@qvac/sdk 0.11.0 → 0.12.1

package/dist/server/bare/plugins/sdcpp-generation/plugin.d.ts

@@ -1,5 +1,13 @@
 import { type CreateModelParams, type PluginModelResult, type ResolveContext, type ResolveResult, type SdcppConfig } from "../../../../schemas/index";
-type DiffusionArtifactKey = "clipLModelPath" | "clipGModelPath" | "t5XxlModelPath" | "llmModelPath" | "vaeModelPath" | "esrganModelPath";
+type DiffusionArtifactKey = "clipLModelPath" | "clipGModelPath" | "t5XxlModelPath" | "llmModelPath" | "vaeModelPath" | "highNoiseDiffusionModelPath" | "esrganModelPath";
+/**
+ * Stable-diffusion.cpp plugin for image diffusion, upscaling, and Wan video.
+ *
+ * Video mode is supported on React Native, but the SDK-published Wan model
+ * set is too large to load on typical mobile devices. Mobile apps should
+ * pass a `delegate` to `loadModel(...)` to run video generation on a
+ * desktop peer instead of loading the model on-device.
+ */
 export declare const diffusionPlugin: {
     modelType: "sdcpp-generation";
     displayName: string;
@@ -8,6 +16,7 @@ export declare const diffusionPlugin: {
         mode: import("zod").ZodDefault<import("zod").ZodEnum<{
             diffusion: "diffusion";
             upscale: "upscale";
+            video: "video";
         }>>;
         threads: import("zod").ZodOptional<import("zod").ZodNumber>;
         device: import("zod").ZodOptional<import("zod").ZodEnum<{
@@ -20,7 +29,6 @@ export declare const diffusionPlugin: {
             v: "v";
             edm_v: "edm_v";
             flow: "flow";
-            flux_flow: "flux_flow";
             flux2_flow: "flux2_flow";
         }>>;
         type: import("zod").ZodOptional<import("zod").ZodEnum<{
@@ -52,7 +60,9 @@ export declare const diffusionPlugin: {
         clip_on_cpu: import("zod").ZodOptional<import("zod").ZodBoolean>;
         vae_on_cpu: import("zod").ZodOptional<import("zod").ZodBoolean>;
         vae_tiling: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        offload_to_cpu: import("zod").ZodOptional<import("zod").ZodBoolean>;
         flash_attn: import("zod").ZodOptional<import("zod").ZodBoolean>;
+        diffusion_fa: import("zod").ZodOptional<import("zod").ZodBoolean>;
         lora_apply_mode: import("zod").ZodOptional<import("zod").ZodEnum<{
             auto: "auto";
             immediately: "immediately";
@@ -75,18 +85,23 @@ export declare const diffusionPlugin: {
                 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";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         clipGModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -105,18 +120,23 @@ export declare const diffusionPlugin: {
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         t5XxlModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -135,18 +155,23 @@ export declare const diffusionPlugin: {
                 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";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         llmModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -165,18 +190,23 @@ export declare const diffusionPlugin: {
                 diffusion: "diffusion";
                 tts: "tts";
                 ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         vaeModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -195,18 +225,58 @@ export declare const diffusionPlugin: {
                 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";
+            }>, import("zod").ZodLiteral<"vad">]>>;
+        }, import("zod/v4/core").$strip>]>>;
+        highNoiseDiffusionModelSrc: 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";
+                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";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         upscaler: import("zod").ZodOptional<import("zod").ZodObject<{
@@ -227,18 +297,23 @@ export declare const diffusionPlugin: {
                     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";
                 }>, import("zod").ZodLiteral<"vad">]>>;
             }, import("zod/v4/core").$strip>]>>;
             tile_size: import("zod").ZodOptional<import("zod").ZodNumber>;
@@ -324,6 +399,125 @@ export declare const diffusionPlugin: {
                 seed: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$strip>>;
         }, import("zod/v4/core").$strip>>;
+        videoStream: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodObject<{
+            modelId: import("zod").ZodString;
+            requestId: import("zod").ZodOptional<import("zod").ZodString>;
+            mode: import("zod").ZodEnum<{
+                txt2vid: "txt2vid";
+            }>;
+            prompt: import("zod").ZodString;
+            negative_prompt: import("zod").ZodOptional<import("zod").ZodString>;
+            width: import("zod").ZodOptional<import("zod").ZodNumber>;
+            height: import("zod").ZodOptional<import("zod").ZodNumber>;
+            video_frames: import("zod").ZodOptional<import("zod").ZodNumber>;
+            fps: import("zod").ZodOptional<import("zod").ZodNumber>;
+            seed: import("zod").ZodOptional<import("zod").ZodNumber>;
+            steps: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sampling_method: import("zod").ZodOptional<import("zod").ZodEnum<{
+                euler: "euler";
+                euler_a: "euler_a";
+                heun: "heun";
+                dpm2: "dpm2";
+                "dpm++2m": "dpm++2m";
+                "dpm++2mv2": "dpm++2mv2";
+                "dpm++2s_a": "dpm++2s_a";
+                lcm: "lcm";
+                ipndm: "ipndm";
+                ipndm_v: "ipndm_v";
+                ddim_trailing: "ddim_trailing";
+                tcd: "tcd";
+                res_multistep: "res_multistep";
+                res_2s: "res_2s";
+            }>>;
+            scheduler: import("zod").ZodOptional<import("zod").ZodEnum<{
+                lcm: "lcm";
+                discrete: "discrete";
+                karras: "karras";
+                exponential: "exponential";
+                ays: "ays";
+                gits: "gits";
+                sgm_uniform: "sgm_uniform";
+                simple: "simple";
+                smoothstep: "smoothstep";
+                kl_optimal: "kl_optimal";
+                bong_tangent: "bong_tangent";
+            }>>;
+            cfg_scale: import("zod").ZodOptional<import("zod").ZodNumber>;
+            flow_shift: import("zod").ZodOptional<import("zod").ZodNumber>;
+            high_noise_steps: import("zod").ZodOptional<import("zod").ZodNumber>;
+            high_noise_sampler: import("zod").ZodOptional<import("zod").ZodEnum<{
+                euler: "euler";
+                euler_a: "euler_a";
+                heun: "heun";
+                dpm2: "dpm2";
+                "dpm++2m": "dpm++2m";
+                "dpm++2mv2": "dpm++2mv2";
+                "dpm++2s_a": "dpm++2s_a";
+                lcm: "lcm";
+                ipndm: "ipndm";
+                ipndm_v: "ipndm_v";
+                ddim_trailing: "ddim_trailing";
+                tcd: "tcd";
+                res_multistep: "res_multistep";
+                res_2s: "res_2s";
+            }>>;
+            high_noise_scheduler: import("zod").ZodOptional<import("zod").ZodEnum<{
+                lcm: "lcm";
+                discrete: "discrete";
+                karras: "karras";
+                exponential: "exponential";
+                ays: "ays";
+                gits: "gits";
+                sgm_uniform: "sgm_uniform";
+                simple: "simple";
+                smoothstep: "smoothstep";
+                kl_optimal: "kl_optimal";
+                bong_tangent: "bong_tangent";
+            }>>;
+            high_noise_cfg_scale: import("zod").ZodOptional<import("zod").ZodNumber>;
+            high_noise_flow_shift: import("zod").ZodOptional<import("zod").ZodNumber>;
+            moe_boundary: import("zod").ZodOptional<import("zod").ZodNumber>;
+            vace_strength: import("zod").ZodOptional<import("zod").ZodNumber>;
+            control_frames: import("zod").ZodOptional<import("zod").ZodArray<import("zod").ZodString>>;
+            vae_tiling: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            vae_tile_size: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodNumber, import("zod").ZodString]>>;
+            vae_tile_overlap: import("zod").ZodOptional<import("zod").ZodNumber>;
+            cache_mode: import("zod").ZodOptional<import("zod").ZodEnum<{
+                disabled: "disabled";
+                easycache: "easycache";
+                ucache: "ucache";
+                dbcache: "dbcache";
+                taylorseer: "taylorseer";
+                "cache-dit": "cache-dit";
+            }>>;
+            cache_preset: import("zod").ZodOptional<import("zod").ZodString>;
+            cache_threshold: import("zod").ZodOptional<import("zod").ZodNumber>;
+        }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
+            type: import("zod").ZodLiteral<"videoStream">;
+            step: import("zod").ZodOptional<import("zod").ZodNumber>;
+            totalSteps: import("zod").ZodOptional<import("zod").ZodNumber>;
+            elapsedMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+            data: import("zod").ZodOptional<import("zod").ZodString>;
+            outputIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            done: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            stats: import("zod").ZodOptional<import("zod").ZodObject<{
+                seed: import("zod").ZodOptional<import("zod").ZodNumber>;
+                modelLoadMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                generationMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalGenerationMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalWallMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalSteps: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalGenerations: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalImages: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalPixels: import("zod").ZodOptional<import("zod").ZodNumber>;
+                width: import("zod").ZodOptional<import("zod").ZodNumber>;
+                height: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalVideos: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalVideoFrames: import("zod").ZodOptional<import("zod").ZodNumber>;
+                videoFrames: import("zod").ZodOptional<import("zod").ZodNumber>;
+                fps: import("zod").ZodOptional<import("zod").ZodNumber>;
+            }, import("zod/v4/core").$strip>>;
+        }, import("zod/v4/core").$strip>>;
         upscaleStream: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodObject<{
             modelId: import("zod").ZodString;
             image: import("zod").ZodString;
@@ -344,6 +538,10 @@ export declare const diffusionPlugin: {
                 width: import("zod").ZodOptional<import("zod").ZodNumber>;
                 height: import("zod").ZodOptional<import("zod").ZodNumber>;
                 repeats: import("zod").ZodOptional<import("zod").ZodNumber>;
+                backendDevice: import("zod").ZodOptional<import("zod").ZodEnum<{
+                    cpu: "cpu";
+                    gpu: "gpu";
+                }>>;
             }, import("zod/v4/core").$strip>>;
         }, import("zod/v4/core").$strip>>;
     };

package/dist/server/bare/plugins/sdcpp-generation/plugin.js

@@ -1,9 +1,10 @@
-import ImgStableDiffusion, { EsrganUpscaler, } from "@qvac/diffusion-cpp";
+import ImgStableDiffusion, { EsrganUpscaler, VideoStableDiffusion, } from "@qvac/diffusion-cpp";
 import addonLogging from "@qvac/diffusion-cpp/addonLogging.js";
-import { definePlugin, defineHandler, sdcppConfigSchema, diffusionRequestSchema, diffusionStreamResponseSchema, upscaleRequestSchema, upscaleStreamResponseSchema, ModelType, ADDON_DIFFUSION, } from "../../../../schemas/index.js";
+import { definePlugin, defineHandler, sdcppConfigSchema, diffusionRequestSchema, diffusionStreamResponseSchema, videoRequestSchema, videoStreamResponseSchema, upscaleRequestSchema, upscaleStreamResponseSchema, ModelType, ADDON_DIFFUSION, } from "../../../../schemas/index.js";
 import { createStreamLogger, registerAddonLogger } from "../../../../logging/index.js";
 import { ModelLoadFailedError } from "../../../../utils/errors-server.js";
 import { diffusion } from "./ops/diffusion.js";
+import { video } from "./ops/video.js";
 import { upscale } from "./ops/upscale.js";
 // Single source of truth for `SdcppConfig.upscaler.*` → addon-config key
 // mapping. Used by both the diffusion-mode (post-generation upscaler) and the
@@ -27,12 +28,24 @@ function flattenUpscalerKeys(upscaler) {
         }),
     };
 }
+// `mode: "upscale"` builds an EsrganUpscaler directly (not via SdCtx), so the
+// top-level `device` field has to be forwarded explicitly — it is not part of
+// the `upscaler.*` block.
 function toEsrganAddonConfig(config) {
     return {
         ...flattenUpscalerKeys(config.upscaler),
+        ...(config.device !== undefined && { device: config.device }),
         ...(config.verbosity !== undefined && { verbosity: config.verbosity }),
     };
 }
+/**
+ * Stable-diffusion.cpp plugin for image diffusion, upscaling, and Wan video.
+ *
+ * Video mode is supported on React Native, but the SDK-published Wan model
+ * set is too large to load on typical mobile devices. Mobile apps should
+ * pass a `delegate` to `loadModel(...)` to run video generation on a
+ * desktop peer instead of loading the model on-device.
+ */
 export const diffusionPlugin = definePlugin({
     modelType: ModelType.sdcppGeneration,
     displayName: "Image Generation & Upscaling (stable-diffusion.cpp)",
@@ -45,27 +58,30 @@ export const diffusionPlugin = definePlugin({
         if (cfg.mode === "upscale") {
             return { config: cfg };
         }
-        const { clipLModelSrc, clipGModelSrc, t5XxlModelSrc, llmModelSrc, vaeModelSrc, upscaler, ...rest } = cfg;
-        const { model_src: esrganModelSrc, ...upscalerRuntime } = upscaler ?? {};
+        const { clipLModelSrc, clipGModelSrc, t5XxlModelSrc, llmModelSrc, vaeModelSrc, highNoiseDiffusionModelSrc, upscaler, ...rest } = cfg;
+        // Video jobs do not apply ESRGAN so we drop the whole `upscaler` object.
+        const effectiveUpscaler = cfg.mode === "video" ? undefined : upscaler;
+        const { model_src: esrganModelSrc, ...upscalerRuntime } = effectiveUpscaler ?? {};
         const runtimeConfig = {
             ...rest,
-            ...(upscaler && { upscaler: upscalerRuntime }),
+            ...(effectiveUpscaler && { upscaler: upscalerRuntime }),
         };
         const sources = {
             clipLModelSrc, clipGModelSrc, t5XxlModelSrc,
-            llmModelSrc, vaeModelSrc, esrganModelSrc,
+            llmModelSrc, vaeModelSrc, highNoiseDiffusionModelSrc, esrganModelSrc,
         };
         const hasSources = Object.values(sources).some(Boolean);
         if (!hasSources) {
             return { config: runtimeConfig };
         }
         const resolve = ctx.resolveModelPath;
-        const [clipLModelPath, clipGModelPath, t5XxlModelPath, llmModelPath, vaeModelPath, esrganModelPath,] = await Promise.all([
+        const [clipLModelPath, clipGModelPath, t5XxlModelPath, llmModelPath, vaeModelPath, highNoiseDiffusionModelPath, esrganModelPath,] = await Promise.all([
             clipLModelSrc ? resolve(clipLModelSrc) : undefined,
             clipGModelSrc ? resolve(clipGModelSrc) : undefined,
             t5XxlModelSrc ? resolve(t5XxlModelSrc) : undefined,
             llmModelSrc ? resolve(llmModelSrc) : undefined,
             vaeModelSrc ? resolve(vaeModelSrc) : undefined,
+            highNoiseDiffusionModelSrc ? resolve(highNoiseDiffusionModelSrc) : undefined,
             esrganModelSrc ? resolve(esrganModelSrc) : undefined,
         ]);
         return {
@@ -76,6 +92,7 @@ export const diffusionPlugin = definePlugin({
                 ...(t5XxlModelPath && { t5XxlModelPath }),
                 ...(llmModelPath && { llmModelPath }),
                 ...(vaeModelPath && { vaeModelPath }),
+                ...(highNoiseDiffusionModelPath && { highNoiseDiffusionModelPath }),
                 ...(esrganModelPath && { esrganModelPath }),
             },
         };
@@ -88,7 +105,7 @@ export const diffusionPlugin = definePlugin({
         // the main diffusion checkpoint. Reject early with a clear error
         // instead of letting the native addon fail mid-load. Done before any
         // logger / native side-effects so callers can recover cleanly.
-        if (config.mode !== "upscale" &&
+        if (config.mode === "diffusion" &&
             config.upscaler !== undefined &&
             !artifacts?.["esrganModelPath"]) {
             throw new ModelLoadFailedError("modelConfig.upscaler.model_src is required when modelConfig.upscaler " +
@@ -107,6 +124,35 @@ export const diffusionPlugin = definePlugin({
             });
             return { model };
         }
+        if (config.mode === "video") {
+            if (!artifacts?.["t5XxlModelPath"]) {
+                throw new ModelLoadFailedError("modelConfig.t5XxlModelSrc is required in video mode. " +
+                    "Provide the Wan text encoder model before loading the video pipeline.");
+            }
+            if (!artifacts?.["vaeModelPath"]) {
+                throw new ModelLoadFailedError("modelConfig.vaeModelSrc is required in video mode. " +
+                    "Provide the Wan VAE model before loading the video pipeline.");
+            }
+            const files = {
+                model: modelPath,
+                t5Xxl: artifacts["t5XxlModelPath"],
+                vae: artifacts["vaeModelPath"],
+                ...(artifacts?.["highNoiseDiffusionModelPath"] && {
+                    highNoiseDiffusionModel: artifacts["highNoiseDiffusionModelPath"],
+                }),
+                ...(artifacts?.["esrganModelPath"] && { esrgan: artifacts["esrganModelPath"] }),
+            };
+            /* eslint-disable @typescript-eslint/no-unused-vars */
+            const { clipLModelSrc, clipGModelSrc, t5XxlModelSrc, llmModelSrc, vaeModelSrc, highNoiseDiffusionModelSrc, upscaler, mode, ...rest } = config;
+            /* eslint-enable @typescript-eslint/no-unused-vars */
+            const model = new VideoStableDiffusion({
+                files,
+                config: rest,
+                logger,
+                opts: { stats: true },
+            });
+            return { model };
+        }
         const files = {
             model: modelPath,
             ...(artifacts?.["clipLModelPath"] && { clipL: artifacts["clipLModelPath"] }),
@@ -143,6 +189,13 @@ export const diffusionPlugin = definePlugin({
             cancel: { scope: "model", hard: true },
             handler: diffusion,
         }),
+        videoStream: defineHandler({
+            requestSchema: videoRequestSchema,
+            responseSchema: videoStreamResponseSchema,
+            streaming: true,
+            cancel: { scope: "model", hard: true },
+            handler: video,
+        }),
         upscaleStream: defineHandler({
             requestSchema: upscaleRequestSchema,
             responseSchema: upscaleStreamResponseSchema,

package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech-stream.js

@@ -10,24 +10,14 @@ function hasRunStreaming(model) {
         "runStreaming" in model &&
         typeof model.runStreaming === "function");
 }
-// Callers (e.g. LLM token deltas) may write raw Buffer chunks whose
-// boundaries split a multi-byte UTF-8 codepoint. A naive `buf.toString("utf8")`
-// per chunk emits U+FFFD at every split point and silently corrupts CJK,
-// emoji, accented scripts, etc. before they reach the tokenizer. Keep a
-// tail-buffer of at most 3 bytes spanning an incomplete trailing codepoint
-// and decode only the complete-codepoint prefix of each combined chunk.
 function findLastCompleteUtf8End(buf) {
     const len = buf.length;
-    // A UTF-8 codepoint is at most 4 bytes, so a split can only dangle in
-    // the last 3 bytes. Scan backwards to find the most recent leading byte.
     for (let i = len - 1; i >= 0 && i >= len - 3; i--) {
         const b = buf[i];
         if ((b & 0x80) === 0) {
-            // ASCII byte: everything up to and including i is complete.
             return i + 1;
         }
         if ((b & 0xc0) === 0xc0) {
-            // Leading byte of a multi-byte sequence.
             let expected;
             if ((b & 0xe0) === 0xc0)
                 expected = 2;
@@ -36,7 +26,7 @@ function findLastCompleteUtf8End(buf) {
             else if ((b & 0xf8) === 0xf0)
                 expected = 4;
             else
-                return len; // Invalid leader; let toString emit U+FFFD.
+                return len;
             return i + expected <= len ? len : i;
         }
     }
@@ -59,9 +49,6 @@ async function* buffersToUtf8Fragments(inputStream) {
                 : Buffer.alloc(0);
     }
     if (pending.length > 0) {
-        // Stream ended with a dangling partial codepoint — flush as-is so the
-        // tokenizer sees the replacement character rather than silently dropping
-        // the final byte(s).
         const s = pending.toString("utf8");
         if (s.length > 0) {
             yield s;
@@ -89,7 +76,7 @@ export async function* textToSpeechStream(params, inputStream) {
     const model = getModel(request.modelId);
     const modelStart = nowMs();
     if (!hasRunStreaming(model)) {
-        throw new TextToSpeechStreamFailedError("textToSpeechStream requires an ONNX TTS model with runStreaming");
+        throw new TextToSpeechStreamFailedError("textToSpeechStream requires a TTS model with runStreaming");
     }
     const textSource = buffersToUtf8Fragments(inputStream);
     const streamOpts = buildRunStreamingOptions(request);

package/dist/server/bare/plugins/{onnx-tts → tts-ggml}/ops/text-to-speech.js

@@ -16,7 +16,7 @@ export async function* textToSpeech(params) {
     const modelStart = nowMs();
     if (sentenceStream) {
         if (!hasRunStream(model)) {
-            throw new TextToSpeechFailedError("sentenceStream requires an ONNX TTS model with runStream (Supertonic)");
+            throw new TextToSpeechFailedError("sentenceStream requires a TTS model with runStream");
         }
         const streamOpts = sentenceStreamLocale !== undefined || sentenceStreamMaxChunkScalars !== undefined
             ? {
@@ -62,6 +62,7 @@ export async function* textToSpeech(params) {
     const response = (await model.run({
         input: text,
         inputType,
+        ...(stream ? { streamOutput: true } : {}),
     }));
     if (!stream) {
         let completeBuffer = [];

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

@@ -0,0 +1,75 @@
+import { type CreateModelParams, type PluginModelResult, type ResolveContext, type ResolveResult } from "../../../../schemas/index";
+export declare const ttsPlugin: {
+    modelType: "tts-ggml";
+    displayName: string;
+    addonPackage: "@qvac/tts-ggml";
+    loadConfigSchema: import("zod").ZodUnion<readonly [import("zod").ZodObject<{
+        [x: string]: import("zod").ZodOptional<import("zod").ZodUnknown>;
+    }, import("zod/v4/core").$strict>, import("zod").ZodObject<{
+        [x: string]: import("zod").ZodOptional<import("zod").ZodUnknown>;
+    }, import("zod/v4/core").$strict>]>;
+    resolveConfig(cfg: Record<string, unknown>, ctx: ResolveContext): Promise<ResolveResult<{
+        ttsEngine: "supertonic";
+        language: "en" | "de" | "es" | "it";
+        voice?: string | undefined;
+        ttsSpeed?: number | undefined;
+        ttsNumInferenceSteps?: number | undefined;
+        useGPU?: boolean | undefined;
+    } | {
+        ttsEngine: "chatterbox";
+        language: "en" | "de" | "es" | "it";
+        voice?: string | undefined;
+        useGPU?: boolean | undefined;
+    }, 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>;
+            sentenceStream: import("zod").ZodDefault<import("zod").ZodBoolean>;
+            sentenceStreamLocale: import("zod").ZodOptional<import("zod").ZodString>;
+            sentenceStreamMaxChunkScalars: import("zod").ZodOptional<import("zod").ZodNumber>;
+            type: import("zod").ZodLiteral<"textToSpeech">;
+        }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
+            type: import("zod").ZodLiteral<"textToSpeech">;
+            buffer: import("zod").ZodArray<import("zod").ZodNumber>;
+            done: import("zod").ZodDefault<import("zod").ZodBoolean>;
+            stats: import("zod").ZodOptional<import("zod").ZodObject<{
+                audioDuration: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalSamples: import("zod").ZodOptional<import("zod").ZodNumber>;
+            }, import("zod/v4/core").$strip>>;
+            chunkIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sentenceChunk: import("zod").ZodOptional<import("zod").ZodString>;
+        }, import("zod/v4/core").$strip>>;
+        textToSpeechStream: import("../../../../schemas/index").PluginHandlerDefinition<import("zod").ZodObject<{
+            modelId: import("zod").ZodString;
+            inputType: import("zod").ZodDefault<import("zod").ZodString>;
+            accumulateSentences: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            sentenceDelimiterPreset: import("zod").ZodOptional<import("zod").ZodEnum<{
+                latin: "latin";
+                cjk: "cjk";
+                multilingual: "multilingual";
+            }>>;
+            maxBufferScalars: import("zod").ZodOptional<import("zod").ZodNumber>;
+            flushAfterMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+            type: import("zod").ZodLiteral<"textToSpeechStream">;
+        }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
+            type: import("zod").ZodLiteral<"textToSpeechStream">;
+            buffer: import("zod").ZodArray<import("zod").ZodNumber>;
+            done: import("zod").ZodDefault<import("zod").ZodBoolean>;
+            stats: import("zod").ZodOptional<import("zod").ZodObject<{
+                audioDuration: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalSamples: import("zod").ZodOptional<import("zod").ZodNumber>;
+            }, import("zod/v4/core").$strip>>;
+            chunkIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            sentenceChunk: import("zod").ZodOptional<import("zod").ZodString>;
+        }, import("zod/v4/core").$strip>>;
+    };
+    logging: {
+        module: import("@qvac/tts-ggml/addonLogging").AddonLogging;
+        namespace: "tts-ggml";
+    };
+};
+//# sourceMappingURL=plugin.d.ts.map