@qvac/sdk 0.13.5 → 0.14.1

package/dist/server/bare/plugins/{onnx-ocr → ggml-ocr}/plugin.js

@@ -2,34 +2,36 @@ import { definePlugin, defineHandler, ocrStreamRequestSchema, ocrStreamResponseS
 import { ModelLoadFailedError } from "../../../../utils/errors-server.js";
 import { hyperdriveUrlSchema } from "../../../../schemas/load-model.js";
 import { createStreamLogger, registerAddonLogger } from "../../../../logging/index.js";
-import { ONNXOcr } from "@qvac/ocr-onnx";
+import ocrAddonLogging from "@qvac/ocr-ggml/addonLogging.js";
+import { OcrGgml } from "@qvac/ocr-ggml";
 import { ocr } from "./ops/ocr-stream.js";
 import { attachModelExecutionMs } from "../../../../profiling/model-execution.js";
-import { OCR_CRAFT_DETECTOR } from "../../../../models/registry/index.js";
-const OCR_DETECTOR_FILENAME = "detector_craft.onnx";
+import { OCR_CRAFT } from "../../../../models/registry/index.js";
+const OCR_DETECTOR_FILENAME = "craft_mlt_25k.gguf";
 function deriveDetectorSource(modelSrc) {
     if (modelSrc.startsWith("pear://")) {
         const { key } = hyperdriveUrlSchema.parse(modelSrc);
         return `pear://${key}/${OCR_DETECTOR_FILENAME}`;
     }
     if (modelSrc.startsWith("registry://")) {
-        return OCR_CRAFT_DETECTOR.src;
+        return OCR_CRAFT.src;
     }
     return undefined;
 }
 function createOCRModel(modelId, detectorPath, recognizerPath, ocrConfig) {
-    const logger = createStreamLogger(modelId, ModelType.onnxOcr);
-    registerAddonLogger(modelId, ModelType.onnxOcr, logger);
+    const logger = createStreamLogger(modelId, ModelType.ggmlOcr);
+    registerAddonLogger(modelId, ModelType.ggmlOcr, logger);
     const params = {
         pathDetector: detectorPath,
         pathRecognizer: recognizerPath,
         langList: ocrConfig.langList || ["en"],
-        useGPU: ocrConfig.useGPU ?? true,
-        ...(ocrConfig.timeout !== undefined && { timeout: ocrConfig.timeout }),
-        ...(ocrConfig.pipelineMode !== undefined && {
-            pipelineMode: ocrConfig.pipelineMode,
+        ...(ocrConfig.pipelineType !== undefined && {
+            pipelineType: ocrConfig.pipelineType,
         }),
         ...(ocrConfig.magRatio !== undefined && { magRatio: ocrConfig.magRatio }),
+        ...(ocrConfig.canvasSize !== undefined && {
+            canvasSize: ocrConfig.canvasSize,
+        }),
         ...(ocrConfig.defaultRotationAngles !== undefined && {
             defaultRotationAngles: ocrConfig.defaultRotationAngles,
         }),
@@ -42,11 +44,12 @@ function createOCRModel(modelId, detectorPath, recognizerPath, ocrConfig) {
         ...(ocrConfig.recognizerBatchSize !== undefined && {
             recognizerBatchSize: ocrConfig.recognizerBatchSize,
         }),
-        ...(ocrConfig.decodingMethod !== undefined && {
-            decodingMethod: ocrConfig.decodingMethod,
+        ...(ocrConfig.nThreads !== undefined && { nThreads: ocrConfig.nThreads }),
+        ...(ocrConfig.backendDevice !== undefined && {
+            backendDevice: ocrConfig.backendDevice,
         }),
-        ...(ocrConfig.straightenPages !== undefined && {
-            straightenPages: ocrConfig.straightenPages,
+        ...(ocrConfig.gpuDevice !== undefined && {
+            gpuDevice: ocrConfig.gpuDevice,
         }),
     };
     const args = {
@@ -54,12 +57,12 @@ function createOCRModel(modelId, detectorPath, recognizerPath, ocrConfig) {
         params,
         opts: { stats: true },
     };
-    const model = new ONNXOcr(args);
+    const model = new OcrGgml(args);
     return model;
 }
 export const ocrPlugin = definePlugin({
-    modelType: ModelType.onnxOcr,
-    displayName: "OCR (ONNX)",
+    modelType: ModelType.ggmlOcr,
+    displayName: "OCR (GGML)",
     addonPackage: ADDON_OCR,
     loadConfigSchema: ocrConfigSchema,
     async resolveConfig(cfg, ctx) {
@@ -89,7 +92,7 @@ export const ocrPlugin = definePlugin({
             requestSchema: ocrStreamRequestSchema,
             responseSchema: ocrStreamResponseSchema,
             streaming: true,
-            // ONNX OCR does not expose a cancel surface — SDK falls back
+            // GGML OCR does not expose a cancel surface — SDK falls back
             // to soft-cancel.
             cancel: { scope: "none" },
             handler: async function* (request) {
@@ -121,4 +124,8 @@ export const ocrPlugin = definePlugin({
             },
         }),
     },
+    logging: {
+        module: ocrAddonLogging,
+        namespace: ModelType.ggmlOcr,
+    },
 });

package/dist/server/bare/plugins/index.d.ts

@@ -5,7 +5,7 @@ export { bciPlugin } from "./bci-whispercpp-transcription/plugin";
 export { parakeetPlugin } from "./parakeet-transcription/plugin";
 export { nmtPlugin } from "./nmtcpp-translation/plugin";
 export { ttsPlugin } from "./tts-ggml/plugin";
-export { ocrPlugin } from "./onnx-ocr/plugin";
+export { ocrPlugin } from "./ggml-ocr/plugin";
 export { diffusionPlugin } from "./sdcpp-generation/plugin";
 export { vlaPlugin } from "./ggml-vla/plugin";
 export { classificationPlugin } from "./ggml-classification/plugin";

package/dist/server/bare/plugins/index.js

@@ -5,7 +5,7 @@ export { bciPlugin } from "./bci-whispercpp-transcription/plugin.js";
 export { parakeetPlugin } from "./parakeet-transcription/plugin.js";
 export { nmtPlugin } from "./nmtcpp-translation/plugin.js";
 export { ttsPlugin } from "./tts-ggml/plugin.js";
-export { ocrPlugin } from "./onnx-ocr/plugin.js";
+export { ocrPlugin } from "./ggml-ocr/plugin.js";
 export { diffusionPlugin } from "./sdcpp-generation/plugin.js";
 export { vlaPlugin } from "./ggml-vla/plugin.js";
 export { classificationPlugin } from "./ggml-classification/plugin.js";

package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stats.d.ts

@@ -0,0 +1,10 @@
+import type { LlmStats } from "../../../types/addon-responses";
+export declare function normalizeCompletionStats(stats: LlmStats | undefined): {
+    timeToFirstToken?: number | undefined;
+    tokensPerSecond?: number | undefined;
+    cacheTokens?: number | undefined;
+    promptTokens?: number | undefined;
+    generatedTokens?: number | undefined;
+    backendDevice?: "cpu" | "gpu" | undefined;
+} | undefined;
+//# sourceMappingURL=completion-stats.d.ts.map

package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stats.js

@@ -0,0 +1,29 @@
+function finiteNumber(value) {
+    return value !== undefined && Number.isFinite(value) ? value : undefined;
+}
+export function normalizeCompletionStats(stats) {
+    if (!stats)
+        return undefined;
+    const timeToFirstToken = finiteNumber(stats.TTFT);
+    const tokensPerSecond = finiteNumber(stats.TPS);
+    const cacheTokens = finiteNumber(stats.CacheTokens);
+    const promptTokens = finiteNumber(stats.promptTokens);
+    const generatedTokens = finiteNumber(stats.generatedTokens);
+    const normalized = {
+        ...(timeToFirstToken !== undefined && { timeToFirstToken }),
+        ...(tokensPerSecond !== undefined && { tokensPerSecond }),
+        ...(cacheTokens !== undefined && { cacheTokens }),
+        ...(promptTokens !== undefined && { promptTokens }),
+        ...(generatedTokens !== undefined && { generatedTokens }),
+        ...(stats.backendDevice !== undefined && { backendDevice: stats.backendDevice }),
+    };
+    if (timeToFirstToken === undefined &&
+        tokensPerSecond === undefined &&
+        cacheTokens === undefined &&
+        promptTokens === undefined &&
+        generatedTokens === undefined &&
+        stats.backendDevice === undefined) {
+        return undefined;
+    }
+    return normalized;
+}

package/dist/server/bare/plugins/llamacpp-completion/ops/completion-stream.js

@@ -11,7 +11,8 @@ import { buildAutoCacheSaveHistory } from "../../../../utils/index.js";
 import { getServerLogger } from "../../../../../logging/index.js";
 import { AttachmentNotFoundError } from "../../../../../utils/errors-server.js";
 import { nowMs } from "../../../../../profiling/index.js";
-import { buildStreamResult, hasDefinedValues, } from "../../../../../profiling/model-execution.js";
+import { buildStreamResult } from "../../../../../profiling/model-execution.js";
+import { normalizeCompletionStats } from "./completion-stats.js";
 import fs from "bare-fs";
 const logger = getServerLogger();
 /**
@@ -195,28 +196,9 @@ async function* processModelResponse(model, messagesToSend, tools, generationPar
         toolCallsResult = toolCalls;
     }
     const responseWithStats = response;
-    const stats = {
-        ...(responseWithStats.stats?.TTFT !== undefined && {
-            timeToFirstToken: responseWithStats.stats.TTFT,
-        }),
-        ...(responseWithStats.stats?.TPS !== undefined && {
-            tokensPerSecond: responseWithStats.stats.TPS,
-        }),
-        ...(responseWithStats.stats?.CacheTokens !== undefined && {
-            cacheTokens: responseWithStats.stats.CacheTokens,
-        }),
-        ...(responseWithStats.stats?.promptTokens !== undefined && {
-            promptTokens: responseWithStats.stats.promptTokens,
-        }),
-        ...(responseWithStats.stats?.generatedTokens !== undefined && {
-            generatedTokens: responseWithStats.stats.generatedTokens,
-        }),
-        ...(responseWithStats.stats?.backendDevice !== undefined && {
-            backendDevice: responseWithStats.stats.backendDevice,
-        }),
-    };
+    const stats = normalizeCompletionStats(responseWithStats.stats);
     return {
-        ...buildStreamResult(modelExecutionMs, hasDefinedValues(stats) ? stats : undefined),
+        ...buildStreamResult(modelExecutionMs, stats),
         toolCalls: toolCallsResult,
         responseText: accumulatedText,
         producedTokens,

package/dist/server/bare/plugins/llamacpp-completion/plugin.d.ts

@@ -39,7 +39,7 @@ export declare const llmPlugin: {
         }>>;
         "tensor-split": import("zod").ZodOptional<import("zod").ZodString>;
         openclCacheDir: import("zod").ZodOptional<import("zod").ZodString>;
-        reasoning_budget: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodLiteral<-1>, import("zod").ZodLiteral<0>]>>;
+        reasoning_budget: import("zod").ZodOptional<import("zod").ZodNumber>;
         projectionModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
             name: import("zod").ZodOptional<import("zod").ZodString>;
@@ -65,7 +65,7 @@ export declare const llmPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -77,6 +77,11 @@ export declare const llmPlugin: {
                 classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
+        image_tile_mode: import("zod").ZodOptional<import("zod").ZodEnum<{
+            disabled: "disabled";
+            batched: "batched";
+            sequential: "sequential";
+        }>>;
     }, import("zod/v4/core").$strip>;
     resolveConfig(cfg: LlmConfigInput, ctx: ResolveContext): Promise<{
         config: {
@@ -105,7 +110,8 @@ export declare const llmPlugin: {
             "split-mode"?: "none" | "layer" | "row" | undefined;
             "tensor-split"?: string | undefined;
             openclCacheDir?: string | undefined;
-            reasoning_budget?: 0 | -1 | undefined;
+            reasoning_budget?: number | undefined;
+            image_tile_mode?: "disabled" | "batched" | "sequential" | undefined;
         };
         artifacts?: never;
     } | {
@@ -135,7 +141,8 @@ export declare const llmPlugin: {
             "split-mode"?: "none" | "layer" | "row" | undefined;
             "tensor-split"?: string | undefined;
             openclCacheDir?: string | undefined;
-            reasoning_budget?: 0 | -1 | undefined;
+            reasoning_budget?: number | undefined;
+            image_tile_mode?: "disabled" | "batched" | "sequential" | undefined;
         };
         artifacts: {
             projectionModelPath: string;
@@ -184,7 +191,8 @@ export declare const llmPlugin: {
                 frequency_penalty: import("zod").ZodOptional<import("zod").ZodNumber>;
                 presence_penalty: import("zod").ZodOptional<import("zod").ZodNumber>;
                 repeat_penalty: import("zod").ZodOptional<import("zod").ZodNumber>;
-                reasoning_budget: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodLiteral<-1>, import("zod").ZodLiteral<0>]>>;
+                reasoning_budget: import("zod").ZodOptional<import("zod").ZodNumber>;
+                remove_thinking_from_context: import("zod").ZodOptional<import("zod").ZodBoolean>;
             }, import("zod/v4/core").$strict>>;
             captureThinking: import("zod").ZodOptional<import("zod").ZodBoolean>;
             emitRawDeltas: import("zod").ZodOptional<import("zod").ZodBoolean>;
@@ -436,7 +444,7 @@ export declare const llmPlugin: {
                 epochs_completed: import("zod").ZodNumber;
             }, import("zod/v4/core").$strict>>;
         }, import("zod/v4/core").$strict>>;
-        translate: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodUnion<readonly [import("zod").ZodObject<{
+        translate: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodDiscriminatedUnion<[import("zod").ZodObject<{
             modelId: import("zod").ZodString;
             text: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodArray<import("zod").ZodString>]>;
             stream: import("zod").ZodBoolean;
@@ -459,7 +467,7 @@ export declare const llmPlugin: {
             context: import("zod").ZodOptional<import("zod").ZodString>;
             type: import("zod").ZodLiteral<"translate">;
             requestId: import("zod").ZodOptional<import("zod").ZodString>;
-        }, import("zod/v4/core").$strip>]>, import("zod").ZodObject<{
+        }, import("zod/v4/core").$strip>], "modelType">, import("zod").ZodObject<{
             type: import("zod").ZodLiteral<"translate">;
             token: import("zod").ZodString;
             done: import("zod").ZodOptional<import("zod").ZodBoolean>;

package/dist/server/bare/plugins/nmtcpp-translation/plugin.d.ts

@@ -153,7 +153,7 @@ export declare const nmtPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -190,7 +190,7 @@ export declare const nmtPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -240,7 +240,7 @@ export declare const nmtPlugin: {
                     "onnx-tts": "onnx-tts";
                     "tts-ggml": "tts-ggml";
                     "parakeet-transcription": "parakeet-transcription";
-                    "onnx-ocr": "onnx-ocr";
+                    "ggml-ocr": "ggml-ocr";
                     "sdcpp-generation": "sdcpp-generation";
                     "ggml-vla": "ggml-vla";
                     "ggml-classification": "ggml-classification";
@@ -277,7 +277,7 @@ export declare const nmtPlugin: {
                     "onnx-tts": "onnx-tts";
                     "tts-ggml": "tts-ggml";
                     "parakeet-transcription": "parakeet-transcription";
-                    "onnx-ocr": "onnx-ocr";
+                    "ggml-ocr": "ggml-ocr";
                     "sdcpp-generation": "sdcpp-generation";
                     "ggml-vla": "ggml-vla";
                     "ggml-classification": "ggml-classification";
@@ -314,7 +314,7 @@ export declare const nmtPlugin: {
                     "onnx-tts": "onnx-tts";
                     "tts-ggml": "tts-ggml";
                     "parakeet-transcription": "parakeet-transcription";
-                    "onnx-ocr": "onnx-ocr";
+                    "ggml-ocr": "ggml-ocr";
                     "sdcpp-generation": "sdcpp-generation";
                     "ggml-vla": "ggml-vla";
                     "ggml-classification": "ggml-classification";
@@ -401,7 +401,7 @@ export declare const nmtPlugin: {
     resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<ResolveResult<Record<string, unknown>>>;
     createModel(params: CreateModelParams): PluginModelResult;
     handlers: {
-        translate: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodUnion<readonly [import("zod").ZodObject<{
+        translate: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodDiscriminatedUnion<[import("zod").ZodObject<{
             modelId: import("zod").ZodString;
             text: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodArray<import("zod").ZodString>]>;
             stream: import("zod").ZodBoolean;
@@ -424,7 +424,7 @@ export declare const nmtPlugin: {
             context: import("zod").ZodOptional<import("zod").ZodString>;
             type: import("zod").ZodLiteral<"translate">;
             requestId: import("zod").ZodOptional<import("zod").ZodString>;
-        }, import("zod/v4/core").$strip>]>, import("zod").ZodObject<{
+        }, import("zod/v4/core").$strip>], "modelType">, import("zod").ZodObject<{
             type: import("zod").ZodLiteral<"translate">;
             token: import("zod").ZodString;
             done: import("zod").ZodOptional<import("zod").ZodBoolean>;

package/dist/server/bare/plugins/parakeet-transcription/plugin.d.ts

@@ -38,6 +38,7 @@ export declare const parakeetPlugin: {
                 melSpecTime: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendDevice: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendId: import("zod").ZodOptional<import("zod").ZodNumber>;
+                gpuUnsupported: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemTotalMb: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemFreeMb: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$strip>>;
@@ -100,6 +101,7 @@ export declare const parakeetPlugin: {
                 melSpecTime: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendDevice: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendId: import("zod").ZodOptional<import("zod").ZodNumber>;
+                gpuUnsupported: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemTotalMb: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemFreeMb: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$strip>>;

package/dist/server/bare/plugins/sdcpp-generation/plugin.d.ts

@@ -94,7 +94,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -131,7 +131,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -168,7 +168,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -205,7 +205,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -242,7 +242,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -279,7 +279,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -316,7 +316,7 @@ export declare const diffusionPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -355,7 +355,7 @@ export declare const diffusionPlugin: {
                     "onnx-tts": "onnx-tts";
                     "tts-ggml": "tts-ggml";
                     "parakeet-transcription": "parakeet-transcription";
-                    "onnx-ocr": "onnx-ocr";
+                    "ggml-ocr": "ggml-ocr";
                     "sdcpp-generation": "sdcpp-generation";
                     "ggml-vla": "ggml-vla";
                     "ggml-classification": "ggml-classification";
@@ -556,6 +556,8 @@ export declare const diffusionPlugin: {
             done: import("zod").ZodOptional<import("zod").ZodBoolean>;
             stats: import("zod").ZodOptional<import("zod").ZodObject<{
                 seed: import("zod").ZodOptional<import("zod").ZodNumber>;
+                width: import("zod").ZodOptional<import("zod").ZodNumber>;
+                height: import("zod").ZodOptional<import("zod").ZodNumber>;
                 modelLoadMs: import("zod").ZodOptional<import("zod").ZodNumber>;
                 generationMs: import("zod").ZodOptional<import("zod").ZodNumber>;
                 totalGenerationMs: import("zod").ZodOptional<import("zod").ZodNumber>;
@@ -564,8 +566,6 @@ export declare const diffusionPlugin: {
                 totalGenerations: import("zod").ZodOptional<import("zod").ZodNumber>;
                 totalImages: import("zod").ZodOptional<import("zod").ZodNumber>;
                 totalPixels: import("zod").ZodOptional<import("zod").ZodNumber>;
-                width: import("zod").ZodOptional<import("zod").ZodNumber>;
-                height: import("zod").ZodOptional<import("zod").ZodNumber>;
                 totalVideos: import("zod").ZodOptional<import("zod").ZodNumber>;
                 totalVideoFrames: import("zod").ZodOptional<import("zod").ZodNumber>;
                 videoFrames: import("zod").ZodOptional<import("zod").ZodNumber>;

package/dist/server/bare/plugins/tts-ggml/plugin.d.ts

@@ -3,22 +3,28 @@ export declare const ttsPlugin: {
     modelType: "tts-ggml";
     displayName: string;
     addonPackage: "@qvac/tts-ggml";
-    loadConfigSchema: import("zod").ZodUnion<readonly [import("zod").ZodObject<{
+    loadConfigSchema: import("zod").ZodDiscriminatedUnion<[import("zod").ZodObject<{
         [x: string]: import("zod").ZodOptional<import("zod").ZodUnknown>;
     }, import("zod/v4/core").$strict>, import("zod").ZodObject<{
         [x: string]: import("zod").ZodOptional<import("zod").ZodUnknown>;
-    }, import("zod/v4/core").$strict>]>;
+    }, import("zod/v4/core").$strict>], "ttsEngine">;
     resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<ResolveResult<{
-        ttsEngine: "supertonic";
-        language: "en" | "es" | "fr" | "ko" | "pt";
+        ttsEngine: "chatterbox";
+        language: "en" | "ar" | "de" | "es" | "fi" | "fr" | "it" | "ko" | "nl" | "pl" | "pt" | "ru" | "zh" | "da" | "el" | "he" | "hi" | "ms" | "no" | "sv" | "tr" | "sw";
         voice?: string | undefined;
-        ttsSpeed?: number | undefined;
-        ttsNumInferenceSteps?: number | undefined;
         useGPU?: boolean | undefined;
+        streamChunkTokens?: number | undefined;
+        streamFirstChunkTokens?: number | undefined;
+        cfmSteps?: number | undefined;
+        threads?: number | undefined;
+        nGpuLayers?: number | undefined;
+        seed?: number | undefined;
     } | {
-        ttsEngine: "chatterbox";
-        language: "en" | "ar" | "de" | "es" | "fi" | "fr" | "it" | "ko" | "nl" | "pl" | "pt" | "da" | "el" | "ms" | "no" | "sv" | "tr" | "sw";
+        ttsEngine: "supertonic";
+        language: "id" | "en" | "ar" | "bg" | "cs" | "de" | "es" | "et" | "fi" | "fr" | "hu" | "it" | "ja" | "ko" | "lt" | "lv" | "nl" | "pl" | "pt" | "ru" | "sk" | "sl" | "uk" | "da" | "el" | "hi" | "hr" | "ro" | "sv" | "tr" | "vi";
         voice?: string | undefined;
+        ttsSpeed?: number | undefined;
+        ttsNumInferenceSteps?: number | undefined;
         useGPU?: boolean | undefined;
     }, string>>;
     createModel(params: CreateModelParams): PluginModelResult;

package/dist/server/bare/plugins/tts-ggml/plugin.js

@@ -49,6 +49,18 @@ function createChatterboxModel(modelId, config, params, artifacts) {
         engine: TTSGgml.ENGINE_CHATTERBOX,
         files: { t3Model, s3genModel },
         ...(referenceAudioPath ? { referenceAudio: referenceAudioPath } : {}),
+        ...(config.streamChunkTokens !== undefined
+            ? { streamChunkTokens: config.streamChunkTokens }
+            : {}),
+        ...(config.streamFirstChunkTokens !== undefined
+            ? { streamFirstChunkTokens: config.streamFirstChunkTokens }
+            : {}),
+        ...(config.cfmSteps !== undefined ? { cfmSteps: config.cfmSteps } : {}),
+        ...(config.threads !== undefined ? { threads: config.threads } : {}),
+        ...(config.nGpuLayers !== undefined
+            ? { nGpuLayers: config.nGpuLayers }
+            : {}),
+        ...(config.seed !== undefined ? { seed: config.seed } : {}),
         config: {
             language: config.language ?? "en",
             ...(config.useGPU !== undefined ? { useGPU: config.useGPU } : {}),

package/dist/server/bare/plugins/whispercpp-transcription/plugin.d.ts

@@ -88,7 +88,7 @@ export declare const whisperPlugin: {
                 "onnx-tts": "onnx-tts";
                 "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
+                "ggml-ocr": "ggml-ocr";
                 "sdcpp-generation": "sdcpp-generation";
                 "ggml-vla": "ggml-vla";
                 "ggml-classification": "ggml-classification";
@@ -249,6 +249,7 @@ export declare const whisperPlugin: {
                 melSpecTime: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendDevice: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendId: import("zod").ZodOptional<import("zod").ZodNumber>;
+                gpuUnsupported: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemTotalMb: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemFreeMb: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$strip>>;
@@ -311,6 +312,7 @@ export declare const whisperPlugin: {
                 melSpecTime: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendDevice: import("zod").ZodOptional<import("zod").ZodNumber>;
                 backendId: import("zod").ZodOptional<import("zod").ZodNumber>;
+                gpuUnsupported: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemTotalMb: import("zod").ZodOptional<import("zod").ZodNumber>;
                 gpuMemFreeMb: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$strip>>;

package/dist/server/bare/registry/logging-stream-registry.d.ts

@@ -11,10 +11,11 @@
  *
  */
 import type { LogLevel } from "@qvac/logging";
+type StreamHandler = (level: LogLevel, namespace: string, message: string, sourceId: string) => void;
 export declare function startLogBuffering(id: string): void;
 export declare function stopLogBufferingWithTimeout(id: string): void;
-export declare function registerLoggingStream(id: string, streamHandler: (level: LogLevel, namespace: string, message: string) => void): void;
-export declare function unregisterLoggingStream(id: string, streamHandler: (level: LogLevel, namespace: string, message: string) => void): void;
+export declare function registerLoggingStream(id: string, streamHandler: StreamHandler): void;
+export declare function unregisterLoggingStream(id: string, streamHandler: StreamHandler): void;
 export declare function unregisterAllLoggingStreams(id: string): void;
 export declare function sendLogToStreams(id: string, level: LogLevel, namespace: string, message: string): void;
 export declare function hasLoggingStreams(id: string): boolean | undefined;
@@ -27,4 +28,5 @@ export declare function getLoggingStreamStats(): {
     activeTimeouts: number;
 };
 export declare function clearAllLoggingStreams(): void;
+export {};
 //# sourceMappingURL=logging-stream-registry.d.ts.map

package/dist/server/bare/registry/logging-stream-registry.js

@@ -10,6 +10,7 @@
  * - Manages stream lifecycle (subscribe/unsubscribe)
  *
  */
+import { SDK_ALL_LOG_ID } from "../../../logging/namespaces.js";
 const loggingStreams = new Map();
 // Buffering for logs emitted during model loading (before client subscribes)
 const MAX_BUFFERED_LOGS_PER_MODEL = 100;
@@ -48,7 +49,7 @@ export function registerLoggingStream(id, streamHandler) {
     if (buffered && buffered.length > 0) {
         for (const log of buffered) {
             try {
-                streamHandler(log.level, log.namespace, log.message);
+                streamHandler(log.level, log.namespace, log.message, log.sourceId);
             }
             catch (error) {
                 console.error(`Error flushing buffered log for ID ${id}:`, error); // fallback (avoid recursion)
@@ -77,12 +78,20 @@ export function unregisterAllLoggingStreams(id) {
     clearBufferingTimeout(id);
 }
 export function sendLogToStreams(id, level, namespace, message) {
+    // The originating id is preserved as `sourceId` so the global stream keeps the
+    // real origin of each log instead of reporting the subscription id.
+    deliverToStream(id, level, namespace, message, id);
+    if (id !== SDK_ALL_LOG_ID) {
+        deliverToStream(SDK_ALL_LOG_ID, level, namespace, message, id);
+    }
+}
+function deliverToStream(id, level, namespace, message, sourceId) {
     const streams = loggingStreams.get(id);
     const isBuffering = modelsWithBuffering.has(id);
     if (streams && streams.size > 0) {
         for (const streamHandler of streams) {
             try {
-                streamHandler(level, namespace, message);
+                streamHandler(level, namespace, message, sourceId);
             }
             catch (error) {
                 console.error(`Error sending log to stream for ID ${id}:`, error); // fallback (avoid recursion)
@@ -99,7 +108,7 @@ export function sendLogToStreams(id, level, namespace, message) {
         if (validLogs.length >= MAX_BUFFERED_LOGS_PER_MODEL) {
             validLogs.shift();
         }
-        validLogs.push({ level, namespace, message, timestamp: now });
+        validLogs.push({ level, namespace, message, sourceId, timestamp: now });
         logBuffer.set(id, validLogs);
     }
 }
@@ -118,7 +127,6 @@ export function getLoggingStreamStats() {
     };
 }
 export function clearAllLoggingStreams() {
-    const count = loggingStreams.size;
     for (const timeout of bufferingTimeouts.values()) {
         clearTimeout(timeout);
     }
@@ -126,7 +134,4 @@ export function clearAllLoggingStreams() {
     logBuffer.clear();
     modelsWithBuffering.clear();
     bufferingTimeouts.clear();
-    if (count > 0) {
-        console.log(`🧹 Cleared logging streams for ${count} ID(s)`); // fallback (avoid recursion)
-    }
 }

package/dist/server/bare/registry/model-config-utils.js

@@ -15,7 +15,7 @@ export const CANONICAL_TO_ALIAS = {
     [ModelType.nmtcppTranslation]: "nmt",
     [ModelType.onnxTts]: "tts",
     [ModelType.ttsGgml]: "tts",
-    [ModelType.onnxOcr]: "ocr",
+    [ModelType.ggmlOcr]: "ocr",
     [ModelType.sdcppGeneration]: "diffusion",
     [ModelType.ggmlVla]: "vla",
     [ModelType.ggmlClassification]: "classification",
@@ -26,7 +26,7 @@ export const MODEL_CONFIG_SCHEMAS = {
     [ModelType.whispercppTranscription]: whisperConfigSchema,
     [ModelType.bciWhispercppTranscription]: bciConfigSchema,
     [ModelType.parakeetTranscription]: parakeetRuntimeConfigSchema.passthrough(),
-    [ModelType.onnxOcr]: ocrConfigSchema,
+    [ModelType.ggmlOcr]: ocrConfigSchema,
     [ModelType.sdcppGeneration]: sdcppConfigSchema,
     [ModelType.ggmlVla]: vlaConfigSchema,
     [ModelType.ggmlClassification]: classificationConfigSchema,