@ai-matrx/desktop-protocol 0.21.93 → 0.21.94

package/dist/client.cjs

@@ -147,7 +147,7 @@ var init_primitives = __esm({
     import_zod = require("zod");
     import_uuid = require("@ai-matrx/kit/uuid");
     PROTOCOL_MAJOR = 1;
-    PROTOCOL_MINOR = 87;
+    PROTOCOL_MINOR = 88;
     PROTOCOL_VERSION = `${PROTOCOL_MAJOR}.${PROTOCOL_MINOR}`;
     ProtocolVersion = import_zod.z.string().regex(/^[1-9][0-9]{0,2}\.[0-9]{1,3}$/);
     RequestId = import_zod.z.string().min(1).max(64).regex(/^[A-Za-z0-9_-]+$/);
@@ -243,7 +243,7 @@ var init_errors = __esm({
       // 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"
@@ -898,7 +898,7 @@ var init_session = __esm({
 });
 
 // src/schema/models.ts
-var import_zod10, ModelKey, TranscriptionModelId, MlRuntimeState, MlInstallPhase, MlInstallProgress, TranscriptionModelInfo, TranscriptSegment, MlStatusParams, MlStatusResult, ModelsInstallParams, ModelsInstallResult, ModelsTranscribeParams, ModelsTranscribeResult, TtsVoiceId, TtsVoice, TtsModelInfo, ModelsTtsStatusParams, ModelsTtsStatusResult, ModelsTtsInstallParams, ModelsTtsInstallResult, TtsBlendPart, ModelsTtsSpeakParams, ModelsTtsSpeakResult, TtsAudioChunk, ModelsTtsVoiceSaveParams, ModelsTtsVoiceSaveResult, ModelsTtsVoiceRemoveParams, ModelsTtsVoiceRemoveResult, MlProgress, TextModelId, TextModelInfo, UnusableTextModel, LlmServerState, LlmRunning, ModelsLlmStatusParams, ModelsLlmStatusResult, ModelsLlmInstallParams, ModelsLlmInstallResult, ModelsLlmStartParams, ModelsLlmStartResult, ModelsLlmStopParams, ModelsLlmStopResult, ModelsLlmChatParams, ModelsLlmChatResult, Sha256Hex, CatalogFile, CatalogModelEntry, LlamaServerBuild, MlCatalogSetParams, MlCatalogSetResult, DownloadKind, DownloadRow, ModelsDownloadsListParams, ModelsDownloadsListResult, ModelsDownloadsCancelParams, ModelsDownloadsCancelResult, LocalModelKind, LocalModelRow, ModelsLocalListParams, ModelsLocalListResult, ModelsLocalRemoveParams, ModelsLocalRemoveResult, MlServiceState, MlServiceStatus, ModelsStatusParams, ModelsStatusResult, ModelsProgressEvent;
+var import_zod10, ModelKey, TranscriptionModelId, MlRuntimeState, MlInstallPhase, MlInstallProgress, TranscriptionModelInfo, TranscriptSegment, MlStatusParams, MlStatusResult, ModelsInstallParams, ModelsInstallResult, ModelsTranscribeParams, ModelsTranscribeResult, TtsVoiceId, TtsVoice, TtsModelInfo, ModelsTtsStatusParams, ModelsTtsStatusResult, ModelsTtsInstallParams, ModelsTtsInstallResult, TtsBlendPart, ModelsTtsSpeakParams, ModelsTtsSpeakResult, TtsAudioChunk, ModelsTtsVoiceSaveParams, ModelsTtsVoiceSaveResult, ModelsTtsVoiceRemoveParams, ModelsTtsVoiceRemoveResult, MlProgress, TextModelId, TextModelInfo, UnusableTextModel, LlmServerState, LlmRunning, LlmSlot, LlmEmbeddingSlot, ModelsLlmStatusParams, ModelsLlmStatusResult, ModelsLlmInstallParams, ModelsLlmInstallResult, ModelsLlmStartParams, ModelsLlmStartResult, ModelsLlmStopParams, ModelsLlmStopResult, ModelsLlmChatParams, ModelsLlmChatResult, ModelsEmbedParams, EmbeddingVector, ModelsEmbedResult, Sha256Hex, CatalogFile, CatalogModelEntry, LlamaServerBuild, MlCatalogSetParams, MlCatalogSetResult, DownloadKind, DownloadRow, ModelsDownloadsListParams, ModelsDownloadsListResult, ModelsDownloadsCancelParams, ModelsDownloadsCancelResult, LocalModelKind, LocalModelRow, ModelsLocalListParams, ModelsLocalListResult, ModelsLocalRemoveParams, ModelsLocalRemoveResult, MlServiceState, MlServiceStatus, ModelsStatusParams, ModelsStatusResult, ModelsProgressEvent;
 var init_models = __esm({
   "src/schema/models.ts"() {
     "use strict";
@@ -1062,6 +1062,14 @@ var init_models = __esm({
       gpu_layers: import_zod10.z.int().min(-1).max(999).optional(),
       threads: import_zod10.z.int().min(1).max(512).optional()
     });
+    LlmSlot = import_zod10.z.enum(["chat", "embed"]);
+    LlmEmbeddingSlot = import_zod10.z.strictObject({
+      state: LlmServerState,
+      models: import_zod10.z.array(TextModelInfo),
+      running: LlmRunning.nullable(),
+      starting: TextModelId.nullable(),
+      error: import_zod10.z.string().max(300).nullable()
+    });
     ModelsLlmStatusParams = import_zod10.z.strictObject({});
     ModelsLlmStatusResult = import_zod10.z.strictObject({
       state: LlmServerState,
@@ -1077,7 +1085,9 @@ var init_models = __esm({
       /** 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_zod10.z.string().max(300).nullable()
+      error: import_zod10.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()
     });
     ModelsLlmInstallParams = import_zod10.z.strictObject({
       model: TextModelId,
@@ -1099,10 +1109,15 @@ var init_models = __esm({
       /** (1.48) CPU threads, computed by the app from the knobs. */
       threads: import_zod10.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_zod10.z.int().min(10).max(3600).optional()
+      load_timeout_s: import_zod10.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()
     });
     ModelsLlmStartResult = import_zod10.z.strictObject({ running: LlmRunning, elapsed_ms: nonNegInt });
-    ModelsLlmStopParams = import_zod10.z.strictObject({});
+    ModelsLlmStopParams = import_zod10.z.strictObject({
+      /** (1.88) Which slot to stop: `chat` (default) or `embed`. */
+      slot: LlmSlot.optional()
+    });
     ModelsLlmStopResult = import_zod10.z.strictObject({ stopped: import_zod10.z.boolean() });
     ModelsLlmChatParams = import_zod10.z.strictObject({
       model: TextModelId,
@@ -1112,6 +1127,23 @@ var init_models = __esm({
       window_bytes: import_zod10.z.int().min(4096).max(16777216).optional()
     });
     ModelsLlmChatResult = import_zod10.z.strictObject({ model: TextModelId, content_type: import_zod10.z.string().max(128) });
+    ModelsEmbedParams = import_zod10.z.strictObject({
+      model: TextModelId,
+      input: import_zod10.z.union([import_zod10.z.string().min(1).max(262144), import_zod10.z.array(import_zod10.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_zod10.z.boolean().optional()
+    });
+    EmbeddingVector = import_zod10.z.strictObject({
+      index: nonNegInt,
+      embedding: import_zod10.z.array(import_zod10.z.number()).min(1).max(16384)
+    });
+    ModelsEmbedResult = import_zod10.z.strictObject({
+      model: TextModelId,
+      /** The vector length (every row has it). */
+      dimensions: import_zod10.z.int().min(1).max(16384),
+      data: import_zod10.z.array(EmbeddingVector).min(1).max(256),
+      usage: import_zod10.z.strictObject({ prompt_tokens: nonNegInt, total_tokens: nonNegInt })
+    });
     Sha256Hex = import_zod10.z.string().regex(/^[0-9a-f]{64}$/);
     CatalogFile = import_zod10.z.strictObject({
       /** Path inside the model's folder (the artifact's path_in_repo). */
@@ -1157,7 +1189,9 @@ var init_models = __esm({
       /** (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(),
       /** (1.81) Text-to-speech models (Kokoro): what models.tts.* runs. */
-      speech: import_zod10.z.array(CatalogModelEntry).optional()
+      speech: import_zod10.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_zod10.z.array(CatalogModelEntry).optional()
     });
     MlCatalogSetResult = import_zod10.z.strictObject({ text: nonNegInt, transcription: nonNegInt });
     DownloadKind = import_zod10.z.enum(["speech_model", "text_model", "runtime"]);
@@ -5847,6 +5881,7 @@ var init_registry = __esm({
       "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"] },
@@ -6414,6 +6449,11 @@ var init_registry = __esm({
       ModelsLlmStopResult,
       ModelsLlmChatParams,
       ModelsLlmChatResult,
+      ModelsEmbedParams,
+      ModelsEmbedResult,
+      EmbeddingVector,
+      LlmSlot,
+      LlmEmbeddingSlot,
       DownloadKind,
       DownloadRow,
       ModelsDownloadsListParams,
@@ -7188,7 +7228,9 @@ 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"]]],