@qvac/sdk 0.17.0 → 0.18.0

package/dist/server/bare/plugins/tts-ggml/plugin.d.ts

@@ -39,12 +39,273 @@ export declare const ttsPlugin: {
             surprise: "surprise";
         }>>;
         pitch: import("zod").ZodOptional<import("zod").ZodString>;
-        pace: import("zod").ZodOptional<import("zod").ZodString>;
+        pace: import("zod").ZodOptional<import("zod").ZodEnum<{
+            slow: "slow";
+            moderate: "moderate";
+            fast: "fast";
+        }>>;
         expressivity: import("zod").ZodOptional<import("zod").ZodString>;
         noise: import("zod").ZodOptional<import("zod").ZodString>;
         reverb: import("zod").ZodOptional<import("zod").ZodString>;
         quality: import("zod").ZodOptional<import("zod").ZodString>;
         ttsEngine: import("zod").ZodLiteral<"parler">;
+    }, import("zod/v4/core").$strict>, import("zod").ZodObject<{
+        lavasrEnhancerModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                audiogen: "audiogen";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
+                "parakeet-transcription": "parakeet-transcription";
+                "ggml-ocr": "ggml-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                "audiogen-ggml": "audiogen-ggml";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
+                llm: "llm";
+                whisper: "whisper";
+                bci: "bci";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                classification: "classification";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>>;
+        lavasrDenoiserModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                audiogen: "audiogen";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
+                "parakeet-transcription": "parakeet-transcription";
+                "ggml-ocr": "ggml-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                "audiogen-ggml": "audiogen-ggml";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
+                llm: "llm";
+                whisper: "whisper";
+                bci: "bci";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                classification: "classification";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>>;
+        ttsEngine: import("zod").ZodLiteral<"cosyvoice3">;
+        emotion: import("zod").ZodOptional<import("zod").ZodEnum<{
+            anger: "anger";
+            happy: "happy";
+            neutral: "neutral";
+            sad: "sad";
+        }>>;
+        pace: import("zod").ZodOptional<import("zod").ZodEnum<{
+            slow: "slow";
+            moderate: "moderate";
+            fast: "fast";
+        }>>;
+        instruct: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            dialect: import("zod").ZodOptional<import("zod").ZodEnum<{
+                cantonese: "cantonese";
+                northeastern: "northeastern";
+                gansu: "gansu";
+                guizhou: "guizhou";
+                henan: "henan";
+                hubei: "hubei";
+                hunan: "hunan";
+                jiangxi: "jiangxi";
+                minnan: "minnan";
+                ningxia: "ningxia";
+                shanxi: "shanxi";
+                shaanxi: "shaanxi";
+                shandong: "shandong";
+                shanghai: "shanghai";
+                sichuan: "sichuan";
+                tianjin: "tianjin";
+                yunnan: "yunnan";
+            }>>;
+            volume: import("zod").ZodOptional<import("zod").ZodEnum<{
+                loud: "loud";
+                soft: "soft";
+            }>>;
+            style: import("zod").ZodOptional<import("zod").ZodEnum<{
+                peppa: "peppa";
+                robot: "robot";
+            }>>;
+        }, import("zod/v4/core").$strict>]>>;
+        useGPU: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        outputSampleRate: import("zod").ZodOptional<import("zod").ZodNumber>;
+        streamChunkTokens: import("zod").ZodOptional<import("zod").ZodNumber>;
+        streamFirstChunkTokens: import("zod").ZodOptional<import("zod").ZodNumber>;
+        threads: import("zod").ZodOptional<import("zod").ZodNumber>;
+        nGpuLayers: import("zod").ZodOptional<import("zod").ZodNumber>;
+        seed: import("zod").ZodOptional<import("zod").ZodNumber>;
+    }, import("zod/v4/core").$strict>, import("zod").ZodObject<{
+        audio8CodecDecoderModelSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                audiogen: "audiogen";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
+                "parakeet-transcription": "parakeet-transcription";
+                "ggml-ocr": "ggml-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                "audiogen-ggml": "audiogen-ggml";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
+                llm: "llm";
+                whisper: "whisper";
+                bci: "bci";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                classification: "classification";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>;
+        audio8CodecEncoderModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                audiogen: "audiogen";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
+                "parakeet-transcription": "parakeet-transcription";
+                "ggml-ocr": "ggml-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                "audiogen-ggml": "audiogen-ggml";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
+                llm: "llm";
+                whisper: "whisper";
+                bci: "bci";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                classification: "classification";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>>;
+        referenceAudioSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
+            src: import("zod").ZodString;
+            name: import("zod").ZodOptional<import("zod").ZodString>;
+            modelId: import("zod").ZodOptional<import("zod").ZodString>;
+            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
+            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
+            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
+            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            engine: import("zod").ZodOptional<import("zod").ZodString>;
+            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
+            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                audiogen: "audiogen";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
+                "parakeet-transcription": "parakeet-transcription";
+                "ggml-ocr": "ggml-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                "audiogen-ggml": "audiogen-ggml";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
+                llm: "llm";
+                whisper: "whisper";
+                bci: "bci";
+                nmt: "nmt";
+                parakeet: "parakeet";
+                classification: "classification";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>>;
+        ttsEngine: import("zod").ZodLiteral<"audio8">;
+        referenceText: import("zod").ZodOptional<import("zod").ZodString>;
+        greedy: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        temperature: import("zod").ZodOptional<import("zod").ZodNumber>;
+        topK: import("zod").ZodOptional<import("zod").ZodNumber>;
+        topP: import("zod").ZodOptional<import("zod").ZodNumber>;
+        maxFrames: import("zod").ZodOptional<import("zod").ZodNumber>;
+        useGPU: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        outputSampleRate: import("zod").ZodOptional<import("zod").ZodNumber>;
+        threads: import("zod").ZodOptional<import("zod").ZodNumber>;
+        nGpuLayers: import("zod").ZodOptional<import("zod").ZodNumber>;
+        seed: import("zod").ZodOptional<import("zod").ZodNumber>;
     }, import("zod/v4/core").$strict>], "ttsEngine">;
     resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<ResolveResult<{
         ttsEngine: "parler";
@@ -66,11 +327,27 @@ export declare const ttsPlugin: {
         voice?: string | undefined;
         emotion?: "command" | "anger" | "narration" | "conversation" | "disgust" | "fear" | "happy" | "neutral" | "proper noun" | "news" | "sad" | "surprise" | undefined;
         pitch?: string | undefined;
-        pace?: string | undefined;
+        pace?: "slow" | "moderate" | "fast" | undefined;
         expressivity?: string | undefined;
         noise?: string | undefined;
         reverb?: string | undefined;
         quality?: string | undefined;
+    } | {
+        ttsEngine: "cosyvoice3";
+        emotion?: "anger" | "happy" | "neutral" | "sad" | undefined;
+        pace?: "slow" | "moderate" | "fast" | undefined;
+        instruct?: string | {
+            dialect?: "cantonese" | "northeastern" | "gansu" | "guizhou" | "henan" | "hubei" | "hunan" | "jiangxi" | "minnan" | "ningxia" | "shanxi" | "shaanxi" | "shandong" | "shanghai" | "sichuan" | "tianjin" | "yunnan" | undefined;
+            volume?: "loud" | "soft" | undefined;
+            style?: "peppa" | "robot" | undefined;
+        } | undefined;
+        useGPU?: boolean | undefined;
+        outputSampleRate?: number | undefined;
+        streamChunkTokens?: number | undefined;
+        streamFirstChunkTokens?: number | undefined;
+        threads?: number | undefined;
+        nGpuLayers?: number | undefined;
+        seed?: number | undefined;
     } | {
         ttsEngine: "chatterbox";
         language: "en" | "ar" | "de" | "es" | "fi" | "fr" | "it" | "ja" | "ko" | "nl" | "pl" | "pt" | "ru" | "zh" | "da" | "el" | "he" | "hi" | "ms" | "no" | "sv" | "tr" | "sw";
@@ -92,6 +369,19 @@ export declare const ttsPlugin: {
         useGPU?: boolean | undefined;
         outputSampleRate?: number | undefined;
         vulkanCacheDir?: string | undefined;
+    } | {
+        ttsEngine: "audio8";
+        referenceText?: string | undefined;
+        greedy?: boolean | undefined;
+        temperature?: number | undefined;
+        topK?: number | undefined;
+        topP?: number | undefined;
+        maxFrames?: number | undefined;
+        useGPU?: boolean | undefined;
+        outputSampleRate?: number | undefined;
+        threads?: number | undefined;
+        nGpuLayers?: number | undefined;
+        seed?: number | undefined;
     }, string>>;
     createModel(params: CreateModelParams): PluginModelResult;
     handlers: {
@@ -115,7 +405,11 @@ export declare const ttsPlugin: {
                 surprise: "surprise";
             }>>;
             pitch: import("zod").ZodOptional<import("zod").ZodString>;
-            pace: import("zod").ZodOptional<import("zod").ZodString>;
+            pace: import("zod").ZodOptional<import("zod").ZodEnum<{
+                slow: "slow";
+                moderate: "moderate";
+                fast: "fast";
+            }>>;
             expressivity: import("zod").ZodOptional<import("zod").ZodString>;
             noise: import("zod").ZodOptional<import("zod").ZodString>;
             reverb: import("zod").ZodOptional<import("zod").ZodString>;
@@ -160,7 +454,11 @@ export declare const ttsPlugin: {
                 surprise: "surprise";
             }>>;
             pitch: import("zod").ZodOptional<import("zod").ZodString>;
-            pace: import("zod").ZodOptional<import("zod").ZodString>;
+            pace: import("zod").ZodOptional<import("zod").ZodEnum<{
+                slow: "slow";
+                moderate: "moderate";
+                fast: "fast";
+            }>>;
             expressivity: import("zod").ZodOptional<import("zod").ZodString>;
             noise: import("zod").ZodOptional<import("zod").ZodString>;
             reverb: import("zod").ZodOptional<import("zod").ZodString>;

package/dist/server/bare/plugins/tts-ggml/plugin.js

@@ -66,6 +66,34 @@ async function resolveSupertonicConfig(config, ctx) {
 function resolveParlerConfig(config) {
     return { config, artifacts: {} };
 }
+// CosyVoice3's multi-file layout (flow/HiFT GGUFs, voice.gguf, vocab.json,
+// merges.txt) rides the primary model's registry companion set, so only the
+// optional LavaSR post-processing sources need resolution here.
+async function resolveCosyvoice3Config(config, ctx) {
+    const { lavasrEnhancerModelSrc, lavasrDenoiserModelSrc, ...runtime } = config;
+    const lavasrArtifacts = await resolveLavasrArtifacts(lavasrEnhancerModelSrc, lavasrDenoiserModelSrc, ctx);
+    return { config: runtime, artifacts: lavasrArtifacts };
+}
+async function resolveAudio8Config(config, ctx) {
+    const { audio8CodecDecoderModelSrc, audio8CodecEncoderModelSrc, referenceAudioSrc, ...runtime } = config;
+    if (!audio8CodecDecoderModelSrc) {
+        throw new TtsArtifactsRequiredError();
+    }
+    const resolve = ctx.resolveModelPath;
+    const [audio8CodecDecoderPath, audio8CodecEncoderPath, referenceAudioPath] = await Promise.all([
+        resolve(audio8CodecDecoderModelSrc),
+        audio8CodecEncoderModelSrc ? resolve(audio8CodecEncoderModelSrc) : Promise.resolve(undefined),
+        referenceAudioSrc ? resolve(referenceAudioSrc) : Promise.resolve(undefined)
+    ]);
+    return {
+        config: runtime,
+        artifacts: {
+            audio8CodecDecoderPath,
+            ...(audio8CodecEncoderPath ? { audio8CodecEncoderPath } : {}),
+            ...(referenceAudioPath ? { referenceAudioPath } : {})
+        }
+    };
+}
 // Build the optional LavaSR `files` entries from resolved artifacts. Supplying
 // a path is what enables the stage in @qvac/tts-ggml — there is no on/off flag.
 function lavasrFiles(artifacts) {
@@ -194,6 +222,97 @@ function createParlerModel(modelId, config, params) {
     registerAddonLogger(modelId, ModelType.ttsGgml, logger);
     return { model };
 }
+// Rebuild the structured instruct without explicit-undefined keys so it
+// satisfies the addon's exactOptionalPropertyTypes constructor typing.
+function toAddonInstruct(instruct) {
+    if (typeof instruct === 'string')
+        return instruct;
+    return {
+        ...(instruct.dialect !== undefined ? { dialect: instruct.dialect } : {}),
+        ...(instruct.volume !== undefined ? { volume: instruct.volume } : {}),
+        ...(instruct.style !== undefined ? { style: instruct.style } : {})
+    };
+}
+function createCosyvoice3Model(modelId, config, params, artifacts) {
+    const cosyvoiceLlmModel = params.modelPath;
+    if (!cosyvoiceLlmModel) {
+        throw new TtsArtifactsRequiredError();
+    }
+    const logger = createStreamLogger(modelId, ModelType.ttsGgml);
+    const model = new TTSGgml({
+        engine: TTSGgml.ENGINE_COSYVOICE3,
+        // The LLM GGUF's registry companion set co-locates the flow/HiFT GGUFs,
+        // voice.gguf, vocab.json and merges.txt next to it, so its containing
+        // directory is the addon's model dir.
+        files: {
+            cosyvoiceModelDir: dirname(cosyvoiceLlmModel),
+            cosyvoiceLlmModel,
+            ...lavasrFiles(artifacts)
+        },
+        ...(config.emotion !== undefined ? { emotion: config.emotion } : {}),
+        ...(config.pace !== undefined ? { pace: config.pace } : {}),
+        ...(config.instruct !== undefined ? { instruct: toAddonInstruct(config.instruct) } : {}),
+        ...(config.streamChunkTokens !== undefined
+            ? { streamChunkTokens: config.streamChunkTokens }
+            : {}),
+        ...(config.streamFirstChunkTokens !== undefined
+            ? { streamFirstChunkTokens: config.streamFirstChunkTokens }
+            : {}),
+        ...(config.threads !== undefined ? { threads: config.threads } : {}),
+        ...(config.nGpuLayers !== undefined ? { nGpuLayers: config.nGpuLayers } : {}),
+        ...(config.seed !== undefined ? { seed: config.seed } : {}),
+        config: {
+            ...(config.useGPU !== undefined ? { useGPU: config.useGPU } : {}),
+            ...(config.outputSampleRate !== undefined
+                ? { outputSampleRate: config.outputSampleRate }
+                : {})
+        },
+        logger,
+        opts: { stats: true },
+        exclusiveRun: true
+    });
+    registerAddonLogger(modelId, ModelType.ttsGgml, logger);
+    return { model };
+}
+function createAudio8Model(modelId, config, params, artifacts) {
+    const audio8Lm = params.modelPath;
+    const audio8CodecDecoder = artifacts['audio8CodecDecoderPath'];
+    const audio8CodecEncoder = artifacts['audio8CodecEncoderPath'];
+    const referenceAudioPath = artifacts['referenceAudioPath'];
+    if (!audio8Lm || !audio8CodecDecoder) {
+        throw new TtsArtifactsRequiredError();
+    }
+    const logger = createStreamLogger(modelId, ModelType.ttsGgml);
+    const model = new TTSGgml({
+        engine: TTSGgml.ENGINE_AUDIO8,
+        files: {
+            audio8Lm,
+            audio8CodecDecoder,
+            ...(audio8CodecEncoder ? { audio8CodecEncoder } : {})
+        },
+        ...(referenceAudioPath ? { referenceAudio: referenceAudioPath } : {}),
+        ...(config.referenceText !== undefined ? { referenceText: config.referenceText } : {}),
+        ...(config.greedy !== undefined ? { greedy: config.greedy } : {}),
+        ...(config.temperature !== undefined ? { temperature: config.temperature } : {}),
+        ...(config.topK !== undefined ? { topK: config.topK } : {}),
+        ...(config.topP !== undefined ? { topP: config.topP } : {}),
+        ...(config.maxFrames !== undefined ? { maxFrames: config.maxFrames } : {}),
+        ...(config.threads !== undefined ? { threads: config.threads } : {}),
+        ...(config.nGpuLayers !== undefined ? { nGpuLayers: config.nGpuLayers } : {}),
+        ...(config.seed !== undefined ? { seed: config.seed } : {}),
+        config: {
+            ...(config.useGPU !== undefined ? { useGPU: config.useGPU } : {}),
+            ...(config.outputSampleRate !== undefined
+                ? { outputSampleRate: config.outputSampleRate }
+                : {})
+        },
+        logger,
+        opts: { stats: true },
+        exclusiveRun: true
+    });
+    registerAddonLogger(modelId, ModelType.ttsGgml, logger);
+    return { model };
+}
 export const ttsPlugin = definePlugin({
     modelType: ModelType.ttsGgml,
     displayName: 'TTS (GGML)',
@@ -205,6 +324,12 @@ export const ttsPlugin = definePlugin({
         if (ttsEngine === 'parler') {
             return resolveParlerConfig(cfg);
         }
+        if (ttsEngine === 'cosyvoice3') {
+            return resolveCosyvoice3Config(cfg, ctx);
+        }
+        if (ttsEngine === 'audio8') {
+            return resolveAudio8Config(cfg, ctx);
+        }
         if (ttsEngine === 'supertonic') {
             return resolveSupertonicConfig(cfg, ctx);
         }
@@ -216,6 +341,12 @@ export const ttsPlugin = definePlugin({
         if (config.ttsEngine === 'parler') {
             return createParlerModel(params.modelId, config, params);
         }
+        if (config.ttsEngine === 'cosyvoice3') {
+            return createCosyvoice3Model(params.modelId, config, params, artifacts);
+        }
+        if (config.ttsEngine === 'audio8') {
+            return createAudio8Model(params.modelId, config, params, artifacts);
+        }
         if (config.ttsEngine === 'supertonic') {
             return createSupertonicModel(params.modelId, config, params, artifacts);
         }

package/dist/server/bare/plugins/whispercpp-transcription/plugin.js

@@ -112,53 +112,62 @@ export const whisperPlugin = definePlugin({
                 const iterator = metadata
                     ? transcribeStream(request.modelId, inputStream, request.prompt, true, streamOpts, request.requestId)
                     : transcribeStream(request.modelId, inputStream, request.prompt, false, streamOpts, request.requestId);
-                for await (const value of iterator) {
-                    if (typeof value === 'object' && value !== null && 'type' in value) {
-                        if (value.type === 'vad') {
-                            yield {
-                                type: 'transcribeStream',
-                                vad: {
-                                    speaking: value.speaking,
-                                    probability: value.probability
-                                }
-                            };
-                            continue;
-                        }
-                        if (value.type === 'endOfTurn') {
-                            // The shared ASR op normalizes addon engine sources to the
-                            // stable SDK-level Whisper/Parakeet variants.
-                            if (value.source === 'parakeet') {
+                try {
+                    let result = await iterator.next();
+                    while (!result.done) {
+                        const value = result.value;
+                        if (typeof value === 'object' && value !== null && 'type' in value) {
+                            if (value.type === 'vad') {
+                                yield {
+                                    type: 'transcribeStream',
+                                    vad: {
+                                        speaking: value.speaking,
+                                        probability: value.probability
+                                    }
+                                };
+                                result = await iterator.next();
                                 continue;
                             }
-                            if (typeof value.silenceDurationMs !== 'number') {
+                            if (value.type === 'endOfTurn') {
+                                // The shared ASR op normalizes addon engine sources to the
+                                // stable SDK-level Whisper/Parakeet variants.
+                                if (value.source !== 'parakeet' && typeof value.silenceDurationMs === 'number') {
+                                    yield {
+                                        type: 'transcribeStream',
+                                        endOfTurn: {
+                                            source: 'whisper',
+                                            silenceDurationMs: value.silenceDurationMs
+                                        }
+                                    };
+                                }
+                                result = await iterator.next();
                                 continue;
                             }
-                            yield {
-                                type: 'transcribeStream',
-                                endOfTurn: {
-                                    source: 'whisper',
-                                    silenceDurationMs: value.silenceDurationMs
-                                }
-                            };
+                            result = await iterator.next();
                             continue;
                         }
-                        continue;
+                        yield metadata
+                            ? {
+                                type: 'transcribeStream',
+                                segment: value
+                            }
+                            : {
+                                type: 'transcribeStream',
+                                text: value
+                            };
+                        result = await iterator.next();
                     }
-                    yield metadata
-                        ? {
-                            type: 'transcribeStream',
-                            segment: value
-                        }
-                        : {
-                            type: 'transcribeStream',
-                            text: value
-                        };
+                    const { modelExecutionMs, stats } = result.value;
+                    yield attachModelExecutionMs({
+                        type: 'transcribeStream',
+                        text: '',
+                        done: true,
+                        ...(stats && { stats })
+                    }, modelExecutionMs);
+                }
+                finally {
+                    await iterator.return?.(undefined);
                 }
-                yield {
-                    type: 'transcribeStream',
-                    text: '',
-                    done: true
-                };
             }
         })
     },

package/dist/server/bare/registry/model-registry.d.ts

@@ -5,6 +5,7 @@ interface AddonInterface {
 interface ModelRunResponse {
     iterate(): AsyncIterable<unknown>;
     await(): Promise<unknown>;
+    cancel(): Promise<void>;
 }
 export interface AnyModel {
     load(force?: boolean): Promise<void>;