@qvac/sdk 0.11.0 → 0.12.0

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

@@ -3,6 +3,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
     mode: z.ZodDefault<z.ZodEnum<{
         diffusion: "diffusion";
         upscale: "upscale";
+        video: "video";
     }>>;
     threads: z.ZodOptional<z.ZodNumber>;
     device: z.ZodOptional<z.ZodEnum<{
@@ -15,7 +16,6 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         v: "v";
         edm_v: "edm_v";
         flow: "flow";
-        flux_flow: "flux_flow";
         flux2_flow: "flux2_flow";
     }>>;
     type: z.ZodOptional<z.ZodEnum<{
@@ -47,7 +47,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
     clip_on_cpu: z.ZodOptional<z.ZodBoolean>;
     vae_on_cpu: z.ZodOptional<z.ZodBoolean>;
     vae_tiling: z.ZodOptional<z.ZodBoolean>;
+    offload_to_cpu: z.ZodOptional<z.ZodBoolean>;
     flash_attn: z.ZodOptional<z.ZodBoolean>;
+    diffusion_fa: z.ZodOptional<z.ZodBoolean>;
     lora_apply_mode: z.ZodOptional<z.ZodEnum<{
         auto: "auto";
         immediately: "immediately";
@@ -70,18 +72,23 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             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";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     clipGModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -100,18 +107,23 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     t5XxlModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -130,18 +142,23 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             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";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     llmModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -160,18 +177,23 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             diffusion: "diffusion";
             tts: "tts";
             ocr: "ocr";
+            vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
             "parakeet-transcription": "parakeet-transcription";
             "onnx-ocr": "onnx-ocr";
             "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     vaeModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -190,18 +212,58 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             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";
+        }>, z.ZodLiteral<"vad">]>>;
+    }, z.core.$strip>]>>;
+    highNoiseDiffusionModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+        src: z.ZodString;
+        name: z.ZodOptional<z.ZodString>;
+        modelId: z.ZodOptional<z.ZodString>;
+        registryPath: z.ZodOptional<z.ZodString>;
+        registrySource: z.ZodOptional<z.ZodString>;
+        blobCoreKey: z.ZodOptional<z.ZodString>;
+        blobIndex: z.ZodOptional<z.ZodNumber>;
+        engine: z.ZodOptional<z.ZodString>;
+        expectedSize: z.ZodOptional<z.ZodNumber>;
+        sha256Checksum: z.ZodOptional<z.ZodString>;
+        addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+            embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
+            vla: "vla";
+            "llamacpp-completion": "llamacpp-completion";
+            "whispercpp-transcription": "whispercpp-transcription";
+            "llamacpp-embedding": "llamacpp-embedding";
+            "nmtcpp-translation": "nmtcpp-translation";
+            "onnx-tts": "onnx-tts";
+            "tts-ggml": "tts-ggml";
+            "parakeet-transcription": "parakeet-transcription";
+            "onnx-ocr": "onnx-ocr";
+            "sdcpp-generation": "sdcpp-generation";
+            "ggml-vla": "ggml-vla";
+            "ggml-classification": "ggml-classification";
+            llm: "llm";
+            whisper: "whisper";
+            nmt: "nmt";
+            parakeet: "parakeet";
+            classification: "classification";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     upscaler: z.ZodOptional<z.ZodObject<{
@@ -222,18 +284,23 @@ export declare const sdcppConfigSchema: z.ZodObject<{
                 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";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>;
         tile_size: z.ZodOptional<z.ZodNumber>;
@@ -257,6 +324,24 @@ export declare const diffusionStatsSchema: z.ZodObject<{
     seed: z.ZodOptional<z.ZodNumber>;
 }, z.core.$strip>;
 export type DiffusionStats = z.infer<typeof diffusionStatsSchema>;
+export declare const videoStatsSchema: z.ZodObject<{
+    seed: z.ZodOptional<z.ZodNumber>;
+    modelLoadMs: z.ZodOptional<z.ZodNumber>;
+    generationMs: z.ZodOptional<z.ZodNumber>;
+    totalGenerationMs: z.ZodOptional<z.ZodNumber>;
+    totalWallMs: z.ZodOptional<z.ZodNumber>;
+    totalSteps: z.ZodOptional<z.ZodNumber>;
+    totalGenerations: z.ZodOptional<z.ZodNumber>;
+    totalImages: z.ZodOptional<z.ZodNumber>;
+    totalPixels: z.ZodOptional<z.ZodNumber>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    totalVideos: z.ZodOptional<z.ZodNumber>;
+    totalVideoFrames: z.ZodOptional<z.ZodNumber>;
+    videoFrames: z.ZodOptional<z.ZodNumber>;
+    fps: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type VideoStats = z.infer<typeof videoStatsSchema>;
 export declare const diffusionStreamResponseSchema: z.ZodObject<{
     type: z.ZodLiteral<"diffusionStream">;
     step: z.ZodOptional<z.ZodNumber>;
@@ -280,6 +365,33 @@ export declare const diffusionStreamResponseSchema: z.ZodObject<{
     }, z.core.$strip>>;
 }, z.core.$strip>;
 export type DiffusionStreamResponse = z.infer<typeof diffusionStreamResponseSchema>;
+export declare const videoStreamResponseSchema: z.ZodObject<{
+    type: z.ZodLiteral<"videoStream">;
+    step: z.ZodOptional<z.ZodNumber>;
+    totalSteps: z.ZodOptional<z.ZodNumber>;
+    elapsedMs: z.ZodOptional<z.ZodNumber>;
+    data: z.ZodOptional<z.ZodString>;
+    outputIndex: z.ZodOptional<z.ZodNumber>;
+    done: z.ZodOptional<z.ZodBoolean>;
+    stats: z.ZodOptional<z.ZodObject<{
+        seed: z.ZodOptional<z.ZodNumber>;
+        modelLoadMs: z.ZodOptional<z.ZodNumber>;
+        generationMs: z.ZodOptional<z.ZodNumber>;
+        totalGenerationMs: z.ZodOptional<z.ZodNumber>;
+        totalWallMs: z.ZodOptional<z.ZodNumber>;
+        totalSteps: z.ZodOptional<z.ZodNumber>;
+        totalGenerations: z.ZodOptional<z.ZodNumber>;
+        totalImages: z.ZodOptional<z.ZodNumber>;
+        totalPixels: z.ZodOptional<z.ZodNumber>;
+        width: z.ZodOptional<z.ZodNumber>;
+        height: z.ZodOptional<z.ZodNumber>;
+        totalVideos: z.ZodOptional<z.ZodNumber>;
+        totalVideoFrames: z.ZodOptional<z.ZodNumber>;
+        videoFrames: z.ZodOptional<z.ZodNumber>;
+        fps: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
+export type VideoStreamResponse = z.infer<typeof videoStreamResponseSchema>;
 export declare const diffusionRequestSchema: z.ZodObject<{
     modelId: z.ZodString;
     prompt: z.ZodString;
@@ -397,6 +509,201 @@ export type DiffusionClientParams = DiffusionClientParamsBase & ({
     init_image?: never;
     init_images?: Uint8Array[];
 });
+export declare const videoRequestSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    requestId: z.ZodOptional<z.ZodString>;
+    mode: z.ZodEnum<{
+        txt2vid: "txt2vid";
+    }>;
+    prompt: z.ZodString;
+    negative_prompt: z.ZodOptional<z.ZodString>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    video_frames: z.ZodOptional<z.ZodNumber>;
+    fps: z.ZodOptional<z.ZodNumber>;
+    seed: z.ZodOptional<z.ZodNumber>;
+    steps: z.ZodOptional<z.ZodNumber>;
+    sampling_method: z.ZodOptional<z.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: z.ZodOptional<z.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: z.ZodOptional<z.ZodNumber>;
+    flow_shift: z.ZodOptional<z.ZodNumber>;
+    high_noise_steps: z.ZodOptional<z.ZodNumber>;
+    high_noise_sampler: z.ZodOptional<z.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: z.ZodOptional<z.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: z.ZodOptional<z.ZodNumber>;
+    high_noise_flow_shift: z.ZodOptional<z.ZodNumber>;
+    moe_boundary: z.ZodOptional<z.ZodNumber>;
+    vace_strength: z.ZodOptional<z.ZodNumber>;
+    control_frames: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    vae_tiling: z.ZodOptional<z.ZodBoolean>;
+    vae_tile_size: z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>;
+    vae_tile_overlap: z.ZodOptional<z.ZodNumber>;
+    cache_mode: z.ZodOptional<z.ZodEnum<{
+        disabled: "disabled";
+        easycache: "easycache";
+        ucache: "ucache";
+        dbcache: "dbcache";
+        taylorseer: "taylorseer";
+        "cache-dit": "cache-dit";
+    }>>;
+    cache_preset: z.ZodOptional<z.ZodString>;
+    cache_threshold: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type VideoRequest = z.input<typeof videoRequestSchema>;
+export declare const videoStreamRequestSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    requestId: z.ZodOptional<z.ZodString>;
+    mode: z.ZodEnum<{
+        txt2vid: "txt2vid";
+    }>;
+    prompt: z.ZodString;
+    negative_prompt: z.ZodOptional<z.ZodString>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    video_frames: z.ZodOptional<z.ZodNumber>;
+    fps: z.ZodOptional<z.ZodNumber>;
+    seed: z.ZodOptional<z.ZodNumber>;
+    steps: z.ZodOptional<z.ZodNumber>;
+    sampling_method: z.ZodOptional<z.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: z.ZodOptional<z.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: z.ZodOptional<z.ZodNumber>;
+    flow_shift: z.ZodOptional<z.ZodNumber>;
+    high_noise_steps: z.ZodOptional<z.ZodNumber>;
+    high_noise_sampler: z.ZodOptional<z.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: z.ZodOptional<z.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: z.ZodOptional<z.ZodNumber>;
+    high_noise_flow_shift: z.ZodOptional<z.ZodNumber>;
+    moe_boundary: z.ZodOptional<z.ZodNumber>;
+    vace_strength: z.ZodOptional<z.ZodNumber>;
+    control_frames: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    vae_tiling: z.ZodOptional<z.ZodBoolean>;
+    vae_tile_size: z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>;
+    vae_tile_overlap: z.ZodOptional<z.ZodNumber>;
+    cache_mode: z.ZodOptional<z.ZodEnum<{
+        disabled: "disabled";
+        easycache: "easycache";
+        ucache: "ucache";
+        dbcache: "dbcache";
+        taylorseer: "taylorseer";
+        "cache-dit": "cache-dit";
+    }>>;
+    cache_preset: z.ZodOptional<z.ZodString>;
+    cache_threshold: z.ZodOptional<z.ZodNumber>;
+    type: z.ZodLiteral<"videoStream">;
+}, z.core.$strip>;
+export type VideoStreamRequest = z.input<typeof videoStreamRequestSchema>;
+type VideoClientParamsBase = Omit<VideoRequest, "requestId" | "control_frames">;
+export type VideoClientParams = VideoClientParamsBase & {
+    control_frames?: Uint8Array[];
+};
 export declare const upscaleStatsSchema: z.ZodObject<{
     modelLoadMs: z.ZodOptional<z.ZodNumber>;
     upscaleMs: z.ZodOptional<z.ZodNumber>;
@@ -408,6 +715,10 @@ export declare const upscaleStatsSchema: z.ZodObject<{
     width: z.ZodOptional<z.ZodNumber>;
     height: z.ZodOptional<z.ZodNumber>;
     repeats: z.ZodOptional<z.ZodNumber>;
+    backendDevice: z.ZodOptional<z.ZodEnum<{
+        cpu: "cpu";
+        gpu: "gpu";
+    }>>;
 }, z.core.$strip>;
 export type UpscaleStats = z.infer<typeof upscaleStatsSchema>;
 export declare const upscaleRequestSchema: z.ZodObject<{
@@ -439,6 +750,10 @@ export declare const upscaleStreamResponseSchema: z.ZodObject<{
         width: z.ZodOptional<z.ZodNumber>;
         height: z.ZodOptional<z.ZodNumber>;
         repeats: z.ZodOptional<z.ZodNumber>;
+        backendDevice: z.ZodOptional<z.ZodEnum<{
+            cpu: "cpu";
+            gpu: "gpu";
+        }>>;
     }, z.core.$strip>>;
 }, z.core.$strip>;
 export type UpscaleStreamResponse = z.infer<typeof upscaleStreamResponseSchema>;