@qvac/sdk 0.11.0 → 0.12.0

package/dist/models/update-models/naming.js

@@ -317,21 +317,35 @@ function generateDiffusionName({ filename, modelName, quantization, params, tags
 }
 function generateParakeetName({ filename, lowerPath, quantization, }) {
     const lower = filename.toLowerCase();
-    // Detect model variant from registry path
     let variant = "";
-    if (lowerPath.includes("parakeet-tdt") || lowerPath.includes("parakeet/")) {
-        variant = "TDT";
+    if (lower.includes("sortformer") || lower.includes("diar_streaming")) {
+        variant = "SORTFORMER";
     }
-    else if (lowerPath.includes("parakeet-ctc")) {
+    else if (lower.includes("ctc") || lowerPath.includes("parakeet-ctc")) {
         variant = "CTC";
     }
-    else if (lowerPath.includes("eou") || lowerPath.includes("parakeet-rs")) {
+    else if (lower.includes("eou") || lowerPath.includes("parakeet-rs")) {
         variant = "EOU";
     }
-    else if (lowerPath.includes("sortformer")) {
-        variant = "SORTFORMER";
+    else if (lower.includes("tdt") || lowerPath.includes("parakeet-tdt")) {
+        variant = "TDT";
+    }
+    if (lower.endsWith(".gguf")) {
+        const paramsMatch = filename.match(/(?:^|[-_])(\d+(?:\.\d+)?[mb])(?=[-_.])/i);
+        const versionMatch = filename.match(/[-_]v(\d+(?:\.\d+)?)/i);
+        const speakerMatch = filename.match(/(\d+spk)/i);
+        const paramsHint = paramsMatch ? paramsMatch[1] : "";
+        const versionHint = versionMatch ? `V${versionMatch[1]}` : "";
+        const speakerHint = speakerMatch ? speakerMatch[1] : "";
+        const nameParts = [
+            variant,
+            paramsHint,
+            speakerHint,
+            versionHint,
+            quantization,
+        ].filter((p) => p && p !== "");
+        return `PARAKEET_${nameParts.map(cleanPart).join("_")}`;
     }
-    // Detect file role from filename
     let fileRole = "";
     if (lower.includes("encoder") && (lower.endsWith(".data") || lower.endsWith(".onnx.data"))) {
         fileRole = "ENCODER_DATA";

package/dist/pear/pre.js

@@ -32,18 +32,22 @@ const BUILTIN_PLUGINS = [
     "@qvac/sdk/llamacpp-embedding/plugin",
     "@qvac/sdk/whispercpp-transcription/plugin",
     "@qvac/sdk/nmtcpp-translation/plugin",
-    "@qvac/sdk/onnx-tts/plugin",
+    "@qvac/sdk/tts-ggml/plugin",
     "@qvac/sdk/onnx-ocr/plugin",
     "@qvac/sdk/sdcpp-generation/plugin",
+    "@qvac/sdk/ggml-vla/plugin",
+    "@qvac/sdk/ggml-classification/plugin",
 ];
 const BUILTIN_PLUGIN_EXPORTS = {
     "llamacpp-completion": "llmPlugin",
     "llamacpp-embedding": "embeddingsPlugin",
     "whispercpp-transcription": "whisperPlugin",
     "nmtcpp-translation": "nmtPlugin",
-    "onnx-tts": "ttsPlugin",
+    "tts-ggml": "ttsPlugin",
     "onnx-ocr": "ocrPlugin",
     "sdcpp-generation": "diffusionPlugin",
+    "ggml-vla": "vlaPlugin",
+    "ggml-classification": "classificationPlugin",
 };
 const SDK_NAME = "@qvac/sdk";
 const DEFAULT_PEAR_WORKER = "worker.js";

package/dist/schemas/cancel.d.ts

@@ -23,6 +23,7 @@ declare const cancelKindSchema: z.ZodEnum<{
     diffusion: "diffusion";
     tts: "tts";
     ocr: "ocr";
+    vla: "vla";
     finetune: "finetune";
     loadModel: "loadModel";
     downloadAsset: "downloadAsset";
@@ -74,6 +75,7 @@ declare const cancelBroadParamsSchema: z.ZodObject<{
         diffusion: "diffusion";
         tts: "tts";
         ocr: "ocr";
+        vla: "vla";
         finetune: "finetune";
         loadModel: "loadModel";
         downloadAsset: "downloadAsset";
@@ -95,6 +97,7 @@ declare const cancelParamsSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
         diffusion: "diffusion";
         tts: "tts";
         ocr: "ocr";
+        vla: "vla";
         finetune: "finetune";
         loadModel: "loadModel";
         downloadAsset: "downloadAsset";
@@ -118,6 +121,7 @@ export declare const cancelRequestSchema: z.ZodIntersection<z.ZodObject<{
         diffusion: "diffusion";
         tts: "tts";
         ocr: "ocr";
+        vla: "vla";
         finetune: "finetune";
         loadModel: "loadModel";
         downloadAsset: "downloadAsset";
@@ -154,6 +158,7 @@ export declare const cancelBroadSugarSchema: z.ZodObject<{
         diffusion: "diffusion";
         tts: "tts";
         ocr: "ocr";
+        vla: "vla";
         finetune: "finetune";
         loadModel: "loadModel";
         downloadAsset: "downloadAsset";

package/dist/schemas/cancel.js

@@ -27,6 +27,7 @@ const cancelKindSchema = z
     "diffusion",
     "tts",
     "ocr",
+    "vla",
     "finetune",
     "loadModel",
     "downloadAsset",

package/dist/schemas/classification.d.ts

@@ -0,0 +1,53 @@
+import { z } from "zod";
+export declare const classificationConfigSchema: z.ZodObject<{
+    modelPath: z.ZodOptional<z.ZodString>;
+    topK: z.ZodOptional<z.ZodNumber>;
+    nativeLogger: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export declare const classifyParamsSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    image: z.ZodString;
+    topK: z.ZodOptional<z.ZodNumber>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    channels: z.ZodOptional<z.ZodLiteral<3>>;
+}, z.core.$strip>;
+export declare const classifyRequestSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    image: z.ZodString;
+    topK: z.ZodOptional<z.ZodNumber>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    channels: z.ZodOptional<z.ZodLiteral<3>>;
+    type: z.ZodLiteral<"classify">;
+}, z.core.$strip>;
+export declare const classificationResultSchema: z.ZodObject<{
+    label: z.ZodString;
+    confidence: z.ZodNumber;
+}, z.core.$strip>;
+export declare const classifyResponseSchema: z.ZodObject<{
+    type: z.ZodLiteral<"classify">;
+    results: z.ZodArray<z.ZodObject<{
+        label: z.ZodString;
+        confidence: z.ZodNumber;
+    }, z.core.$strip>>;
+    done: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export type ClassificationConfig = z.infer<typeof classificationConfigSchema>;
+export type ClassifyParams = z.infer<typeof classifyParamsSchema>;
+export type ClassifyRequest = z.infer<typeof classifyRequestSchema>;
+export type ClassificationResult = z.infer<typeof classificationResultSchema>;
+export type ClassifyResponse = z.infer<typeof classifyResponseSchema>;
+export interface ClassifyClientParams {
+    modelId: string;
+    /** JPEG or PNG buffer. */
+    image: Uint8Array;
+    topK?: number;
+    /** Raw RGB image width (required for raw bytes). */
+    width?: number;
+    /** Raw RGB image height (required for raw bytes). */
+    height?: number;
+    /** Channel count — must be 3 for raw RGB. */
+    channels?: 3;
+}
+//# sourceMappingURL=classification.d.ts.map

package/dist/schemas/classification.js

@@ -0,0 +1,33 @@
+import { z } from "zod";
+export const classificationConfigSchema = z.object({
+    /** Absolute path to the GGUF weights file. Defaults to the bundled model inside @qvac/classification-ggml. */
+    modelPath: z.string().optional(),
+    /** Limit returned results to the top-K classes. Default: all classes. */
+    topK: z.number().int().optional(),
+    /** Forward native C++ log lines through the SDK logger. Off by default. */
+    nativeLogger: z.boolean().optional(),
+});
+export const classifyParamsSchema = z.object({
+    modelId: z.string(),
+    /** JPEG or PNG buffer encoded as base64, or raw RGB bytes. */
+    image: z.string(),
+    topK: z.number().int().optional(),
+    /** Raw RGB image width (required for raw bytes). */
+    width: z.number().int().optional(),
+    /** Raw RGB image height (required for raw bytes). */
+    height: z.number().int().optional(),
+    /** Channel count — must be 3 for raw RGB. */
+    channels: z.literal(3).optional(),
+});
+export const classifyRequestSchema = classifyParamsSchema.extend({
+    type: z.literal("classify"),
+});
+export const classificationResultSchema = z.object({
+    label: z.string(),
+    confidence: z.number(),
+});
+export const classifyResponseSchema = z.object({
+    type: z.literal("classify"),
+    results: z.array(classificationResultSchema),
+    done: z.boolean().optional(),
+});