@qvac/sdk 0.10.2 → 0.12.0

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

@@ -11,38 +11,65 @@ const ttsLanguageSchema = z.enum(TTS_LANGUAGES);
 export const ttsChatterboxRuntimeConfigSchema = z.object({
     ttsEngine: z.literal("chatterbox"),
     language: ttsLanguageSchema,
+    voice: z.string().optional(),
+    useGPU: z.boolean().optional(),
 });
 export const ttsSupertonicRuntimeConfigSchema = z.object({
     ttsEngine: z.literal("supertonic"),
     language: ttsLanguageSchema,
+    voice: z.string().optional(),
     ttsSpeed: z.number().optional(),
     ttsNumInferenceSteps: z.number().optional(),
-    ttsSupertonicMultilingual: z.boolean().optional(),
+    useGPU: z.boolean().optional(),
 });
 export const ttsRuntimeConfigSchema = z.union([
     ttsChatterboxRuntimeConfigSchema,
     ttsSupertonicRuntimeConfigSchema,
 ]);
-export const ttsChatterboxConfigSchema = ttsChatterboxRuntimeConfigSchema.extend({
-    ttsTokenizerSrc: modelSrcInputSchema,
-    ttsSpeechEncoderSrc: modelSrcInputSchema,
-    ttsEmbedTokensSrc: modelSrcInputSchema,
-    ttsConditionalDecoderSrc: modelSrcInputSchema,
-    ttsLanguageModelSrc: modelSrcInputSchema,
-    referenceAudioSrc: modelSrcInputSchema,
-});
-export const ttsSupertonicConfigSchema = ttsSupertonicRuntimeConfigSchema.extend({
-    ttsTextEncoderSrc: modelSrcInputSchema,
-    ttsDurationPredictorSrc: modelSrcInputSchema,
-    ttsVectorEstimatorSrc: modelSrcInputSchema,
-    ttsVocoderSrc: modelSrcInputSchema,
-    ttsUnicodeIndexerSrc: modelSrcInputSchema,
-    ttsTtsConfigSrc: modelSrcInputSchema,
-    ttsVoiceStyleSrc: modelSrcInputSchema,
+export const ttsChatterboxLoadConfigSchema = ttsChatterboxRuntimeConfigSchema.extend({
+    // Optional at schema time so legacy ONNX configs (no s3genModelSrc) reach
+    // the plugin's resolveConfig and raise LegacyTtsModelDeprecatedError.
+    s3genModelSrc: modelSrcInputSchema.optional(),
+    referenceAudioSrc: modelSrcInputSchema.optional(),
 });
+export const ttsSupertonicLoadConfigSchema = ttsSupertonicRuntimeConfigSchema;
+export const ttsLoadConfigSchema = z.union([
+    ttsChatterboxLoadConfigSchema,
+    ttsSupertonicLoadConfigSchema,
+]);
+// === Legacy ONNX modelConfig fields (deprecated) ===
+//
+// Pre-@qvac/tts-ggml multi-file ONNX `modelConfig` fields are kept ONLY so
+// callers migrating from earlier SDK versions hit a structured
+// `LegacyTtsModelDeprecatedError` from the TTS plugin's `resolveConfig`,
+// rather than a generic Zod `Unrecognized key` error.
+export const LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS = [
+    // ONNX runtime flag; GGML uses modelSrc (GGUF) + `language` instead.
+    "ttsSupertonicMultilingual",
+    "ttsTokenizerSrc",
+    "ttsSpeechEncoderSrc",
+    "ttsEmbedTokensSrc",
+    "ttsConditionalDecoderSrc",
+    "ttsLanguageModelSrc",
+    "ttsTextEncoderSrc",
+    "ttsDurationPredictorSrc",
+    "ttsVectorEstimatorSrc",
+    "ttsVocoderSrc",
+    "ttsUnicodeIndexerSrc",
+    "ttsTtsConfigSrc",
+    "ttsVoiceStyleSrc",
+];
+const legacyTtsOnnxFieldsShape = LEGACY_TTS_ONNX_MODEL_CONFIG_FIELDS.reduce((acc, name) => {
+    acc[name] = z.unknown().optional();
+    return acc;
+}, {});
+// Strict load schema used by `loadModel` and the tts-ggml plugin's
+// `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.union([
-    ttsChatterboxConfigSchema,
-    ttsSupertonicConfigSchema,
+    ttsChatterboxLoadConfigSchema.extend(legacyTtsOnnxFieldsShape).strict(),
+    ttsSupertonicLoadConfigSchema.extend(legacyTtsOnnxFieldsShape).strict(),
 ]);
 export const ttsClientParamsSchema = z.object({
     modelId: z.string(),

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

@@ -77,267 +77,78 @@ export declare const whisperConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            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";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
 }, z.core.$strip>;
 export type WhisperConfig = z.infer<typeof whisperConfigSchema>;
-export declare const parakeetModelTypeEnumSchema: z.ZodEnum<{
-    tdt: "tdt";
-    ctc: "ctc";
-    sortformer: "sortformer";
-}>;
-export type ParakeetModelVariant = z.infer<typeof parakeetModelTypeEnumSchema>;
 export declare const parakeetRuntimeConfigSchema: z.ZodObject<{
-    modelType: z.ZodDefault<z.ZodEnum<{
-        tdt: "tdt";
-        ctc: "ctc";
-        sortformer: "sortformer";
-    }>>;
     maxThreads: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     sampleRate: z.ZodOptional<z.ZodNumber>;
     channels: z.ZodOptional<z.ZodNumber>;
     captionEnabled: z.ZodOptional<z.ZodBoolean>;
     timestampsEnabled: z.ZodOptional<z.ZodBoolean>;
+    seed: z.ZodOptional<z.ZodNumber>;
+    streaming: z.ZodOptional<z.ZodBoolean>;
+    streamingChunkMs: z.ZodOptional<z.ZodNumber>;
+    streamingHistoryMs: z.ZodOptional<z.ZodNumber>;
+    streamingEmitPartials: z.ZodOptional<z.ZodBoolean>;
+    streamingEnergyVad: z.ZodOptional<z.ZodBoolean>;
+    streamingLeftContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingRightLookaheadMs: z.ZodOptional<z.ZodNumber>;
+    streamingSpkCacheEnable: z.ZodOptional<z.ZodBoolean>;
+    streamingSpkCacheLen: z.ZodOptional<z.ZodNumber>;
+    streamingFifoLen: z.ZodOptional<z.ZodNumber>;
+    streamingChunkLeftContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingChunkRightContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingSpkCacheUpdatePeriod: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export declare const parakeetConfigSchema: z.ZodObject<{
-    modelType: z.ZodDefault<z.ZodEnum<{
-        tdt: "tdt";
-        ctc: "ctc";
-        sortformer: "sortformer";
-    }>>;
     maxThreads: z.ZodOptional<z.ZodNumber>;
     useGPU: z.ZodOptional<z.ZodBoolean>;
     sampleRate: z.ZodOptional<z.ZodNumber>;
     channels: z.ZodOptional<z.ZodNumber>;
     captionEnabled: z.ZodOptional<z.ZodBoolean>;
     timestampsEnabled: z.ZodOptional<z.ZodBoolean>;
-    parakeetEncoderSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetDecoderSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetVocabSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetPreprocessorSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetCtcModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetTokenizerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
-    parakeetSortformerSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        src: z.ZodString;
-        name: z.ZodOptional<z.ZodString>;
-        modelId: z.ZodOptional<z.ZodString>;
-        registryPath: z.ZodOptional<z.ZodString>;
-        registrySource: z.ZodOptional<z.ZodString>;
-        blobCoreKey: z.ZodOptional<z.ZodString>;
-        blobIndex: z.ZodOptional<z.ZodNumber>;
-        engine: z.ZodOptional<z.ZodString>;
-        expectedSize: z.ZodOptional<z.ZodNumber>;
-        sha256Checksum: z.ZodOptional<z.ZodString>;
-        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
-            embeddings: "embeddings";
-            "llamacpp-completion": "llamacpp-completion";
-            "whispercpp-transcription": "whispercpp-transcription";
-            "llamacpp-embedding": "llamacpp-embedding";
-            "nmtcpp-translation": "nmtcpp-translation";
-            "onnx-tts": "onnx-tts";
-            "parakeet-transcription": "parakeet-transcription";
-            "onnx-ocr": "onnx-ocr";
-            "sdcpp-generation": "sdcpp-generation";
-            llm: "llm";
-            whisper: "whisper";
-            nmt: "nmt";
-            parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
-        }>, z.ZodLiteral<"vad">]>>;
-    }, z.core.$strip>]>>;
+    seed: z.ZodOptional<z.ZodNumber>;
+    streaming: z.ZodOptional<z.ZodBoolean>;
+    streamingChunkMs: z.ZodOptional<z.ZodNumber>;
+    streamingHistoryMs: z.ZodOptional<z.ZodNumber>;
+    streamingEmitPartials: z.ZodOptional<z.ZodBoolean>;
+    streamingEnergyVad: z.ZodOptional<z.ZodBoolean>;
+    streamingLeftContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingRightLookaheadMs: z.ZodOptional<z.ZodNumber>;
+    streamingSpkCacheEnable: z.ZodOptional<z.ZodBoolean>;
+    streamingSpkCacheLen: z.ZodOptional<z.ZodNumber>;
+    streamingFifoLen: z.ZodOptional<z.ZodNumber>;
+    streamingChunkLeftContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingChunkRightContextMs: z.ZodOptional<z.ZodNumber>;
+    streamingSpkCacheUpdatePeriod: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export type ParakeetRuntimeConfig = z.infer<typeof parakeetRuntimeConfigSchema>;
 export type ParakeetConfig = z.infer<typeof parakeetConfigSchema>;
+export declare const LEGACY_PARAKEET_ONNX_MODEL_CONFIG_FIELDS: readonly ["parakeetEncoderSrc", "parakeetDecoderSrc", "parakeetVocabSrc", "parakeetPreprocessorSrc", "parakeetCtcModelSrc", "parakeetTokenizerSrc", "parakeetSortformerSrc", "parakeetModelSrc", "modelType"];
+export declare const parakeetLoadConfigSchema: z.ZodObject<{
+    [x: string]: z.ZodOptional<z.ZodUnknown>;
+}, z.core.$strict>;
 //# sourceMappingURL=transcription-config.d.ts.map

package/dist/schemas/transcription-config.js

@@ -68,26 +68,87 @@ export const whisperConfigSchema = z.object({
     miscConfig: miscConfigSchema,
     vadModelSrc: modelSrcInputSchema.optional(),
 });
-// === Parakeet (NVIDIA NeMo ONNX) engine config ===
-export const parakeetModelTypeEnumSchema = z.enum(["tdt", "ctc", "sortformer"]);
+// === Parakeet (NVIDIA NeMo GGML) engine config ===
+//
+// Backed by the ggml-based qvac-parakeet.cpp engine. A single GGUF
+// checkpoint covers every variant (TDT, CTC, EOU, Sortformer); the
+// addon auto-detects the model type from `parakeet.model.type` GGUF
+// metadata, so callers no longer pass a `modelType` discriminator and
+// only ever supply a single `modelSrc` at `loadModel` time.
+//
+// The `streaming*` knobs below configure the addon at load time. To
+// override any of them per `transcribeStream` call, see
+// `parakeetStreamingRunConfigSchema` in `./transcription.ts` — the
+// per-call schema intentionally drops the `streaming` prefix because
+// every field on it is already namespaced under `parakeetStreamingConfig`.
 export const parakeetRuntimeConfigSchema = z.object({
-    modelType: parakeetModelTypeEnumSchema.default("tdt"),
     maxThreads: z.number().int().optional(),
     useGPU: z.boolean().optional(),
     sampleRate: z.number().int().optional(),
     channels: z.number().int().optional(),
     captionEnabled: z.boolean().optional(),
     timestampsEnabled: z.boolean().optional(),
+    seed: z.number().int().optional(),
+    streaming: z.boolean().optional(),
+    streamingChunkMs: z.number().int().positive().optional(),
+    streamingHistoryMs: z.number().int().positive().optional(),
+    streamingEmitPartials: z.boolean().optional(),
+    /**
+     * CTC/TDT-only energy-based voice-activity hint. Forwarded to
+     * parakeet-cpp's `StreamingOptions::enable_energy_vad`. Influences
+     * how the engine segments speech (segment cadence, partial vs final
+     * emission) but does NOT add new event types to the transcribeStream
+     * output. Use the whisper engine if you need standalone VAD
+     * `speaking`/`probability` events.
+     */
+    streamingEnergyVad: z.boolean().optional(),
+    streamingLeftContextMs: z.number().int().nonnegative().optional(),
+    streamingRightLookaheadMs: z.number().int().nonnegative().optional(),
+    // === AOSC (Audio-Online Speaker Cache; v2.1+ Sortformer only) =========
+    // Auto-enabled when the loaded GGUF carries
+    // `parakeet.model_variant == "sortformer-streaming-v2.1-aosc"`. Ignored
+    // by v1/v2 Sortformer and by non-Sortformer engines.
+    streamingSpkCacheEnable: z.boolean().optional(),
+    streamingSpkCacheLen: z.number().int().positive().optional(),
+    streamingFifoLen: z.number().int().positive().optional(),
+    streamingChunkLeftContextMs: z.number().int().nonnegative().optional(),
+    streamingChunkRightContextMs: z.number().int().nonnegative().optional(),
+    streamingSpkCacheUpdatePeriod: z.number().int().positive().optional(),
 });
-export const parakeetConfigSchema = parakeetRuntimeConfigSchema.extend({
-    // TDT sources
-    parakeetEncoderSrc: modelSrcInputSchema.optional(),
-    parakeetDecoderSrc: modelSrcInputSchema.optional(),
-    parakeetVocabSrc: modelSrcInputSchema.optional(),
-    parakeetPreprocessorSrc: modelSrcInputSchema.optional(),
-    // CTC sources
-    parakeetCtcModelSrc: modelSrcInputSchema.optional(),
-    parakeetTokenizerSrc: modelSrcInputSchema.optional(),
-    // Sortformer source
-    parakeetSortformerSrc: modelSrcInputSchema.optional(),
-});
+// Parakeet's load-time config currently has no fields beyond the
+// runtime knobs (single GGUF model is supplied via the top-level
+// `modelSrc` of `loadModel`). The alias is retained so consumers can
+// keep importing `ParakeetConfig` / `parakeetConfigSchema`.
+export const parakeetConfigSchema = parakeetRuntimeConfigSchema;
+// === Parakeet legacy ONNX modelConfig fields (deprecated) ===
+//
+// As of @qvac/transcription-parakeet 0.6.0 the addon ships as a single
+// GGUF that auto-detects TDT / CTC / EOU / Sortformer from GGUF
+// metadata. The pre-0.4 multi-file ONNX `modelConfig` fields below are
+// kept ONLY so callers migrating from earlier SDK versions hit a
+// structured `LegacyParakeetModelDeprecatedError` (with a migration
+// message) raised from the parakeet plugin's `resolveConfig`, rather
+// than a generic Zod `Unrecognized key` error.
+export const LEGACY_PARAKEET_ONNX_MODEL_CONFIG_FIELDS = [
+    "parakeetEncoderSrc",
+    "parakeetDecoderSrc",
+    "parakeetVocabSrc",
+    "parakeetPreprocessorSrc",
+    "parakeetCtcModelSrc",
+    "parakeetTokenizerSrc",
+    "parakeetSortformerSrc",
+    "parakeetModelSrc",
+    "modelType",
+];
+const legacyParakeetOnnxFieldsShape = LEGACY_PARAKEET_ONNX_MODEL_CONFIG_FIELDS.reduce((acc, name) => {
+    acc[name] = z.unknown().optional();
+    return acc;
+}, {});
+// Strict schema used by `loadModel` and the parakeet plugin's
+// `loadConfigSchema`. Permits the deprecated ONNX field names so the
+// plugin's `resolveConfig` can raise a structured
+// `LegacyParakeetModelDeprecatedError` instead of a generic Zod error;
+// other unknown keys are still rejected by `.strict()`.
+export const parakeetLoadConfigSchema = parakeetRuntimeConfigSchema
+    .extend(legacyParakeetOnnxFieldsShape)
+    .strict();

package/dist/schemas/transcription.d.ts

@@ -37,6 +37,23 @@ export declare const transcribeSegmentSchema: z.ZodObject<{
     append: z.ZodBoolean;
     id: z.ZodNumber;
 }, z.core.$strip>;
+export declare const vadStateEventSchema: z.ZodObject<{
+    speaking: z.ZodBoolean;
+    probability: z.ZodNumber;
+}, z.core.$strip>;
+export declare const whisperEndOfTurnEventSchema: z.ZodObject<{
+    source: z.ZodLiteral<"whisper">;
+    silenceDurationMs: z.ZodNumber;
+}, z.core.$strip>;
+export declare const parakeetEndOfTurnEventSchema: z.ZodObject<{
+    source: z.ZodLiteral<"parakeet">;
+}, z.core.$strip>;
+export declare const endOfTurnEventSchema: z.ZodPreprocess<z.ZodDiscriminatedUnion<[z.ZodObject<{
+    source: z.ZodLiteral<"whisper">;
+    silenceDurationMs: z.ZodNumber;
+}, z.core.$strip>, z.ZodObject<{
+    source: z.ZodLiteral<"parakeet">;
+}, z.core.$strip>], "source">>;
 export declare const transcribeRequestSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodOptional<z.ZodString>;
@@ -49,6 +66,7 @@ export declare const transcribeRequestSchema: z.ZodObject<{
         value: z.ZodString;
     }, z.core.$strip>], "type">;
     type: z.ZodLiteral<"transcribe">;
+    requestId: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 export declare const transcribeResponseSchema: z.ZodObject<{
     text: z.ZodOptional<z.ZodString>;
@@ -73,6 +91,16 @@ export declare const transcribeResponseSchema: z.ZodObject<{
         append: z.ZodBoolean;
         id: z.ZodNumber;
     }, z.core.$strip>>;
+    vad: z.ZodOptional<z.ZodObject<{
+        speaking: z.ZodBoolean;
+        probability: z.ZodNumber;
+    }, z.core.$strip>>;
+    endOfTurn: z.ZodOptional<z.ZodPreprocess<z.ZodDiscriminatedUnion<[z.ZodObject<{
+        source: z.ZodLiteral<"whisper">;
+        silenceDurationMs: z.ZodNumber;
+    }, z.core.$strip>, z.ZodObject<{
+        source: z.ZodLiteral<"parakeet">;
+    }, z.core.$strip>], "source">>>;
     type: z.ZodLiteral<"transcribe">;
 }, z.core.$strip>;
 export type AudioInput = z.infer<typeof audioInputSchema>;
@@ -86,11 +114,54 @@ export type TranscribeClientParams = {
 };
 export type TranscribeRequest = z.infer<typeof transcribeRequestSchema>;
 export type TranscribeResponse = z.infer<typeof transcribeResponseSchema>;
+/**
+ * Per-call overrides for parakeet's duplex streaming session.
+ *
+ * Each field maps to its `streaming*`-prefixed counterpart in
+ * `parakeetConfig` (see `parakeetRuntimeConfigSchema`). The `streaming`
+ * prefix is intentionally dropped here because every field on this
+ * object is already namespaced under the `parakeetStreamingConfig`
+ * field of `transcribeStream({ ... })`. Any field omitted falls back
+ * to the load-time value.
+ */
+export declare const parakeetStreamingRunConfigSchema: z.ZodObject<{
+    chunkMs: z.ZodOptional<z.ZodNumber>;
+    historyMs: z.ZodOptional<z.ZodNumber>;
+    leftContextMs: z.ZodOptional<z.ZodNumber>;
+    rightLookaheadMs: z.ZodOptional<z.ZodNumber>;
+    emitPartials: z.ZodOptional<z.ZodBoolean>;
+    emitEnergyVad: z.ZodOptional<z.ZodBoolean>;
+    spkCacheEnable: z.ZodOptional<z.ZodBoolean>;
+    spkCacheLen: z.ZodOptional<z.ZodNumber>;
+    fifoLen: z.ZodOptional<z.ZodNumber>;
+    chunkLeftContextMs: z.ZodOptional<z.ZodNumber>;
+    chunkRightContextMs: z.ZodOptional<z.ZodNumber>;
+    spkCacheUpdatePeriod: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type ParakeetStreamingRunConfig = z.infer<typeof parakeetStreamingRunConfigSchema>;
 export declare const transcribeStreamRequestSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodOptional<z.ZodString>;
     metadata: z.ZodOptional<z.ZodBoolean>;
     type: z.ZodLiteral<"transcribeStream">;
+    emitVadEvents: z.ZodOptional<z.ZodBoolean>;
+    endOfTurnSilenceMs: z.ZodOptional<z.ZodNumber>;
+    vadRunIntervalMs: z.ZodOptional<z.ZodNumber>;
+    parakeetStreamingConfig: z.ZodOptional<z.ZodObject<{
+        chunkMs: z.ZodOptional<z.ZodNumber>;
+        historyMs: z.ZodOptional<z.ZodNumber>;
+        leftContextMs: z.ZodOptional<z.ZodNumber>;
+        rightLookaheadMs: z.ZodOptional<z.ZodNumber>;
+        emitPartials: z.ZodOptional<z.ZodBoolean>;
+        emitEnergyVad: z.ZodOptional<z.ZodBoolean>;
+        spkCacheEnable: z.ZodOptional<z.ZodBoolean>;
+        spkCacheLen: z.ZodOptional<z.ZodNumber>;
+        fifoLen: z.ZodOptional<z.ZodNumber>;
+        chunkLeftContextMs: z.ZodOptional<z.ZodNumber>;
+        chunkRightContextMs: z.ZodOptional<z.ZodNumber>;
+        spkCacheUpdatePeriod: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+    requestId: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 export declare const transcribeStreamResponseSchema: z.ZodObject<{
     text: z.ZodOptional<z.ZodString>;
@@ -115,6 +186,16 @@ export declare const transcribeStreamResponseSchema: z.ZodObject<{
         append: z.ZodBoolean;
         id: z.ZodNumber;
     }, z.core.$strip>>;
+    vad: z.ZodOptional<z.ZodObject<{
+        speaking: z.ZodBoolean;
+        probability: z.ZodNumber;
+    }, z.core.$strip>>;
+    endOfTurn: z.ZodOptional<z.ZodPreprocess<z.ZodDiscriminatedUnion<[z.ZodObject<{
+        source: z.ZodLiteral<"whisper">;
+        silenceDurationMs: z.ZodNumber;
+    }, z.core.$strip>, z.ZodObject<{
+        source: z.ZodLiteral<"parakeet">;
+    }, z.core.$strip>], "source">>>;
     type: z.ZodLiteral<"transcribeStream">;
 }, z.core.$strip>;
 export type TranscribeStreamRequest = z.infer<typeof transcribeStreamRequestSchema>;
@@ -123,18 +204,43 @@ export type TranscribeStreamClientParams = {
     modelId: string;
     prompt?: string;
     metadata?: boolean;
+    emitVadEvents?: boolean;
+    endOfTurnSilenceMs?: number;
+    vadRunIntervalMs?: number;
+    parakeetStreamingConfig?: ParakeetStreamingRunConfig;
 };
 export interface TranscribeStreamSession {
-    write(audioChunk: Buffer): void;
+    write(audioChunk: Uint8Array): void;
     end(): void;
     destroy(): void;
     [Symbol.asyncIterator](): AsyncIterator<string>;
 }
 export interface TranscribeStreamMetadataSession {
-    write(audioChunk: Buffer): void;
+    write(audioChunk: Uint8Array): void;
     end(): void;
     destroy(): void;
     [Symbol.asyncIterator](): AsyncIterator<TranscribeSegment>;
 }
+export type VadStateEvent = z.infer<typeof vadStateEventSchema>;
+export type WhisperEndOfTurnEvent = z.infer<typeof whisperEndOfTurnEventSchema>;
+export type ParakeetEndOfTurnEvent = z.infer<typeof parakeetEndOfTurnEventSchema>;
+export type EndOfTurnEvent = z.infer<typeof endOfTurnEventSchema>;
+export type TranscribeStreamEvent = {
+    type: "text";
+    text: string;
+} | {
+    type: "segment";
+    segment: TranscribeSegment;
+} | ({
+    type: "vad";
+} & VadStateEvent) | ({
+    type: "endOfTurn";
+} & EndOfTurnEvent);
+export interface TranscribeStreamConversationSession {
+    write(audioChunk: Uint8Array): void;
+    end(): void;
+    destroy(): void;
+    [Symbol.asyncIterator](): AsyncIterator<TranscribeStreamEvent>;
+}
 export type TranscribeStats = z.infer<typeof transcribeStatsSchema>;
 //# sourceMappingURL=transcription.d.ts.map