@qvac/sdk 0.15.0 → 0.16.0

package/dist/schemas/sdk-errors-server.js

@@ -459,7 +459,9 @@ const serverErrorDefinitions = {
     // Plugin Errors (53,850-53,899)
     [SDK_SERVER_ERROR_CODES.PLUGIN_NOT_FOUND]: {
         name: 'PLUGIN_NOT_FOUND',
-        message: (modelType) => `Plugin not found for model type "${modelType}". If using a custom worker bundle, ensure the plugin is included in your qvac.config plugins array and rebuild with "npx qvac bundle sdk".`
+        message: (modelType) => modelType === 'onnx-ocr'
+            ? 'Plugin not found for model type "onnx-ocr": the ONNX OCR engine was removed in @qvac/sdk 0.15.0. Use modelType "ggml-ocr" with GGUF registry models instead (EasyOCR: OCR_LATIN; DocTR: OCR_DOCTR).'
+            : `Plugin not found for model type "${modelType}". If using a custom worker bundle, ensure the plugin is included in your qvac.config plugins array and rebuild with "npx qvac bundle sdk".`
     },
     [SDK_SERVER_ERROR_CODES.PLUGIN_HANDLER_NOT_FOUND]: {
         name: 'PLUGIN_HANDLER_NOT_FOUND',

package/dist/schemas/text-to-speech.d.ts

@@ -34,6 +34,7 @@ export declare const ttsChatterboxRuntimeConfigSchema: z.ZodObject<{
     streamChunkTokens: z.ZodOptional<z.ZodNumber>;
     streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
     cfmSteps: z.ZodOptional<z.ZodNumber>;
+    cfgRate: z.ZodOptional<z.ZodNumber>;
     threads: z.ZodOptional<z.ZodNumber>;
     nGpuLayers: z.ZodOptional<z.ZodNumber>;
     seed: z.ZodOptional<z.ZodNumber>;
@@ -78,6 +79,7 @@ export declare const ttsSupertonicRuntimeConfigSchema: z.ZodObject<{
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     outputSampleRate: z.ZodOptional<z.ZodNumber>;
+    vulkanCacheDir: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 export declare const ttsRuntimeConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
     ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -111,6 +113,7 @@ export declare const ttsRuntimeConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObjec
     streamChunkTokens: z.ZodOptional<z.ZodNumber>;
     streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
     cfmSteps: z.ZodOptional<z.ZodNumber>;
+    cfgRate: z.ZodOptional<z.ZodNumber>;
     threads: z.ZodOptional<z.ZodNumber>;
     nGpuLayers: z.ZodOptional<z.ZodNumber>;
     seed: z.ZodOptional<z.ZodNumber>;
@@ -154,6 +157,7 @@ export declare const ttsRuntimeConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObjec
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     outputSampleRate: z.ZodOptional<z.ZodNumber>;
+    vulkanCacheDir: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>], "ttsEngine">;
 export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
     ttsEngine: z.ZodLiteral<"chatterbox">;
@@ -187,6 +191,7 @@ export declare const ttsChatterboxLoadConfigSchema: z.ZodObject<{
     streamChunkTokens: z.ZodOptional<z.ZodNumber>;
     streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
     cfmSteps: z.ZodOptional<z.ZodNumber>;
+    cfgRate: z.ZodOptional<z.ZodNumber>;
     threads: z.ZodOptional<z.ZodNumber>;
     nGpuLayers: z.ZodOptional<z.ZodNumber>;
     seed: z.ZodOptional<z.ZodNumber>;
@@ -453,6 +458,7 @@ export declare const ttsSupertonicLoadConfigSchema: z.ZodObject<{
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     outputSampleRate: z.ZodOptional<z.ZodNumber>;
+    vulkanCacheDir: z.ZodOptional<z.ZodString>;
     lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
         name: z.ZodOptional<z.ZodString>;
@@ -560,6 +566,7 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
     streamChunkTokens: z.ZodOptional<z.ZodNumber>;
     streamFirstChunkTokens: z.ZodOptional<z.ZodNumber>;
     cfmSteps: z.ZodOptional<z.ZodNumber>;
+    cfgRate: z.ZodOptional<z.ZodNumber>;
     threads: z.ZodOptional<z.ZodNumber>;
     nGpuLayers: z.ZodOptional<z.ZodNumber>;
     seed: z.ZodOptional<z.ZodNumber>;
@@ -825,6 +832,7 @@ export declare const ttsLoadConfigSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
     ttsNumInferenceSteps: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     outputSampleRate: z.ZodOptional<z.ZodNumber>;
+    vulkanCacheDir: z.ZodOptional<z.ZodString>;
     lavasrEnhancerModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
         name: z.ZodOptional<z.ZodString>;
@@ -928,6 +936,8 @@ export declare const ttsRequestSchema: z.ZodObject<{
 export declare const ttsStatsSchema: z.ZodObject<{
     audioDuration: z.ZodOptional<z.ZodNumber>;
     totalSamples: z.ZodOptional<z.ZodNumber>;
+    enhancerBackendDevice: z.ZodOptional<z.ZodNumber>;
+    enhancerBackendId: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export declare const ttsResponseSchema: z.ZodObject<{
     type: z.ZodLiteral<"textToSpeech">;
@@ -936,6 +946,8 @@ export declare const ttsResponseSchema: z.ZodObject<{
     stats: z.ZodOptional<z.ZodObject<{
         audioDuration: z.ZodOptional<z.ZodNumber>;
         totalSamples: z.ZodOptional<z.ZodNumber>;
+        enhancerBackendDevice: z.ZodOptional<z.ZodNumber>;
+        enhancerBackendId: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     chunkIndex: z.ZodOptional<z.ZodNumber>;
     sentenceChunk: z.ZodOptional<z.ZodString>;
@@ -972,6 +984,8 @@ export declare const textToSpeechStreamResponseSchema: z.ZodObject<{
     stats: z.ZodOptional<z.ZodObject<{
         audioDuration: z.ZodOptional<z.ZodNumber>;
         totalSamples: z.ZodOptional<z.ZodNumber>;
+        enhancerBackendDevice: z.ZodOptional<z.ZodNumber>;
+        enhancerBackendId: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>>;
     chunkIndex: z.ZodOptional<z.ZodNumber>;
     sentenceChunk: z.ZodOptional<z.ZodString>;

package/dist/schemas/text-to-speech.js

@@ -105,6 +105,7 @@ export const ttsChatterboxRuntimeConfigSchema = z.object({
     streamChunkTokens: ttsNonNegativeIntegerSchema.optional(),
     streamFirstChunkTokens: ttsNonNegativeIntegerSchema.optional(),
     cfmSteps: ttsNonNegativeIntegerSchema.optional(),
+    cfgRate: z.number().nonnegative().optional(),
     threads: ttsPositiveIntegerSchema.optional(),
     nGpuLayers: ttsIntegerSchema.optional(),
     seed: ttsIntegerSchema.optional()
@@ -116,7 +117,8 @@ export const ttsSupertonicRuntimeConfigSchema = z.object({
     ttsSpeed: z.number().optional(),
     ttsNumInferenceSteps: z.number().optional(),
     useGPU: z.boolean().optional(),
-    outputSampleRate: ttsOutputSampleRateSchema.optional()
+    outputSampleRate: ttsOutputSampleRateSchema.optional(),
+    vulkanCacheDir: z.string().min(1).optional()
 });
 export const ttsRuntimeConfigSchema = z.discriminatedUnion('ttsEngine', [
     ttsChatterboxRuntimeConfigSchema,
@@ -193,10 +195,12 @@ const legacyTtsOnnxFieldsShape = LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS.reduce((acc
 // `loadConfigSchema`. Permits deprecated ONNX field names so
 // `resolveConfig` can raise LegacyTtsModelDeprecatedError instead of a
 // generic Zod error; other unknown keys are still rejected by `.strict()`.
-export const ttsConfigSchema = z.discriminatedUnion('ttsEngine', [
+export const ttsConfigSchema = z
+    .discriminatedUnion('ttsEngine', [
     ttsChatterboxLoadConfigSchema.extend(legacyTtsOnnxFieldsShape).strict(),
     ttsSupertonicLoadConfigSchema.extend(legacyTtsOnnxFieldsShape).strict()
-]).superRefine(refineChatterboxTokenizerAssets);
+])
+    .superRefine(refineChatterboxTokenizerAssets);
 export const ttsClientParamsSchema = z.object({
     modelId: z.string(),
     inputType: z.string().default('text'),
@@ -211,7 +215,9 @@ export const ttsRequestSchema = ttsClientParamsSchema.extend({
 });
 export const ttsStatsSchema = z.object({
     audioDuration: z.number().optional(),
-    totalSamples: z.number().optional()
+    totalSamples: z.number().optional(),
+    enhancerBackendDevice: z.number().optional(),
+    enhancerBackendId: z.number().optional()
 });
 export const ttsResponseSchema = z.object({
     type: z.literal('textToSpeech'),

package/dist/schemas/tools.d.ts

@@ -31,7 +31,7 @@ export declare const toolSchema: z.ZodObject<{
                 integer: "integer";
             }>;
             description: z.ZodOptional<z.ZodString>;
-            enum: z.ZodOptional<z.ZodArray<z.ZodString>>;
+            enum: z.ZodOptional<z.ZodArray<z.ZodUnion<readonly [z.ZodString, z.ZodNumber, z.ZodBoolean, z.ZodNull]>>>;
         }, z.core.$strip>>;
         required: z.ZodOptional<z.ZodArray<z.ZodString>>;
     }, z.core.$strip>;

package/dist/schemas/tools.js

@@ -14,6 +14,7 @@ export const TOOLS_MODE = {
     static: 'static',
     dynamic: 'dynamic'
 };
+const jsonSchemaEnumValueSchema = z.union([z.string(), z.number(), z.boolean(), z.null()]);
 export const toolSchema = z.object({
     type: z.literal('function'),
     name: z.string(),
@@ -23,7 +24,7 @@ export const toolSchema = z.object({
         properties: z.record(z.string(), z.object({
             type: z.enum(['string', 'number', 'integer', 'boolean', 'object', 'array']),
             description: z.string().optional(),
-            enum: z.array(z.string()).optional()
+            enum: z.array(jsonSchemaEnumValueSchema).optional()
         })),
         required: z.array(z.string()).optional()
     })

package/dist/schemas/translate.js

@@ -58,14 +58,18 @@ const translateRequestIdField = z
     .optional()
     .describe('Stable identifier for this in-flight translation, generated by the client at call time. Surfaces on the registry so callers can target it with `cancel({ requestId })`. Optional on the wire — the server falls back to a server-generated id when the field is missing.');
 export const translateRequestSchema = z.discriminatedUnion('modelType', [
-    translateParamsNmtSchema.extend({
+    translateParamsNmtSchema
+        .extend({
         type: z.literal('translate'),
         requestId: translateRequestIdField
-    }),
-    translateParamsLlmSchema.extend({
+    })
+        .meta({ title: 'TranslateNmtRequest' }),
+    translateParamsLlmSchema
+        .extend({
         type: z.literal('translate'),
         requestId: translateRequestIdField
     })
+        .meta({ title: 'TranslateLlmRequest' })
 ]);
 // Valid model types for translation (aliases and canonical)
 const validTranslationModelTypes = ['nmt', 'nmtcpp-translation', 'llm', 'llamacpp-completion'];
@@ -78,11 +82,13 @@ export const translateServerParamsSchema = translateParamsSchema
     .refine((data) => {
     if (!llmModelTypes.includes(data.modelType))
         return true;
-    // For LLM, check from/to exist
+    // For LLM, `to` is required; `from` is resolved by the worker
+    // (explicit, else auto-detected in server/bare/ops/translate.ts), so an
+    // absent `from` is valid here — it no longer has to arrive on the wire.
     const llmData = data;
-    return llmData.from && llmData.to;
+    return !!llmData.to;
 }, {
-    message: "Both 'from' and 'to' languages are required for LLM translation models"
+    message: "A 'to' language is required for LLM translation models"
 })
     .transform((data) => ({
     ...data,

package/dist/schemas/vla.d.ts

@@ -19,6 +19,11 @@ export declare const vlaHparamsSchema: z.ZodObject<{
         discrete: "discrete";
         continuous: "continuous";
     }>>;
+    imageInputMode: z.ZodOptional<z.ZodEnum<{
+        pixels: "pixels";
+        patches: "patches";
+    }>>;
+    imagePatchElems: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export type VlaHparams = z.infer<typeof vlaHparamsSchema>;
 export declare const vlaStatsSchema: z.ZodObject<{
@@ -78,6 +83,11 @@ export declare const vlaHparamsResponseSchema: z.ZodObject<{
             discrete: "discrete";
             continuous: "continuous";
         }>>;
+        imageInputMode: z.ZodOptional<z.ZodEnum<{
+            pixels: "pixels";
+            patches: "patches";
+        }>>;
+        imagePatchElems: z.ZodOptional<z.ZodNumber>;
     }, z.core.$strip>;
     backendName: z.ZodNullable<z.ZodString>;
 }, z.core.$strip>;

package/dist/schemas/vla.js

@@ -31,16 +31,33 @@ export const vlaHparamsSchema = z.object({
         .int()
         .positive()
         .optional()
-        .describe('Number of camera views the model expects (2 for SmolVLA, 3 for π₀.₅). ' +
-        'Pass exactly this many preprocessed frames in `images`. Optional for ' +
-        'back-compat — older addon builds may omit it.'),
+        .describe('Number of camera views the model expects (2 for SmolVLA and GR00T, 3 ' +
+        'for π₀.₅). Pass exactly this many preprocessed frames in `images`. ' +
+        'Optional for back-compat — older addon builds may omit it.'),
     stateInputMode: z
         .enum(['continuous', 'discrete'])
         .optional()
-        .describe("How the robot state is consumed. `'continuous'` (SmolVLA): the `state` " +
-        "Float32Array is projected by an in-model linear layer. `'discrete'` " +
-        '(π₀.₅): the state is tokenised into the language prompt and the `state` ' +
-        'buffer is ignored — pass an empty `Float32Array(0)`. Optional for back-compat.')
+        .describe("How the robot state is consumed. `'continuous'` (SmolVLA, GR00T): the " +
+        '`state` Float32Array is projected by an in-model linear layer. ' +
+        "`'discrete'` (π₀.₅): the state is tokenised into the language prompt " +
+        'and the `state` buffer is ignored — pass an empty `Float32Array(0)`. ' +
+        'Optional for back-compat.'),
+    imageInputMode: z
+        .enum(['pixels', 'patches'])
+        .optional()
+        .describe("How images are supplied. `'pixels'` (SmolVLA, π₀.₅): each `images` " +
+        'entry is a `3 · imgWidth · imgHeight` CHW plane from ' +
+        "`vlaPreprocessImage`. `'patches'` (GR00T): each entry is a " +
+        'pre-patchified buffer of length `imagePatchElems`. Optional for ' +
+        'back-compat.'),
+    imagePatchElems: z
+        .number()
+        .int()
+        .nonnegative()
+        .optional()
+        .describe("For patch-input models (`imageInputMode === 'patches'`), the exact " +
+        'per-camera patch buffer length each `images` entry must have. Absent ' +
+        'for pixel-input models.')
 });
 // ============================================
 // Stats
@@ -75,15 +92,18 @@ export const vlaRunRequestSchema = z.object({
     images: z
         .array(z.string().min(1).regex(BASE64_PATTERN))
         .min(1)
-        .describe('Base64-encoded preprocessed images. Each entry is the underlying ' +
-        'ArrayBuffer of a `Float32Array` produced by `vlaPreprocessImage(...)`. ' +
-        'Length per image must equal `3 * imgWidth * imgHeight`.'),
+        .describe('Base64-encoded preprocessed images, one per camera. Each entry is the ' +
+        'underlying ArrayBuffer of a `Float32Array`. For pixel-input models ' +
+        '(SmolVLA, π₀.₅) it comes from `vlaPreprocessImage(...)` and its length ' +
+        'must equal `3 * imgWidth * imgHeight`. For patch-input models (GR00T, ' +
+        "`hparams.imageInputMode === 'patches'`) it is a pre-patchified buffer " +
+        'of length `hparams.imagePatchElems`.'),
     imgWidth: z.number().int().positive(),
     imgHeight: z.number().int().positive(),
     state: z
         .string()
         .regex(BASE64_PATTERN)
-        .describe('Base64-encoded `Float32Array`. For continuous-state models (SmolVLA) ' +
+        .describe('Base64-encoded `Float32Array`. For continuous-state models (SmolVLA, GR00T) ' +
         'this is length `hparams.maxStateDim` — use ' +
         '`vlaPadState(state, hparams.maxStateDim)` to zero-pad. For ' +
         "discrete-state models (π₀.₅, `stateInputMode: 'discrete'`) the state " +

package/dist/scripts/contract/build-contract.d.ts

@@ -0,0 +1,38 @@
+import { z } from 'zod';
+export declare const contractDir: URL;
+export declare const callShapeByHandlerType: {
+    readonly reply: "request-reply";
+    readonly stream: "server-stream";
+    readonly duplex: "duplex";
+};
+type JsonSchema = Record<string, unknown>;
+export declare function toWireJsonSchema(schema: z.ZodType, io: 'input' | 'output', defName: string): JsonSchema;
+export declare function buildContract(): {
+    schemaDocument: {
+        $schema: string;
+        title: string;
+        description: string;
+        $defs: Record<string, JsonSchema>;
+    };
+    manifest: {
+        methods: {
+            progress?: {
+                condition: string;
+                responseSchema: string;
+            };
+            name: "cancel" | "transcribe" | "translate" | "finetune" | "loadModel" | "downloadAsset" | "rag" | "pluginInvoke" | "pluginInvokeStream" | "modelRegistryList" | "modelRegistrySearch" | "modelRegistryGetModel" | "completionStream" | "completionOrchestrate" | "batchCompletionStream" | "deleteCache" | "heartbeat" | "unloadModel" | "transcribeStream" | "bciTranscribe" | "bciTranscribeStream" | "loggingStream" | "embed" | "textToSpeech" | "textToSpeechStream" | "provide" | "stopProvide" | "getModelInfo" | "getLoadedModelInfo" | "ocrStream" | "diffusionStream" | "videoStream" | "upscaleStream" | "resume" | "suspend" | "state" | "classify";
+            callShape: "duplex" | "request-reply" | "server-stream";
+            requestSchema: string;
+            responseSchema: string;
+        }[];
+    };
+};
+export declare function renderContractFiles(): Promise<{
+    'schema.json': string;
+    'manifest.json': string;
+    'models.json': string;
+    'model-type-maps.json': string;
+    'error-codes.json': string;
+}>;
+export {};
+//# sourceMappingURL=build-contract.d.ts.map