@qvac/sdk 0.8.0 → 1.0.0

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

@@ -1,373 +0,0 @@
-import { type CreateModelParams, type PluginModelResult, type ResolveContext, type TtsChatterboxRuntimeConfig, type TtsSupertonicRuntimeConfig } from "../../../../schemas/index";
-export declare const ttsPlugin: {
-    modelType: "onnx-tts";
-    displayName: string;
-    addonPackage: "@qvac/tts-onnx";
-    loadConfigSchema: import("zod").ZodUnion<readonly [import("zod").ZodObject<{
-        ttsEngine: import("zod").ZodLiteral<"chatterbox">;
-        language: import("zod").ZodEnum<{
-            en: "en";
-            de: "de";
-            es: "es";
-            it: "it";
-        }>;
-        ttsTokenizerSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsSpeechEncoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
-            src: import("zod").ZodString;
-            name: import("zod").ZodOptional<import("zod").ZodString>;
-            modelId: import("zod").ZodOptional<import("zod").ZodString>;
-            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
-            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
-            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
-            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
-            engine: import("zod").ZodOptional<import("zod").ZodString>;
-            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
-            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
-            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
-                "llamacpp-completion": "llamacpp-completion";
-                "whispercpp-transcription": "whispercpp-transcription";
-                "llamacpp-embedding": "llamacpp-embedding";
-                "nmtcpp-translation": "nmtcpp-translation";
-                "onnx-tts": "onnx-tts";
-                "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsEmbedTokensSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsConditionalDecoderSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
-            src: import("zod").ZodString;
-            name: import("zod").ZodOptional<import("zod").ZodString>;
-            modelId: import("zod").ZodOptional<import("zod").ZodString>;
-            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
-            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
-            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
-            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
-            engine: import("zod").ZodOptional<import("zod").ZodString>;
-            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
-            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
-            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
-                "llamacpp-completion": "llamacpp-completion";
-                "whispercpp-transcription": "whispercpp-transcription";
-                "llamacpp-embedding": "llamacpp-embedding";
-                "nmtcpp-translation": "nmtcpp-translation";
-                "onnx-tts": "onnx-tts";
-                "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsLanguageModelSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        referenceAudioSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
-            src: import("zod").ZodString;
-            name: import("zod").ZodOptional<import("zod").ZodString>;
-            modelId: import("zod").ZodOptional<import("zod").ZodString>;
-            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
-            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
-            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
-            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
-            engine: import("zod").ZodOptional<import("zod").ZodString>;
-            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
-            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
-            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
-                "llamacpp-completion": "llamacpp-completion";
-                "whispercpp-transcription": "whispercpp-transcription";
-                "llamacpp-embedding": "llamacpp-embedding";
-                "nmtcpp-translation": "nmtcpp-translation";
-                "onnx-tts": "onnx-tts";
-                "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-    }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
-        ttsEngine: import("zod").ZodLiteral<"supertonic">;
-        language: import("zod").ZodEnum<{
-            en: "en";
-            de: "de";
-            es: "es";
-            it: "it";
-        }>;
-        ttsSpeed: import("zod").ZodOptional<import("zod").ZodNumber>;
-        ttsNumInferenceSteps: import("zod").ZodOptional<import("zod").ZodNumber>;
-        ttsTokenizerSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsTextEncoderSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsLatentDenoiserSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
-            src: import("zod").ZodString;
-            name: import("zod").ZodOptional<import("zod").ZodString>;
-            modelId: import("zod").ZodOptional<import("zod").ZodString>;
-            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
-            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
-            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
-            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
-            engine: import("zod").ZodOptional<import("zod").ZodString>;
-            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
-            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
-            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
-                "llamacpp-completion": "llamacpp-completion";
-                "whispercpp-transcription": "whispercpp-transcription";
-                "llamacpp-embedding": "llamacpp-embedding";
-                "nmtcpp-translation": "nmtcpp-translation";
-                "onnx-tts": "onnx-tts";
-                "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsVoiceDecoderSrc: 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<{
-                "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";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-        ttsVoiceSrc: import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
-            src: import("zod").ZodString;
-            name: import("zod").ZodOptional<import("zod").ZodString>;
-            modelId: import("zod").ZodOptional<import("zod").ZodString>;
-            registryPath: import("zod").ZodOptional<import("zod").ZodString>;
-            registrySource: import("zod").ZodOptional<import("zod").ZodString>;
-            blobCoreKey: import("zod").ZodOptional<import("zod").ZodString>;
-            blobIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
-            engine: import("zod").ZodOptional<import("zod").ZodString>;
-            expectedSize: import("zod").ZodOptional<import("zod").ZodNumber>;
-            sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
-            addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
-                "llamacpp-completion": "llamacpp-completion";
-                "whispercpp-transcription": "whispercpp-transcription";
-                "llamacpp-embedding": "llamacpp-embedding";
-                "nmtcpp-translation": "nmtcpp-translation";
-                "onnx-tts": "onnx-tts";
-                "parakeet-transcription": "parakeet-transcription";
-                "onnx-ocr": "onnx-ocr";
-                llm: "llm";
-                whisper: "whisper";
-                embeddings: "embeddings";
-                nmt: "nmt";
-                parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-            }>, import("zod").ZodLiteral<"vad">]>>;
-        }, import("zod/v4/core").$strip>]>;
-    }, import("zod/v4/core").$strip>]>;
-    skipPrimaryModelPathValidation: true;
-    resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<{
-        config: TtsChatterboxRuntimeConfig;
-        artifacts: {
-            tokenizerPath: string;
-            speechEncoderPath: string;
-            embedTokensPath: string;
-            conditionalDecoderPath: string;
-            languageModelPath: string;
-            referenceAudioPath: string;
-        };
-    } | {
-        config: TtsSupertonicRuntimeConfig;
-        artifacts: {
-            tokenizerPath: string;
-            textEncoderPath: string;
-            latentDenoiserPath: string;
-            voiceDecoderPath: string;
-            voicePath: string;
-        };
-    }>;
-    createModel(params: CreateModelParams): PluginModelResult;
-    handlers: {
-        textToSpeech: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodObject<{
-            modelId: import("zod").ZodString;
-            inputType: import("zod").ZodDefault<import("zod").ZodString>;
-            text: import("zod").ZodString;
-            stream: import("zod").ZodDefault<import("zod").ZodBoolean>;
-            type: import("zod").ZodLiteral<"textToSpeech">;
-        }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
-            type: import("zod").ZodLiteral<"textToSpeech">;
-            buffer: import("zod").ZodArray<import("zod").ZodNumber>;
-            done: import("zod").ZodDefault<import("zod").ZodBoolean>;
-        }, import("zod/v4/core").$strip>>;
-    };
-    logging: {
-        module: import("@qvac/tts-onnx/addonLogging").AddonLogging;
-        namespace: "onnx-tts";
-    };
-};
-//# sourceMappingURL=plugin.d.ts.map

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

@@ -1,182 +0,0 @@
-import path from "bare-path";
-import ttsAddonLogging from "@qvac/tts-onnx/addonLogging.js";
-import ONNXTTS from "@qvac/tts-onnx";
-import { definePlugin, defineHandler, ttsRequestSchema, ttsResponseSchema, ModelType, ttsConfigSchema, ADDON_TTS, } from "../../../../schemas/index.js";
-import { createStreamLogger, registerAddonLogger } from "../../../../logging/index.js";
-import { TtsArtifactsRequiredError, TtsReferenceAudioRequiredError, } from "../../../../utils/errors-server.js";
-import { textToSpeech } from "./ops/text-to-speech.js";
-import { loadReferenceAudioAt24k } from "./wav-helper.js";
-async function resolveChatterboxConfig(config, ctx) {
-    const { ttsTokenizerSrc, ttsSpeechEncoderSrc, ttsEmbedTokensSrc, ttsConditionalDecoderSrc, ttsLanguageModelSrc, referenceAudioSrc, language, } = config;
-    if (!ttsTokenizerSrc ||
-        !ttsSpeechEncoderSrc ||
-        !ttsEmbedTokensSrc ||
-        !ttsConditionalDecoderSrc ||
-        !ttsLanguageModelSrc) {
-        throw new TtsArtifactsRequiredError();
-    }
-    if (!referenceAudioSrc) {
-        throw new TtsReferenceAudioRequiredError();
-    }
-    const resolve = ctx.resolveModelPath;
-    const [tokenizerPath, speechEncoderPath, embedTokensPath, conditionalDecoderPath, languageModelPath, referenceAudioPath,] = await Promise.all([
-        resolve(ttsTokenizerSrc),
-        resolve(ttsSpeechEncoderSrc),
-        resolve(ttsEmbedTokensSrc),
-        resolve(ttsConditionalDecoderSrc),
-        resolve(ttsLanguageModelSrc),
-        resolve(referenceAudioSrc),
-    ]);
-    return {
-        config: {
-            ttsEngine: "chatterbox",
-            language,
-        },
-        artifacts: {
-            tokenizerPath,
-            speechEncoderPath,
-            embedTokensPath,
-            conditionalDecoderPath,
-            languageModelPath,
-            referenceAudioPath,
-        },
-    };
-}
-async function resolveSupertonicConfig(config, ctx) {
-    const { ttsTokenizerSrc, ttsTextEncoderSrc, ttsLatentDenoiserSrc, ttsVoiceDecoderSrc, ttsVoiceSrc, ttsSpeed, ttsNumInferenceSteps, language, } = config;
-    if (!ttsTokenizerSrc ||
-        !ttsTextEncoderSrc ||
-        !ttsLatentDenoiserSrc ||
-        !ttsVoiceDecoderSrc ||
-        !ttsVoiceSrc) {
-        throw new TtsArtifactsRequiredError();
-    }
-    const resolve = ctx.resolveModelPath;
-    const [tokenizerPath, textEncoderPath, latentDenoiserPath, voiceDecoderPath, voicePath,] = await Promise.all([
-        resolve(ttsTokenizerSrc),
-        resolve(ttsTextEncoderSrc),
-        resolve(ttsLatentDenoiserSrc),
-        resolve(ttsVoiceDecoderSrc),
-        resolve(ttsVoiceSrc),
-    ]);
-    return {
-        config: {
-            ttsEngine: "supertonic",
-            language,
-            ttsSpeed,
-            ttsNumInferenceSteps,
-        },
-        artifacts: {
-            tokenizerPath,
-            textEncoderPath,
-            latentDenoiserPath,
-            voiceDecoderPath,
-            voicePath,
-        },
-    };
-}
-function createChatterboxModel(modelId, config, artifacts) {
-    const tokenizerPath = artifacts["tokenizerPath"];
-    const speechEncoderPath = artifacts["speechEncoderPath"];
-    const embedTokensPath = artifacts["embedTokensPath"];
-    const conditionalDecoderPath = artifacts["conditionalDecoderPath"];
-    const languageModelPath = artifacts["languageModelPath"];
-    const referenceAudioPath = artifacts["referenceAudioPath"];
-    if (!tokenizerPath ||
-        !speechEncoderPath ||
-        !embedTokensPath ||
-        !conditionalDecoderPath ||
-        !languageModelPath) {
-        throw new TtsArtifactsRequiredError();
-    }
-    if (!referenceAudioPath) {
-        throw new TtsReferenceAudioRequiredError();
-    }
-    const logger = createStreamLogger(modelId, ModelType.onnxTts);
-    registerAddonLogger(modelId, ModelType.onnxTts, logger);
-    const referenceAudio = loadReferenceAudioAt24k(referenceAudioPath);
-    const args = {
-        tokenizerPath,
-        speechEncoderPath,
-        embedTokensPath,
-        conditionalDecoderPath,
-        languageModelPath,
-        referenceAudio,
-        logger,
-        opts: { stats: true },
-    };
-    const modelConfig = { language: config.language ?? "en", useGPU: false };
-    const model = new ONNXTTS(args, modelConfig);
-    return { model, loader: undefined };
-}
-function createSupertonicModel(modelId, config, artifacts) {
-    const tokenizerPath = artifacts["tokenizerPath"];
-    const textEncoderPath = artifacts["textEncoderPath"];
-    const latentDenoiserPath = artifacts["latentDenoiserPath"];
-    const voiceDecoderPath = artifacts["voiceDecoderPath"];
-    const voicePath = artifacts["voicePath"];
-    if (!tokenizerPath ||
-        !textEncoderPath ||
-        !latentDenoiserPath ||
-        !voiceDecoderPath ||
-        !voicePath) {
-        throw new TtsArtifactsRequiredError();
-    }
-    const logger = createStreamLogger(modelId, ModelType.onnxTts);
-    registerAddonLogger(modelId, ModelType.onnxTts, logger);
-    const voicesDir = path.dirname(voicePath);
-    const voiceName = path.basename(voicePath).replace(/\.bin$/i, "") || "voice";
-    const args = {
-        tokenizerPath,
-        textEncoderPath,
-        latentDenoiserPath,
-        voiceDecoderPath,
-        voicesDir,
-        voiceName,
-        speed: config.ttsSpeed ?? 1,
-        numInferenceSteps: config.ttsNumInferenceSteps ?? 5,
-        logger,
-        opts: { stats: true },
-    };
-    const modelConfig = { language: config.language ?? "en" };
-    const model = new ONNXTTS(args, modelConfig);
-    return { model, loader: undefined };
-}
-export const ttsPlugin = definePlugin({
-    modelType: ModelType.onnxTts,
-    displayName: "TTS (ONNX)",
-    addonPackage: ADDON_TTS,
-    loadConfigSchema: ttsConfigSchema,
-    skipPrimaryModelPathValidation: true,
-    async resolveConfig(cfg, ctx) {
-        const { ttsEngine } = cfg;
-        if (ttsEngine === "supertonic") {
-            return resolveSupertonicConfig(cfg, ctx);
-        }
-        return resolveChatterboxConfig(cfg, ctx);
-    },
-    createModel(params) {
-        const config = (params.modelConfig ?? {});
-        const artifacts = params.artifacts ?? {};
-        if (config.ttsEngine === "supertonic") {
-            return createSupertonicModel(params.modelId, config, artifacts);
-        }
-        return createChatterboxModel(params.modelId, config, artifacts);
-    },
-    handlers: {
-        textToSpeech: defineHandler({
-            requestSchema: ttsRequestSchema,
-            responseSchema: ttsResponseSchema,
-            streaming: true,
-            handler: async function* (request) {
-                for await (const response of textToSpeech(request)) {
-                    yield response;
-                }
-            },
-        }),
-    },
-    logging: {
-        module: ttsAddonLogging,
-        namespace: ModelType.onnxTts,
-    },
-});

package/dist/server/bare/plugins/onnx-tts/wav-helper.d.ts

@@ -1,14 +0,0 @@
-export type ReadWavResult = {
-    samples: Float32Array;
-    sampleRate: number;
-};
-/**
- * Read a WAV file and return Float32Array of mono samples in [-1, 1].
- * Supports 16-bit PCM and 32-bit float; stereo is converted to mono (left channel).
- */
-export declare function readWavAsFloat32(wavPath: string): ReadWavResult;
-/**
- * Load reference audio from a WAV path and return Float32Array at 24kHz for Chatterbox.
- */
-export declare function loadReferenceAudioAt24k(wavPath: string): Float32Array;
-//# sourceMappingURL=wav-helper.d.ts.map