech: import_zod11.z.array(CatalogModelEntry).optional()
+  speech: import_zod11.z.array(CatalogModelEntry).optional(),
+  /** (1.88) Embedding models (GGUF, run by llama-server with --embeddings in the embeddings slot): what models.embed runs. */
+  embedding: import_zod11.z.array(CatalogModelEntry).optional()
 });
 var MlCatalogSetResult = import_zod11.z.strictObject({ text: nonNegInt, transcription: nonNegInt });
 var DownloadKind = import_zod11.z.enum(["speech_model", "text_model", "runtime"]);
@@ -6491,6 +6530,7 @@ var OPS = {
   "models.llm.start": { params: ModelsLlmStartParams, result: ModelsLlmStartResult, kind: "unary", since: "1.31", errors: ["NOT_FOUND", "RESOURCE_BUSY", "RESOURCE_GONE", "TIMEOUT", "CANCELLED", "PLATFORM_UNSUPPORTED", "INTERNAL"] },
   "models.llm.stop": { params: ModelsLlmStopParams, result: ModelsLlmStopResult, kind: "unary", since: "1.31", errors: ["INTERNAL"] },
   "models.llm.chat": { params: ModelsLlmChatParams, result: ModelsLlmChatResult, kind: "stream", since: "1.76", errors: ["NOT_FOUND", "RESOURCE_BUSY", "RESOURCE_GONE", "INVALID_PARAMS", "TIMEOUT", "CANCELLED", "RATE_LIMITED", "INTERNAL"] },
+  "models.embed": { params: ModelsEmbedParams, result: ModelsEmbedResult, kind: "unary", since: "1.88", errors: ["NOT_FOUND", "RESOURCE_BUSY", "RESOURCE_GONE", "INVALID_PARAMS", "TIMEOUT", "CANCELLED", "INTERNAL"] },
   "models.downloads.list": { params: ModelsDownloadsListParams, result: ModelsDownloadsListResult, kind: "unary", since: "1.45", errors: ["INTERNAL"] },
   "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"] },
@@ -7058,6 +7098,11 @@ var NAMED_SCHEMAS = {
   ModelsLlmStopResult,
   ModelsLlmChatParams,
   ModelsLlmChatResult,
+  ModelsEmbedParams,
+  ModelsEmbedResult,
+  EmbeddingVector,
+  LlmSlot,
+  LlmEmbeddingSlot,
   DownloadKind,
   DownloadRow,
   ModelsDownloadsListParams,
@@ -7770,9 +7815,11 @@ 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", speech: "1.81" },
+  MlCatalogSetParams: { unusable: "1.69", speech: "1.81", embedding: "1.88" },
+  ModelsLlmStatusResult: { embedding: "1.88" },
+  ModelsLlmStopParams: { slot: "1.88" },
   MlProgress: { audio: "1.81" },
-  ModelsLlmStartParams: { gpu_layers: "1.48", threads: "1.48", load_timeout_s: "1.48" },
+  ModelsLlmStartParams: { gpu_layers: "1.48", threads: "1.48", load_timeout_s: "1.48", slot: "1.88" },
   FsPatchResult: { edits_applied: "1.1", edits_failed: "1.1" },
   "ProtocolErrorBody.data": { edits_failed: "1.1" },
   "RelayCallResponse|0": { protocol_version: "1.1" },
@@ -7956,7 +8003,9 @@ var RESULT_PATH = /* @__PURE__ */ new Map([
   // 1.67: Matrx Local's local_mac_apps result fields.
   ["photos.get", "PhotosGetResult"],
   ["location.get", "LocationGetResult"],
-  ["account.agents.restore_version", "AccountAgentsRestoreVersionResult"]
+  ["account.agents.restore_version", "AccountAgentsRestoreVersionResult"],
+  // 1.88: the embeddings slot beside the chat model.
+  ["models.llm.status", "ModelsLlmStatusResult"]
 ]);
 var NESTED_RESULTS = /* @__PURE__ */ new Map([
   ["apps.chats.list", [["chats", "ChatRow"]]],
@@ -8120,7 +8169,7 @@ 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 = "6bafb641835624f0c1624f585cf5cd59d4594107b4661936bab4b771e54924e2";
+var CONTRACT_SHA256 = "24650f34e966ce74c997fefcc7e6474b71fb9737ea81e81bfd1513a5e5018768";
 
 // src/close-policy.ts
 var CLOSE_RULES = {