@qvac/sdk 0.14.1 → 0.15.0

package/dist/schemas/sdk-config.d.ts

@@ -1,4 +1,4 @@
-import { z } from "zod";
+import { z } from 'zod';
 /**
  * Device match criteria for device-specific config defaults.
  * All specified criteria must match (AND logic).
@@ -40,23 +40,24 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
         repeat_penalty: z.ZodOptional<z.ZodNumber>;
         stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
         n_discarded: z.ZodOptional<z.ZodNumber>;
+        parallel: z.ZodOptional<z.ZodNumber>;
         tools: z.ZodOptional<z.ZodBoolean>;
         toolsMode: z.ZodOptional<z.ZodEnum<{
             static: "static";
             dynamic: "dynamic";
         }>>;
-        "cache-type-k": z.ZodOptional<z.ZodString>;
-        "cache-type-v": z.ZodOptional<z.ZodString>;
-        "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+        'cache-type-k': z.ZodOptional<z.ZodString>;
+        'cache-type-v': z.ZodOptional<z.ZodString>;
+        'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
             integrated: "integrated";
             dedicated: "dedicated";
         }>]>>;
-        "split-mode": z.ZodOptional<z.ZodEnum<{
+        'split-mode': z.ZodOptional<z.ZodEnum<{
             none: "none";
             layer: "layer";
             row: "row";
         }>>;
-        "tensor-split": z.ZodOptional<z.ZodString>;
+        'tensor-split': z.ZodOptional<z.ZodString>;
         openclCacheDir: z.ZodOptional<z.ZodString>;
         reasoning_budget: z.ZodOptional<z.ZodNumber>;
         projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -101,12 +102,13 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
             batched: "batched";
             sequential: "sequential";
         }>>;
+        'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strip>>>;
     "llamacpp-embedding": z.ZodOptional<z.ZodOptional<z.ZodObject<{
         gpuLayers: z.ZodOptional<z.ZodNumber>;
         device: z.ZodOptional<z.ZodEnum<{
-            cpu: "cpu";
             gpu: "gpu";
+            cpu: "cpu";
         }>>;
         batchSize: z.ZodOptional<z.ZodNumber>;
         pooling: z.ZodOptional<z.ZodEnum<{
@@ -329,8 +331,8 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
         }>>>;
         threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
         device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-            cpu: "cpu";
             gpu: "gpu";
+            cpu: "cpu";
         }>>>;
         prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
             auto: "auto";
@@ -707,23 +709,24 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
         repeat_penalty: z.ZodOptional<z.ZodNumber>;
         stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
         n_discarded: z.ZodOptional<z.ZodNumber>;
+        parallel: z.ZodOptional<z.ZodNumber>;
         tools: z.ZodOptional<z.ZodBoolean>;
         toolsMode: z.ZodOptional<z.ZodEnum<{
             static: "static";
             dynamic: "dynamic";
         }>>;
-        "cache-type-k": z.ZodOptional<z.ZodString>;
-        "cache-type-v": z.ZodOptional<z.ZodString>;
-        "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+        'cache-type-k': z.ZodOptional<z.ZodString>;
+        'cache-type-v': z.ZodOptional<z.ZodString>;
+        'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
             integrated: "integrated";
             dedicated: "dedicated";
         }>]>>;
-        "split-mode": z.ZodOptional<z.ZodEnum<{
+        'split-mode': z.ZodOptional<z.ZodEnum<{
             none: "none";
             layer: "layer";
             row: "row";
         }>>;
-        "tensor-split": z.ZodOptional<z.ZodString>;
+        'tensor-split': z.ZodOptional<z.ZodString>;
         openclCacheDir: z.ZodOptional<z.ZodString>;
         reasoning_budget: z.ZodOptional<z.ZodNumber>;
         projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -768,12 +771,13 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
             batched: "batched";
             sequential: "sequential";
         }>>;
+        'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
     }, z.core.$strip>>>;
     embeddings: z.ZodOptional<z.ZodOptional<z.ZodObject<{
         gpuLayers: z.ZodOptional<z.ZodNumber>;
         device: z.ZodOptional<z.ZodEnum<{
-            cpu: "cpu";
             gpu: "gpu";
+            cpu: "cpu";
         }>>;
         batchSize: z.ZodOptional<z.ZodNumber>;
         pooling: z.ZodOptional<z.ZodEnum<{
@@ -996,8 +1000,8 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
         }>>>;
         threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
         device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-            cpu: "cpu";
             gpu: "gpu";
+            cpu: "cpu";
         }>>>;
         prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
             auto: "auto";
@@ -1394,23 +1398,24 @@ export declare const devicePatternSchema: z.ZodObject<{
             repeat_penalty: z.ZodOptional<z.ZodNumber>;
             stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
             n_discarded: z.ZodOptional<z.ZodNumber>;
+            parallel: z.ZodOptional<z.ZodNumber>;
             tools: z.ZodOptional<z.ZodBoolean>;
             toolsMode: z.ZodOptional<z.ZodEnum<{
                 static: "static";
                 dynamic: "dynamic";
             }>>;
-            "cache-type-k": z.ZodOptional<z.ZodString>;
-            "cache-type-v": z.ZodOptional<z.ZodString>;
-            "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+            'cache-type-k': z.ZodOptional<z.ZodString>;
+            'cache-type-v': z.ZodOptional<z.ZodString>;
+            'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
                 integrated: "integrated";
                 dedicated: "dedicated";
             }>]>>;
-            "split-mode": z.ZodOptional<z.ZodEnum<{
+            'split-mode': z.ZodOptional<z.ZodEnum<{
                 none: "none";
                 layer: "layer";
                 row: "row";
             }>>;
-            "tensor-split": z.ZodOptional<z.ZodString>;
+            'tensor-split': z.ZodOptional<z.ZodString>;
             openclCacheDir: z.ZodOptional<z.ZodString>;
             reasoning_budget: z.ZodOptional<z.ZodNumber>;
             projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -1455,12 +1460,13 @@ export declare const devicePatternSchema: z.ZodObject<{
                 batched: "batched";
                 sequential: "sequential";
             }>>;
+            'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
         }, z.core.$strip>>>;
         "llamacpp-embedding": z.ZodOptional<z.ZodOptional<z.ZodObject<{
             gpuLayers: z.ZodOptional<z.ZodNumber>;
             device: z.ZodOptional<z.ZodEnum<{
-                cpu: "cpu";
                 gpu: "gpu";
+                cpu: "cpu";
             }>>;
             batchSize: z.ZodOptional<z.ZodNumber>;
             pooling: z.ZodOptional<z.ZodEnum<{
@@ -1683,8 +1689,8 @@ export declare const devicePatternSchema: z.ZodObject<{
             }>>>;
             threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
             device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-                cpu: "cpu";
                 gpu: "gpu";
+                cpu: "cpu";
             }>>>;
             prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
                 auto: "auto";
@@ -2061,23 +2067,24 @@ export declare const devicePatternSchema: z.ZodObject<{
             repeat_penalty: z.ZodOptional<z.ZodNumber>;
             stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
             n_discarded: z.ZodOptional<z.ZodNumber>;
+            parallel: z.ZodOptional<z.ZodNumber>;
             tools: z.ZodOptional<z.ZodBoolean>;
             toolsMode: z.ZodOptional<z.ZodEnum<{
                 static: "static";
                 dynamic: "dynamic";
             }>>;
-            "cache-type-k": z.ZodOptional<z.ZodString>;
-            "cache-type-v": z.ZodOptional<z.ZodString>;
-            "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+            'cache-type-k': z.ZodOptional<z.ZodString>;
+            'cache-type-v': z.ZodOptional<z.ZodString>;
+            'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
                 integrated: "integrated";
                 dedicated: "dedicated";
             }>]>>;
-            "split-mode": z.ZodOptional<z.ZodEnum<{
+            'split-mode': z.ZodOptional<z.ZodEnum<{
                 none: "none";
                 layer: "layer";
                 row: "row";
             }>>;
-            "tensor-split": z.ZodOptional<z.ZodString>;
+            'tensor-split': z.ZodOptional<z.ZodString>;
             openclCacheDir: z.ZodOptional<z.ZodString>;
             reasoning_budget: z.ZodOptional<z.ZodNumber>;
             projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -2122,12 +2129,13 @@ export declare const devicePatternSchema: z.ZodObject<{
                 batched: "batched";
                 sequential: "sequential";
             }>>;
+            'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
         }, z.core.$strip>>>;
         embeddings: z.ZodOptional<z.ZodOptional<z.ZodObject<{
             gpuLayers: z.ZodOptional<z.ZodNumber>;
             device: z.ZodOptional<z.ZodEnum<{
-                cpu: "cpu";
                 gpu: "gpu";
+                cpu: "cpu";
             }>>;
             batchSize: z.ZodOptional<z.ZodNumber>;
             pooling: z.ZodOptional<z.ZodEnum<{
@@ -2350,8 +2358,8 @@ export declare const devicePatternSchema: z.ZodObject<{
             }>>>;
             threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
             device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-                cpu: "cpu";
                 gpu: "gpu";
+                cpu: "cpu";
             }>>>;
             prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
                 auto: "auto";
@@ -2768,23 +2776,24 @@ export declare const qvacConfigSchema: z.ZodObject<{
                 repeat_penalty: z.ZodOptional<z.ZodNumber>;
                 stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
                 n_discarded: z.ZodOptional<z.ZodNumber>;
+                parallel: z.ZodOptional<z.ZodNumber>;
                 tools: z.ZodOptional<z.ZodBoolean>;
                 toolsMode: z.ZodOptional<z.ZodEnum<{
                     static: "static";
                     dynamic: "dynamic";
                 }>>;
-                "cache-type-k": z.ZodOptional<z.ZodString>;
-                "cache-type-v": z.ZodOptional<z.ZodString>;
-                "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+                'cache-type-k': z.ZodOptional<z.ZodString>;
+                'cache-type-v': z.ZodOptional<z.ZodString>;
+                'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
                     integrated: "integrated";
                     dedicated: "dedicated";
                 }>]>>;
-                "split-mode": z.ZodOptional<z.ZodEnum<{
+                'split-mode': z.ZodOptional<z.ZodEnum<{
                     none: "none";
                     layer: "layer";
                     row: "row";
                 }>>;
-                "tensor-split": z.ZodOptional<z.ZodString>;
+                'tensor-split': z.ZodOptional<z.ZodString>;
                 openclCacheDir: z.ZodOptional<z.ZodString>;
                 reasoning_budget: z.ZodOptional<z.ZodNumber>;
                 projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -2829,12 +2838,13 @@ export declare const qvacConfigSchema: z.ZodObject<{
                     batched: "batched";
                     sequential: "sequential";
                 }>>;
+                'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
             }, z.core.$strip>>>;
             "llamacpp-embedding": z.ZodOptional<z.ZodOptional<z.ZodObject<{
                 gpuLayers: z.ZodOptional<z.ZodNumber>;
                 device: z.ZodOptional<z.ZodEnum<{
-                    cpu: "cpu";
                     gpu: "gpu";
+                    cpu: "cpu";
                 }>>;
                 batchSize: z.ZodOptional<z.ZodNumber>;
                 pooling: z.ZodOptional<z.ZodEnum<{
@@ -3057,8 +3067,8 @@ export declare const qvacConfigSchema: z.ZodObject<{
                 }>>>;
                 threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
                 device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-                    cpu: "cpu";
                     gpu: "gpu";
+                    cpu: "cpu";
                 }>>>;
                 prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
                     auto: "auto";
@@ -3435,23 +3445,24 @@ export declare const qvacConfigSchema: z.ZodObject<{
                 repeat_penalty: z.ZodOptional<z.ZodNumber>;
                 stop_sequences: z.ZodOptional<z.ZodArray<z.ZodString>>;
                 n_discarded: z.ZodOptional<z.ZodNumber>;
+                parallel: z.ZodOptional<z.ZodNumber>;
                 tools: z.ZodOptional<z.ZodBoolean>;
                 toolsMode: z.ZodOptional<z.ZodEnum<{
                     static: "static";
                     dynamic: "dynamic";
                 }>>;
-                "cache-type-k": z.ZodOptional<z.ZodString>;
-                "cache-type-v": z.ZodOptional<z.ZodString>;
-                "main-gpu": z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
+                'cache-type-k': z.ZodOptional<z.ZodString>;
+                'cache-type-v': z.ZodOptional<z.ZodString>;
+                'main-gpu': z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodEnum<{
                     integrated: "integrated";
                     dedicated: "dedicated";
                 }>]>>;
-                "split-mode": z.ZodOptional<z.ZodEnum<{
+                'split-mode': z.ZodOptional<z.ZodEnum<{
                     none: "none";
                     layer: "layer";
                     row: "row";
                 }>>;
-                "tensor-split": z.ZodOptional<z.ZodString>;
+                'tensor-split': z.ZodOptional<z.ZodString>;
                 openclCacheDir: z.ZodOptional<z.ZodString>;
                 reasoning_budget: z.ZodOptional<z.ZodNumber>;
                 projectionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -3496,12 +3507,13 @@ export declare const qvacConfigSchema: z.ZodObject<{
                     batched: "batched";
                     sequential: "sequential";
                 }>>;
+                'mmproj-use-gpu': z.ZodOptional<z.ZodBoolean>;
             }, z.core.$strip>>>;
             embeddings: z.ZodOptional<z.ZodOptional<z.ZodObject<{
                 gpuLayers: z.ZodOptional<z.ZodNumber>;
                 device: z.ZodOptional<z.ZodEnum<{
-                    cpu: "cpu";
                     gpu: "gpu";
+                    cpu: "cpu";
                 }>>;
                 batchSize: z.ZodOptional<z.ZodNumber>;
                 pooling: z.ZodOptional<z.ZodEnum<{
@@ -3724,8 +3736,8 @@ export declare const qvacConfigSchema: z.ZodObject<{
                 }>>>;
                 threads: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
                 device: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
-                    cpu: "cpu";
                     gpu: "gpu";
+                    cpu: "cpu";
                 }>>>;
                 prediction: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
                     auto: "auto";

package/dist/schemas/sdk-config.js

@@ -1,23 +1,23 @@
-import { z } from "zod";
-import { logLevelSchema } from "./logging-stream.js";
-import { ModelType } from "./model-types.js";
-import { llmConfigBaseSchema, embedConfigBaseSchema } from "./llamacpp-config.js";
-import { whisperConfigSchema, parakeetConfigSchema, } from "./transcription-config.js";
-import { ocrConfigSchema } from "./ocr.js";
-import { sdcppConfigSchema } from "./sdcpp-config.js";
-import { vlaConfigSchema } from "./vla.js";
-import { runtimeContextSchema } from "./runtime-context.js";
+import { z } from 'zod';
+import { logLevelSchema } from './logging-stream.js';
+import { ModelType } from './model-types.js';
+import { llmConfigBaseSchema, embedConfigBaseSchema } from './llamacpp-config.js';
+import { whisperConfigSchema, parakeetConfigSchema } from './transcription-config.js';
+import { ocrConfigSchema } from './ocr.js';
+import { sdcppConfigSchema } from './sdcpp-config.js';
+import { vlaConfigSchema } from './vla.js';
+import { runtimeContextSchema } from './runtime-context.js';
 // Alias keys for user convenience (maps to canonical types)
 const AliasKeys = {
-    llm: "llm",
-    whisper: "whisper",
-    embeddings: "embeddings",
-    nmt: "nmt",
-    parakeet: "parakeet",
-    tts: "tts",
-    ocr: "ocr",
-    diffusion: "diffusion",
-    vla: "vla",
+    llm: 'llm',
+    whisper: 'whisper',
+    embeddings: 'embeddings',
+    nmt: 'nmt',
+    parakeet: 'parakeet',
+    tts: 'tts',
+    ocr: 'ocr',
+    diffusion: 'diffusion',
+    vla: 'vla'
 };
 /**
  * Device match criteria for device-specific config defaults.
@@ -31,7 +31,7 @@ export const deviceMatchSchema = z.object({
     /** Device model prefix to match (e.g., "Pixel 10" matches "Pixel 10 Pro") */
     deviceModelPrefix: z.string().optional(),
     /** Device model substring to match (e.g., "Galaxy" matches "Samsung Galaxy S25") */
-    deviceModelContains: z.string().optional(),
+    deviceModelContains: z.string().optional()
 });
 /**
  * Device-specific model config defaults.
@@ -43,12 +43,8 @@ export const deviceConfigDefaultsSchema = z
     // Canonical keys
     [ModelType.llamacppCompletion]: llmConfigBaseSchema.optional(),
     [ModelType.llamacppEmbedding]: embedConfigBaseSchema.optional(),
-    [ModelType.whispercppTranscription]: whisperConfigSchema
-        .partial()
-        .optional(),
-    [ModelType.parakeetTranscription]: parakeetConfigSchema
-        .partial()
-        .optional(),
+    [ModelType.whispercppTranscription]: whisperConfigSchema.partial().optional(),
+    [ModelType.parakeetTranscription]: parakeetConfigSchema.partial().optional(),
     [ModelType.nmtcppTranslation]: z.record(z.string(), z.unknown()).optional(),
     [ModelType.ttsGgml]: z.record(z.string(), z.unknown()).optional(),
     [ModelType.ggmlOcr]: ocrConfigSchema.partial().optional(),
@@ -63,7 +59,7 @@ export const deviceConfigDefaultsSchema = z
     [AliasKeys.tts]: z.record(z.string(), z.unknown()).optional(),
     [AliasKeys.ocr]: ocrConfigSchema.partial().optional(),
     [AliasKeys.diffusion]: sdcppConfigSchema.partial().optional(),
-    [AliasKeys.vla]: vlaConfigSchema.partial().optional(),
+    [AliasKeys.vla]: vlaConfigSchema.partial().optional()
 })
     .partial();
 /**
@@ -75,9 +71,9 @@ export const devicePatternSchema = z.object({
     /** Match criteria - all specified fields must match */
     match: deviceMatchSchema,
     /** Config defaults to apply when matched */
-    defaults: deviceConfigDefaultsSchema,
+    defaults: deviceConfigDefaultsSchema
 });
-const directoryPath = z.string().transform((s) => s.replace(/\/+$/, ""));
+const directoryPath = z.string().transform((s) => s.replace(/\/+$/, ''));
 /**
  * QVAC SDK Configuration Schema
  *
@@ -167,5 +163,5 @@ export const qvacConfigSchema = z.object({
      * Bare runtime version for native addon ABI verification during bundling.
      * When omitted, verify auto-detects from node_modules (bare-runtime, then bare).
      */
-    bareRuntimeVersion: z.string().optional(),
+    bareRuntimeVersion: z.string().optional()
 });

package/dist/schemas/sdk-errors-client.d.ts

@@ -1,4 +1,4 @@
-import { type ErrorCodesMap } from "@qvac/error";
+import { type ErrorCodesMap } from '@qvac/error';
 export declare const SDK_CLIENT_ERROR_CODES: {
     readonly INVALID_RESPONSE_TYPE: 50001;
     readonly INVALID_OPERATION_IN_RESPONSE: 50002;