@qvac/sdk 0.10.2 → 0.12.0

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

@@ -1,10 +1,23 @@
 import { type CreateModelParams, type PluginModelResult, type ResolveContext, type ResolveResult, type SdcppConfig } from "../../../../schemas/index";
-type DiffusionArtifactKey = "clipLModelPath" | "clipGModelPath" | "t5XxlModelPath" | "llmModelPath" | "vaeModelPath";
+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;
     addonPackage: "@qvac/diffusion-cpp";
     loadConfigSchema: import("zod").ZodObject<{
+        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<{
             cpu: "cpu";
@@ -16,7 +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<{
@@ -48,7 +60,14 @@ 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";
+            at_runtime: "at_runtime";
+        }>>;
         verbosity: import("zod").ZodOptional<import("zod").ZodNumber>;
         clipLModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
             src: import("zod").ZodString;
@@ -63,21 +82,26 @@ export declare const diffusionPlugin: {
             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";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         clipGModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -93,21 +117,26 @@ export declare const diffusionPlugin: {
             sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
             addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         t5XxlModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -123,21 +152,26 @@ export declare const diffusionPlugin: {
             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";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         llmModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -153,21 +187,26 @@ export declare const diffusionPlugin: {
             sha256Checksum: import("zod").ZodOptional<import("zod").ZodString>;
             addon: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
+                vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
+                "tts-ggml": "tts-ggml";
                 "parakeet-transcription": "parakeet-transcription";
                 "onnx-ocr": "onnx-ocr";
                 "sdcpp-generation": "sdcpp-generation";
+                "ggml-vla": "ggml-vla";
+                "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
+                classification: "classification";
             }>, import("zod").ZodLiteral<"vad">]>>;
         }, import("zod/v4/core").$strip>]>>;
         vaeModelSrc: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodString, import("zod").ZodObject<{
@@ -183,23 +222,105 @@ export declare const diffusionPlugin: {
             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>]>>;
+        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";
-                diffusion: "diffusion";
+                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<{
+            type: import("zod").ZodOptional<import("zod").ZodLiteral<"esrgan">>;
+            model_src: 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").$strip>]>>;
+            tile_size: import("zod").ZodOptional<import("zod").ZodNumber>;
+            direct: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            offload_params_to_cpu: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            threads: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodLiteral<-1>, import("zod").ZodNumber]>>;
+        }, import("zod/v4/core").$strict>>;
     }, import("zod/v4/core").$strip>;
     resolveConfig(cfg: SdcppConfig, ctx: ResolveContext): Promise<ResolveResult<SdcppConfig, DiffusionArtifactKey>>;
     createModel(params: CreateModelParams): PluginModelResult;
@@ -248,7 +369,14 @@ export declare const diffusionPlugin: {
             vae_tiling: import("zod").ZodOptional<import("zod").ZodBoolean>;
             cache_preset: import("zod").ZodOptional<import("zod").ZodString>;
             init_image: import("zod").ZodOptional<import("zod").ZodString>;
+            init_images: import("zod").ZodOptional<import("zod").ZodArray<import("zod").ZodString>>;
+            increase_ref_index: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            auto_resize_ref_image: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            lora: import("zod").ZodOptional<import("zod").ZodString>;
             strength: import("zod").ZodOptional<import("zod").ZodNumber>;
+            upscale: import("zod").ZodOptional<import("zod").ZodUnion<readonly [import("zod").ZodBoolean, import("zod").ZodObject<{
+                repeats: import("zod").ZodOptional<import("zod").ZodNumber>;
+            }, import("zod/v4/core").$strict>]>>;
         }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
             type: import("zod").ZodLiteral<"diffusionStream">;
             step: import("zod").ZodOptional<import("zod").ZodNumber>;
@@ -271,6 +399,151 @@ export declare const diffusionPlugin: {
                 seed: import("zod").ZodOptional<import("zod").ZodNumber>;
             }, import("zod/v4/core").$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;
+            repeats: import("zod").ZodOptional<import("zod").ZodNumber>;
+        }, import("zod/v4/core").$strip>, import("zod").ZodObject<{
+            type: import("zod").ZodLiteral<"upscaleStream">;
+            data: import("zod").ZodOptional<import("zod").ZodString>;
+            outputIndex: import("zod").ZodOptional<import("zod").ZodNumber>;
+            done: import("zod").ZodOptional<import("zod").ZodBoolean>;
+            stats: import("zod").ZodOptional<import("zod").ZodObject<{
+                modelLoadMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                upscaleMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalUpscaleMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalWallMs: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalUpscales: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalImages: import("zod").ZodOptional<import("zod").ZodNumber>;
+                totalPixels: import("zod").ZodOptional<import("zod").ZodNumber>;
+                width: import("zod").ZodOptional<import("zod").ZodNumber>;
+                height: import("zod").ZodOptional<import("zod").ZodNumber>;
+                repeats: import("zod").ZodOptional<import("zod").ZodNumber>;
+                backendDevice: import("zod").ZodOptional<import("zod").ZodEnum<{
+                    cpu: "cpu";
+                    gpu: "gpu";
+                }>>;
+            }, import("zod/v4/core").$strip>>;
+        }, import("zod/v4/core").$strip>>;
     };
     logging: {
         module: import("@qvac/diffusion-cpp/addonLogging").AddonLogging;

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

@@ -1,27 +1,88 @@
-import ImgStableDiffusion, {} 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, 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
+// standalone-upscale-mode branches; keeping the mapping in one place avoids
+// drift if `@qvac/diffusion-cpp` ever adds or renames an `upscaler_*` key.
+function flattenUpscalerKeys(upscaler) {
+    if (!upscaler)
+        return {};
+    return {
+        ...(upscaler.tile_size !== undefined && {
+            upscaler_tile_size: upscaler.tile_size,
+        }),
+        ...(upscaler.direct !== undefined && {
+            upscaler_direct: upscaler.direct,
+        }),
+        ...(upscaler.offload_params_to_cpu !== undefined && {
+            upscaler_offload_params_to_cpu: upscaler.offload_params_to_cpu,
+        }),
+        ...(upscaler.threads !== undefined && {
+            upscaler_threads: upscaler.threads,
+        }),
+    };
+}
+// `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 (stable-diffusion.cpp)",
+    displayName: "Image Generation & Upscaling (stable-diffusion.cpp)",
     addonPackage: ADDON_DIFFUSION,
     loadConfigSchema: sdcppConfigSchema,
     async resolveConfig(cfg, ctx) {
-        const { clipLModelSrc, clipGModelSrc, t5XxlModelSrc, llmModelSrc, vaeModelSrc, ...runtimeConfig } = cfg;
-        const sources = { clipLModelSrc, clipGModelSrc, t5XxlModelSrc, llmModelSrc, vaeModelSrc };
+        // Standalone-upscaler mode never references auxiliary models: the primary
+        // modelSrc IS the ESRGAN file. Skip resolution to avoid downloading
+        // unused encoders/VAEs and to keep load fast.
+        if (cfg.mode === "upscale") {
+            return { config: cfg };
+        }
+        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,
+            ...(effectiveUpscaler && { upscaler: upscalerRuntime }),
+        };
+        const sources = {
+            clipLModelSrc, clipGModelSrc, t5XxlModelSrc,
+            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] = 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 {
             config: runtimeConfig,
@@ -31,14 +92,67 @@ export const diffusionPlugin = definePlugin({
                 ...(t5XxlModelPath && { t5XxlModelPath }),
                 ...(llmModelPath && { llmModelPath }),
                 ...(vaeModelPath && { vaeModelPath }),
+                ...(highNoiseDiffusionModelPath && { highNoiseDiffusionModelPath }),
+                ...(esrganModelPath && { esrganModelPath }),
             },
         };
     },
     createModel(params) {
         const { modelId, modelPath, modelConfig, artifacts } = params;
         const config = (modelConfig ?? {});
+        // In diffusion mode the ESRGAN file (when post-generation upscale is
+        // wanted) must come from upscaler.model_src — the primary modelPath is
+        // 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 === "diffusion" &&
+            config.upscaler !== undefined &&
+            !artifacts?.["esrganModelPath"]) {
+            throw new ModelLoadFailedError("modelConfig.upscaler.model_src is required when modelConfig.upscaler " +
+                "is set in diffusion mode. Provide the ESRGAN model, omit the " +
+                "upscaler block, or switch to modelConfig.mode = 'upscale' to load " +
+                "a standalone upscaler.");
+        }
         const logger = createStreamLogger(modelId, ModelType.sdcppGeneration);
         registerAddonLogger(modelId, ModelType.sdcppGeneration, logger);
+        if (config.mode === "upscale") {
+            const model = new EsrganUpscaler({
+                files: { esrgan: modelPath },
+                config: toEsrganAddonConfig(config),
+                logger,
+                opts: { stats: true },
+            });
+            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"] }),
@@ -46,10 +160,20 @@ export const diffusionPlugin = definePlugin({
             ...(artifacts?.["t5XxlModelPath"] && { t5Xxl: artifacts["t5XxlModelPath"] }),
             ...(artifacts?.["llmModelPath"] && { llm: artifacts["llmModelPath"] }),
             ...(artifacts?.["vaeModelPath"] && { vae: artifacts["vaeModelPath"] }),
+            ...(artifacts?.["esrganModelPath"] && { esrgan: artifacts["esrganModelPath"] }),
+        };
+        // `mode` is consumed by this plugin to select the model class above; the
+        // stable-diffusion.cpp native config does not understand it. Strip it
+        // before forwarding `rest` to the addon.
+        // eslint-disable-next-line @typescript-eslint/no-unused-vars
+        const { upscaler, mode, ...rest } = config;
+        const addonConfig = {
+            ...rest,
+            ...flattenUpscalerKeys(upscaler),
         };
         const model = new ImgStableDiffusion({
             files,
-            config: config,
+            config: addonConfig,
             logger,
             opts: { stats: true },
         });
@@ -60,8 +184,27 @@ export const diffusionPlugin = definePlugin({
             requestSchema: diffusionRequestSchema,
             responseSchema: diffusionStreamResponseSchema,
             streaming: true,
+            // sdcpp diffusion exposes a model-wide hard cancel — compute
+            // is interrupted on the currently-running generation.
+            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,
+            streaming: true,
+            // sdcpp upscale path has no cancel surface today — SDK falls
+            // back to soft-cancel.
+            cancel: { scope: "none" },
+            handler: upscale,
+        }),
     },
     logging: {
         module: addonLogging,