@qvac/sdk 0.11.0 → 0.12.0

package/dist/schemas/transcription.js

@@ -11,6 +11,11 @@ export const audioInputSchema = z.discriminatedUnion("type", [
 ]);
 const transcribeBaseSchema = z.object({
     modelId: z.string(),
+    /**
+     * Initial transcription prompt. Whisper engine only — silently
+     * ignored by the parakeet engine, which has no equivalent prompting
+     * surface in `qvac-parakeet.cpp`.
+     */
     prompt: z.string().optional(),
     metadata: z.boolean().optional(),
 });
@@ -40,9 +45,33 @@ export const vadStateEventSchema = z.object({
     speaking: z.boolean(),
     probability: z.number(),
 });
-export const endOfTurnEventSchema = z.object({
+export const whisperEndOfTurnEventSchema = z.object({
+    source: z.literal("whisper"),
     silenceDurationMs: z.number(),
 });
+export const parakeetEndOfTurnEventSchema = z.object({
+    source: z.literal("parakeet"),
+});
+const endOfTurnDiscriminatedSchema = z.discriminatedUnion("source", [
+    whisperEndOfTurnEventSchema,
+    parakeetEndOfTurnEventSchema,
+]);
+/** Legacy whisper wire frames omitted `source` and only sent `silenceDurationMs`. */
+function normalizeEndOfTurnWire(value) {
+    if (typeof value === "object" &&
+        value !== null &&
+        !("source" in value) &&
+        typeof value.silenceDurationMs ===
+            "number") {
+        return {
+            source: "whisper",
+            silenceDurationMs: value
+                .silenceDurationMs,
+        };
+    }
+    return value;
+}
+export const endOfTurnEventSchema = z.preprocess(normalizeEndOfTurnWire, endOfTurnDiscriminatedSchema);
 export const transcribeRequestSchema = transcribeParamsSchema.extend({
     type: z.literal("transcribe"),
     requestId: z
@@ -63,6 +92,30 @@ const transcriptionResultBase = z.object({
 export const transcribeResponseSchema = transcriptionResultBase.extend({
     type: z.literal("transcribe"),
 });
+/**
+ * Per-call overrides for parakeet's duplex streaming session.
+ *
+ * Each field maps to its `streaming*`-prefixed counterpart in
+ * `parakeetConfig` (see `parakeetRuntimeConfigSchema`). The `streaming`
+ * prefix is intentionally dropped here because every field on this
+ * object is already namespaced under the `parakeetStreamingConfig`
+ * field of `transcribeStream({ ... })`. Any field omitted falls back
+ * to the load-time value.
+ */
+export const parakeetStreamingRunConfigSchema = z.object({
+    chunkMs: z.number().int().positive().optional(),
+    historyMs: z.number().int().positive().optional(),
+    leftContextMs: z.number().int().nonnegative().optional(),
+    rightLookaheadMs: z.number().int().nonnegative().optional(),
+    emitPartials: z.boolean().optional(),
+    emitEnergyVad: z.boolean().optional(),
+    spkCacheEnable: z.boolean().optional(),
+    spkCacheLen: z.number().int().positive().optional(),
+    fifoLen: z.number().int().positive().optional(),
+    chunkLeftContextMs: z.number().int().nonnegative().optional(),
+    chunkRightContextMs: z.number().int().nonnegative().optional(),
+    spkCacheUpdatePeriod: z.number().int().positive().optional(),
+});
 export const transcribeStreamRequestSchema = transcribeBaseSchema.extend({
     type: z.literal("transcribeStream"),
     emitVadEvents: z.boolean().optional(),
@@ -71,8 +124,9 @@ export const transcribeStreamRequestSchema = transcribeBaseSchema.extend({
         .int()
         .nonnegative()
         .optional()
-        .describe("Silence (ms) before an endOfTurn event fires. 0 or unset disables end-of-turn detection entirely."),
+        .describe("Silence (ms) before an endOfTurn event fires. 0 or unset disables end-of-turn detection entirely. Whisper engine only."),
     vadRunIntervalMs: z.number().int().positive().optional(),
+    parakeetStreamingConfig: parakeetStreamingRunConfigSchema.optional(),
     requestId: z
         .string()
         .min(1)

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

@@ -144,18 +144,23 @@ export declare const nmtConfigBaseSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     dstVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -174,18 +179,23 @@ export declare const nmtConfigBaseSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     normalize: z.ZodOptional<z.ZodNumber>;
@@ -217,18 +227,23 @@ export declare const nmtConfigBaseSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>;
         srcVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -247,18 +262,23 @@ export declare const nmtConfigBaseSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>;
         dstVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -277,18 +297,23 @@ export declare const nmtConfigBaseSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>;
         normalize: z.ZodOptional<z.ZodNumber>;
@@ -500,18 +525,23 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     dstVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -530,18 +560,23 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     normalize: z.ZodOptional<z.ZodNumber>;
@@ -573,18 +608,23 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>;
         srcVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -603,18 +643,23 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>;
         dstVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -633,18 +678,23 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>;
         normalize: z.ZodOptional<z.ZodNumber>;
@@ -742,7 +792,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
         engine?: string | undefined;
         expectedSize?: number | undefined;
         sha256Checksum?: string | undefined;
-        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
     } | undefined;
     dstVocabSrc?: string | {
         src: string;
@@ -755,7 +805,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
         engine?: string | undefined;
         expectedSize?: number | undefined;
         sha256Checksum?: string | undefined;
-        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
     } | undefined;
     normalize?: number | undefined;
     pivotModel?: {
@@ -770,7 +820,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         };
         mode?: "full" | undefined;
         beamsize?: number | undefined;
@@ -792,7 +842,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         } | undefined;
         dstVocabSrc?: string | {
             src: string;
@@ -805,7 +855,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         } | undefined;
         normalize?: number | undefined;
     } | undefined;
@@ -846,7 +896,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
         engine?: string | undefined;
         expectedSize?: number | undefined;
         sha256Checksum?: string | undefined;
-        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
     } | undefined;
     dstVocabSrc?: string | {
         src: string;
@@ -859,7 +909,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
         engine?: string | undefined;
         expectedSize?: number | undefined;
         sha256Checksum?: string | undefined;
-        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+        addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
     } | undefined;
     normalize?: number | undefined;
     pivotModel?: {
@@ -874,7 +924,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         };
         mode?: "full" | undefined;
         beamsize?: number | undefined;
@@ -896,7 +946,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         } | undefined;
         dstVocabSrc?: string | {
             src: string;
@@ -909,7 +959,7 @@ export declare const nmtConfigSchema: z.ZodPipe<z.ZodDiscriminatedUnion<[z.ZodOb
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "vla" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "tts-ggml" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "ggml-vla" | "ggml-classification" | "llm" | "whisper" | "nmt" | "parakeet" | "classification" | "vad" | undefined;
         } | undefined;
         normalize?: number | undefined;
     } | undefined;

package/dist/schemas/vla.d.ts

@@ -0,0 +1,87 @@
+import { z } from "zod";
+export declare const vlaConfigSchema: z.ZodObject<{
+    backend: z.ZodOptional<z.ZodEnum<{
+        cpu: "cpu";
+        auto: "auto";
+    }>>;
+    verbosity: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type VlaConfig = z.input<typeof vlaConfigSchema>;
+export declare const vlaHparamsSchema: z.ZodObject<{
+    chunkSize: z.ZodNumber;
+    actionDim: z.ZodNumber;
+    maxActionDim: z.ZodNumber;
+    maxStateDim: z.ZodNumber;
+    tokenizerMaxLength: z.ZodNumber;
+    visionImageSize: z.ZodNumber;
+}, z.core.$strip>;
+export type VlaHparams = z.infer<typeof vlaHparamsSchema>;
+export declare const vlaStatsSchema: z.ZodObject<{
+    vision_ms: z.ZodOptional<z.ZodNumber>;
+    smollm2_compute_ms: z.ZodOptional<z.ZodNumber>;
+    smollm2_total_ms: z.ZodOptional<z.ZodNumber>;
+    ode_ms: z.ZodOptional<z.ZodNumber>;
+    total_ms: z.ZodOptional<z.ZodNumber>;
+    backendDevice: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type VlaStats = z.infer<typeof vlaStatsSchema>;
+export declare const vlaRunRequestSchema: z.ZodObject<{
+    type: z.ZodLiteral<"vlaRun">;
+    modelId: z.ZodString;
+    images: z.ZodArray<z.ZodString>;
+    imgWidth: z.ZodNumber;
+    imgHeight: z.ZodNumber;
+    state: z.ZodString;
+    tokens: z.ZodString;
+    mask: z.ZodString;
+    noise: z.ZodOptional<z.ZodString>;
+}, z.core.$strip>;
+export type VlaRunRequest = z.input<typeof vlaRunRequestSchema>;
+export declare const vlaRunResponseSchema: z.ZodObject<{
+    actions: z.ZodString;
+    actionDim: z.ZodNumber;
+    chunkSize: z.ZodNumber;
+    stats: z.ZodOptional<z.ZodObject<{
+        vision_ms: z.ZodOptional<z.ZodNumber>;
+        smollm2_compute_ms: z.ZodOptional<z.ZodNumber>;
+        smollm2_total_ms: z.ZodOptional<z.ZodNumber>;
+        ode_ms: z.ZodOptional<z.ZodNumber>;
+        total_ms: z.ZodOptional<z.ZodNumber>;
+        backendDevice: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
+export type VlaRunResponse = z.infer<typeof vlaRunResponseSchema>;
+export declare const vlaHparamsRequestSchema: z.ZodObject<{
+    type: z.ZodLiteral<"vlaHparams">;
+    modelId: z.ZodString;
+}, z.core.$strip>;
+export type VlaHparamsRequest = z.input<typeof vlaHparamsRequestSchema>;
+export declare const vlaHparamsResponseSchema: z.ZodObject<{
+    hparams: z.ZodObject<{
+        chunkSize: z.ZodNumber;
+        actionDim: z.ZodNumber;
+        maxActionDim: z.ZodNumber;
+        maxStateDim: z.ZodNumber;
+        tokenizerMaxLength: z.ZodNumber;
+        visionImageSize: z.ZodNumber;
+    }, z.core.$strip>;
+    backendName: z.ZodNullable<z.ZodString>;
+}, z.core.$strip>;
+export type VlaHparamsResponse = z.infer<typeof vlaHparamsResponseSchema>;
+export interface VlaClientRunParams {
+    modelId: string;
+    images: Float32Array[];
+    imgWidth: number;
+    imgHeight: number;
+    state: Float32Array;
+    tokens: Int32Array;
+    mask: Uint8Array;
+    noise?: Float32Array;
+}
+export interface VlaClientRunResult {
+    actions: Float32Array;
+    actionDim: number;
+    chunkSize: number;
+    stats?: VlaStats;
+}
+//# sourceMappingURL=vla.d.ts.map

package/dist/schemas/vla.js

@@ -0,0 +1,108 @@
+import { z } from "zod";
+const BASE64_PATTERN = /^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/;
+// ============================================
+// Load-time config
+// ============================================
+export const vlaConfigSchema = z.object({
+    backend: z
+        .enum(["auto", "cpu"])
+        .optional()
+        .describe("Backend selection passed to `VlaModel.load({ backend })`. " +
+        "`'auto'` (default) prefers an accepted GPU (Vulkan / Metal / OpenCL) and falls back to CPU. " +
+        "`'cpu'` forces CPU regardless of available accelerators."),
+    verbosity: z
+        .number()
+        .int()
+        .optional()
+        .describe("Native log verbosity forwarded to the addon (0=ERROR, 1=WARN, 2=INFO, 3=DEBUG)."),
+});
+// ============================================
+// Hparams (returned by the addon after load)
+// ============================================
+export const vlaHparamsSchema = z.object({
+    chunkSize: z.number().int().nonnegative(),
+    actionDim: z.number().int().nonnegative(),
+    maxActionDim: z.number().int().nonnegative(),
+    maxStateDim: z.number().int().nonnegative(),
+    tokenizerMaxLength: z.number().int().nonnegative(),
+    visionImageSize: z.number().int().nonnegative(),
+});
+// ============================================
+// Stats
+// ============================================
+export const vlaStatsSchema = z.object({
+    vision_ms: z.number().optional(),
+    smollm2_compute_ms: z.number().optional(),
+    smollm2_total_ms: z.number().optional(),
+    ode_ms: z.number().optional(),
+    total_ms: z.number().optional(),
+    backendDevice: z
+        .number()
+        .optional()
+        .describe("0 = CPU backend, 1 = GPU backend (Vulkan / Metal / OpenCL)."),
+});
+// ============================================
+// Run request / response (wire format)
+//
+// Typed arrays travel as base64-encoded ArrayBuffers because JSON-RPC can't
+// carry them natively. The client API helpers (`vla()` in client/api/vla.ts)
+// keep the consumer-facing `Float32Array | Int32Array | Uint8Array` shape
+// and handle the encoding internally.
+// ============================================
+export const vlaRunRequestSchema = z.object({
+    type: z.literal("vlaRun"),
+    modelId: z.string(),
+    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`."),
+    imgWidth: z.number().int().positive(),
+    imgHeight: z.number().int().positive(),
+    state: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .describe("Base64-encoded `Float32Array` of length `hparams.maxStateDim`. " +
+        "Use `vlaPadState(state, hparams.maxStateDim)` to zero-pad."),
+    tokens: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .describe("Base64-encoded `Int32Array` of length `hparams.tokenizerMaxLength`."),
+    mask: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .describe("Base64-encoded `Uint8Array` of length `hparams.tokenizerMaxLength`."),
+    noise: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .optional()
+        .describe("Optional base64-encoded `Float32Array` of length " +
+        "`hparams.chunkSize * hparams.maxActionDim`. When omitted the addon " +
+        "samples its own prior."),
+});
+export const vlaRunResponseSchema = z.object({
+    actions: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .describe("Base64-encoded `Float32Array` of length `hparams.chunkSize * hparams.actionDim`."),
+    actionDim: z.number().int().positive(),
+    chunkSize: z.number().int().positive(),
+    stats: vlaStatsSchema.optional(),
+});
+// ============================================
+// Hparams request / response (plugin handler)
+// ============================================
+export const vlaHparamsRequestSchema = z.object({
+    type: z.literal("vlaHparams"),
+    modelId: z.string(),
+});
+export const vlaHparamsResponseSchema = z.object({
+    hparams: vlaHparamsSchema,
+    backendName: z.string().nullable(),
+});

package/dist/scripts/run-unit-tests.js

@@ -1,4 +1,4 @@
-import { readdirSync } from "fs";
+import { readFileSync, readdirSync } from "fs";
 import { join, dirname } from "path";
 import { spawnSync } from "child_process";
 import { fileURLToPath } from "url";
@@ -19,8 +19,14 @@ function collectTestFiles(dir) {
 }
 const testFiles = collectTestFiles(testDir);
 let hasFailure = false;
+function usesNodeTestRunner(filePath) {
+    const source = readFileSync(filePath, "utf8");
+    return (source.includes("from 'node:test'") ||
+        source.includes('from "node:test"'));
+}
 for (const file of testFiles) {
-    const result = spawnSync("bun", ["run", file], {
+    const args = usesNodeTestRunner(file) ? ["test", file] : ["run", file];
+    const result = spawnSync("bun", args, {
         stdio: "inherit",
     });
     if (result.status !== 0) {

package/dist/server/bare/ops/transcribe.d.ts

@@ -1,4 +1,4 @@
-import { type TranscribeParams, type TranscribeSegment, type TranscribeStats, type TranscribeStreamEvent } from "../../../schemas/index";
+import { type TranscribeParams, type TranscribeSegment, type TranscribeStats, type TranscribeStreamEvent, type ParakeetStreamingRunConfig } from "../../../schemas/index";
 import { assertMetadataSupported, toTranscribeSegment, type WhisperAddonSegment } from "../utils/transcribe-metadata";
 export { assertMetadataSupported, toTranscribeSegment, type WhisperAddonSegment, };
 type TranscribeReturn = {
@@ -13,6 +13,7 @@ export interface TranscribeStreamOpts {
     emitVadEvents?: boolean;
     endOfTurnSilenceMs?: number;
     vadRunIntervalMs?: number;
+    parakeetStreamingConfig?: ParakeetStreamingRunConfig;
 }
 export declare function transcribeStream(modelId: string, audioInputStream: AsyncIterable<Buffer>, prompt: string | undefined, metadata: true, opts?: TranscribeStreamOpts, requestId?: string): AsyncGenerator<TranscribeSegment | TranscribeStreamEvent, void, void>;
 export declare function transcribeStream(modelId: string, audioInputStream: AsyncIterable<Buffer>, prompt?: string, metadata?: boolean, opts?: TranscribeStreamOpts, requestId?: string): AsyncGenerator<string | TranscribeStreamEvent, void, void>;