@qvac/sdk 0.10.2 → 0.11.0

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

@@ -79,6 +79,9 @@ export declare const reloadConfigOptionsSchema: z.ZodUnion<readonly [z.ZodObject
             sha256Checksum: z.ZodOptional<z.ZodString>;
             addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
@@ -91,9 +94,6 @@ export declare const reloadConfigOptionsSchema: z.ZodUnion<readonly [z.ZodObject
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>>;
     }, z.core.$strict>;
@@ -177,6 +177,9 @@ export declare const reloadConfigOptionsToRequestSchema: z.ZodUnion<readonly [z.
             sha256Checksum: z.ZodOptional<z.ZodString>;
             addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
@@ -189,9 +192,6 @@ export declare const reloadConfigOptionsToRequestSchema: z.ZodUnion<readonly [z.
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>>;
     }, z.core.$strict>;
@@ -264,7 +264,7 @@ export declare const reloadConfigOptionsToRequestSchema: z.ZodUnion<readonly [z.
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "tts" | "ocr" | "diffusion" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
         } | undefined;
     };
 }, {
@@ -335,7 +335,7 @@ export declare const reloadConfigOptionsToRequestSchema: z.ZodUnion<readonly [z.
             engine?: string | undefined;
             expectedSize?: number | undefined;
             sha256Checksum?: string | undefined;
-            addon?: "embeddings" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "tts" | "ocr" | "diffusion" | "vad" | undefined;
+            addon?: "embeddings" | "diffusion" | "tts" | "ocr" | "llamacpp-completion" | "whispercpp-transcription" | "llamacpp-embedding" | "nmtcpp-translation" | "onnx-tts" | "parakeet-transcription" | "onnx-ocr" | "sdcpp-generation" | "llm" | "whisper" | "nmt" | "parakeet" | "vad" | undefined;
         } | undefined;
     };
 }>>]>;
@@ -423,6 +423,9 @@ export declare const reloadConfigWhisperRequestSchema: z.ZodObject<{
             sha256Checksum: z.ZodOptional<z.ZodString>;
             addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
@@ -435,9 +438,6 @@ export declare const reloadConfigWhisperRequestSchema: z.ZodObject<{
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>>;
     }, z.core.$strip>;
@@ -526,6 +526,9 @@ export declare const reloadConfigRequestSchema: z.ZodUnion<readonly [z.ZodObject
             sha256Checksum: z.ZodOptional<z.ZodString>;
             addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
                 embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
@@ -538,9 +541,6 @@ export declare const reloadConfigRequestSchema: z.ZodUnion<readonly [z.ZodObject
                 whisper: "whisper";
                 nmt: "nmt";
                 parakeet: "parakeet";
-                tts: "tts";
-                ocr: "ocr";
-                diffusion: "diffusion";
             }>, z.ZodLiteral<"vad">]>>;
         }, z.core.$strip>]>>>;
     }, z.core.$strip>;

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

@@ -1,5 +1,9 @@
 import { z } from "zod";
 export declare const sdcppConfigSchema: z.ZodObject<{
+    mode: z.ZodDefault<z.ZodEnum<{
+        diffusion: "diffusion";
+        upscale: "upscale";
+    }>>;
     threads: z.ZodOptional<z.ZodNumber>;
     device: z.ZodOptional<z.ZodEnum<{
         cpu: "cpu";
@@ -44,6 +48,11 @@ export declare const sdcppConfigSchema: z.ZodObject<{
     vae_on_cpu: z.ZodOptional<z.ZodBoolean>;
     vae_tiling: z.ZodOptional<z.ZodBoolean>;
     flash_attn: z.ZodOptional<z.ZodBoolean>;
+    lora_apply_mode: z.ZodOptional<z.ZodEnum<{
+        auto: "auto";
+        immediately: "immediately";
+        at_runtime: "at_runtime";
+    }>>;
     verbosity: z.ZodOptional<z.ZodNumber>;
     clipLModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
         src: z.ZodString;
@@ -58,6 +67,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -70,9 +82,6 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     clipGModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -88,6 +97,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -100,9 +112,6 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     t5XxlModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -118,6 +127,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -130,9 +142,6 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     llmModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -148,6 +157,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -160,9 +172,6 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
     vaeModelSrc: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
@@ -178,6 +187,9 @@ export declare const sdcppConfigSchema: z.ZodObject<{
         sha256Checksum: z.ZodOptional<z.ZodString>;
         addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
             embeddings: "embeddings";
+            diffusion: "diffusion";
+            tts: "tts";
+            ocr: "ocr";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
@@ -190,13 +202,47 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             whisper: "whisper";
             nmt: "nmt";
             parakeet: "parakeet";
-            tts: "tts";
-            ocr: "ocr";
-            diffusion: "diffusion";
         }>, z.ZodLiteral<"vad">]>>;
     }, z.core.$strip>]>>;
+    upscaler: z.ZodOptional<z.ZodObject<{
+        type: z.ZodOptional<z.ZodLiteral<"esrgan">>;
+        model_src: z.ZodOptional<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+            src: z.ZodString;
+            name: z.ZodOptional<z.ZodString>;
+            modelId: z.ZodOptional<z.ZodString>;
+            registryPath: z.ZodOptional<z.ZodString>;
+            registrySource: z.ZodOptional<z.ZodString>;
+            blobCoreKey: z.ZodOptional<z.ZodString>;
+            blobIndex: z.ZodOptional<z.ZodNumber>;
+            engine: z.ZodOptional<z.ZodString>;
+            expectedSize: z.ZodOptional<z.ZodNumber>;
+            sha256Checksum: z.ZodOptional<z.ZodString>;
+            addon: z.ZodOptional<z.ZodUnion<readonly [z.ZodEnum<{
+                embeddings: "embeddings";
+                diffusion: "diffusion";
+                tts: "tts";
+                ocr: "ocr";
+                "llamacpp-completion": "llamacpp-completion";
+                "whispercpp-transcription": "whispercpp-transcription";
+                "llamacpp-embedding": "llamacpp-embedding";
+                "nmtcpp-translation": "nmtcpp-translation";
+                "onnx-tts": "onnx-tts";
+                "parakeet-transcription": "parakeet-transcription";
+                "onnx-ocr": "onnx-ocr";
+                "sdcpp-generation": "sdcpp-generation";
+                llm: "llm";
+                whisper: "whisper";
+                nmt: "nmt";
+                parakeet: "parakeet";
+            }>, z.ZodLiteral<"vad">]>>;
+        }, z.core.$strip>]>>;
+        tile_size: z.ZodOptional<z.ZodNumber>;
+        direct: z.ZodOptional<z.ZodBoolean>;
+        offload_params_to_cpu: z.ZodOptional<z.ZodBoolean>;
+        threads: z.ZodOptional<z.ZodUnion<readonly [z.ZodLiteral<-1>, z.ZodNumber]>>;
+    }, z.core.$strict>>;
 }, z.core.$strip>;
-export type SdcppConfig = z.infer<typeof sdcppConfigSchema>;
+export type SdcppConfig = z.input<typeof sdcppConfigSchema>;
 export declare const diffusionStatsSchema: z.ZodObject<{
     modelLoadMs: z.ZodOptional<z.ZodNumber>;
     generationMs: z.ZodOptional<z.ZodNumber>;
@@ -278,7 +324,14 @@ export declare const diffusionRequestSchema: z.ZodObject<{
     vae_tiling: z.ZodOptional<z.ZodBoolean>;
     cache_preset: z.ZodOptional<z.ZodString>;
     init_image: z.ZodOptional<z.ZodString>;
+    init_images: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    increase_ref_index: z.ZodOptional<z.ZodBoolean>;
+    auto_resize_ref_image: z.ZodOptional<z.ZodBoolean>;
+    lora: z.ZodOptional<z.ZodString>;
     strength: z.ZodOptional<z.ZodNumber>;
+    upscale: z.ZodOptional<z.ZodUnion<readonly [z.ZodBoolean, z.ZodObject<{
+        repeats: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strict>]>>;
 }, z.core.$strip>;
 export type DiffusionRequest = z.input<typeof diffusionRequestSchema>;
 export declare const diffusionStreamRequestSchema: z.ZodObject<{
@@ -325,11 +378,72 @@ export declare const diffusionStreamRequestSchema: z.ZodObject<{
     vae_tiling: z.ZodOptional<z.ZodBoolean>;
     cache_preset: z.ZodOptional<z.ZodString>;
     init_image: z.ZodOptional<z.ZodString>;
+    init_images: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    increase_ref_index: z.ZodOptional<z.ZodBoolean>;
+    auto_resize_ref_image: z.ZodOptional<z.ZodBoolean>;
+    lora: z.ZodOptional<z.ZodString>;
     strength: z.ZodOptional<z.ZodNumber>;
+    upscale: z.ZodOptional<z.ZodUnion<readonly [z.ZodBoolean, z.ZodObject<{
+        repeats: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strict>]>>;
     type: z.ZodLiteral<"diffusionStream">;
 }, z.core.$strip>;
 export type DiffusionStreamRequest = z.input<typeof diffusionStreamRequestSchema>;
-export type DiffusionClientParams = Omit<DiffusionRequest, "init_image"> & {
+type DiffusionClientParamsBase = Omit<DiffusionRequest, "init_image" | "init_images">;
+export type DiffusionClientParams = DiffusionClientParamsBase & ({
     init_image?: Uint8Array;
+    init_images?: never;
+} | {
+    init_image?: never;
+    init_images?: Uint8Array[];
+});
+export declare const upscaleStatsSchema: z.ZodObject<{
+    modelLoadMs: z.ZodOptional<z.ZodNumber>;
+    upscaleMs: z.ZodOptional<z.ZodNumber>;
+    totalUpscaleMs: z.ZodOptional<z.ZodNumber>;
+    totalWallMs: z.ZodOptional<z.ZodNumber>;
+    totalUpscales: z.ZodOptional<z.ZodNumber>;
+    totalImages: z.ZodOptional<z.ZodNumber>;
+    totalPixels: z.ZodOptional<z.ZodNumber>;
+    width: z.ZodOptional<z.ZodNumber>;
+    height: z.ZodOptional<z.ZodNumber>;
+    repeats: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type UpscaleStats = z.infer<typeof upscaleStatsSchema>;
+export declare const upscaleRequestSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    image: z.ZodString;
+    repeats: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+export type UpscaleRequest = z.input<typeof upscaleRequestSchema>;
+export declare const upscaleStreamRequestSchema: z.ZodObject<{
+    modelId: z.ZodString;
+    image: z.ZodString;
+    repeats: z.ZodOptional<z.ZodNumber>;
+    type: z.ZodLiteral<"upscaleStream">;
+}, z.core.$strip>;
+export type UpscaleStreamRequest = z.input<typeof upscaleStreamRequestSchema>;
+export declare const upscaleStreamResponseSchema: z.ZodObject<{
+    type: z.ZodLiteral<"upscaleStream">;
+    data: z.ZodOptional<z.ZodString>;
+    outputIndex: z.ZodOptional<z.ZodNumber>;
+    done: z.ZodOptional<z.ZodBoolean>;
+    stats: z.ZodOptional<z.ZodObject<{
+        modelLoadMs: z.ZodOptional<z.ZodNumber>;
+        upscaleMs: z.ZodOptional<z.ZodNumber>;
+        totalUpscaleMs: z.ZodOptional<z.ZodNumber>;
+        totalWallMs: z.ZodOptional<z.ZodNumber>;
+        totalUpscales: z.ZodOptional<z.ZodNumber>;
+        totalImages: z.ZodOptional<z.ZodNumber>;
+        totalPixels: z.ZodOptional<z.ZodNumber>;
+        width: z.ZodOptional<z.ZodNumber>;
+        height: z.ZodOptional<z.ZodNumber>;
+        repeats: z.ZodOptional<z.ZodNumber>;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
+export type UpscaleStreamResponse = z.infer<typeof upscaleStreamResponseSchema>;
+export type UpscaleClientParams = Omit<UpscaleRequest, "image"> & {
+    image: Uint8Array;
 };
+export {};
 //# sourceMappingURL=sdcpp-config.d.ts.map

package/dist/schemas/sdcpp-config.js

@@ -1,8 +1,16 @@
 import { z } from "zod";
 import { modelSrcInputSchema } from "./model-src-utils.js";
 const BASE64_PATTERN = /^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/;
+const ABSOLUTE_PATH_PATTERN = /^(\/|[A-Za-z]:[\\/]|\\\\)/;
 export const sdcppConfigSchema = z
     .object({
+    mode: z.enum(["diffusion", "upscale"]).default("diffusion")
+        .describe("Operation mode for the diffusion plugin. " +
+        "`'diffusion'` (default) builds a full SD / SDXL / SD3 / FLUX pipeline from " +
+        "the primary model plus optional auxiliary text encoders, VAE, and ESRGAN " +
+        "upscaler, and exposes diffusion({ ... }). " +
+        "`'upscale'` builds a standalone ESRGAN upscaler from the primary model " +
+        "file alone (auxiliary model sources are ignored) and exposes upscale({ ... })."),
     threads: z.number().optional(),
     device: z.enum(["gpu", "cpu"]).optional(),
     prediction: z
@@ -23,6 +31,14 @@ export const sdcppConfigSchema = z
     vae_on_cpu: z.boolean().optional().describe("Force VAE decoder to run on CPU"),
     vae_tiling: z.boolean().optional().describe("Enable VAE tiling for large images on limited VRAM"),
     flash_attn: z.boolean().optional().describe("Enable flash attention to reduce memory usage"),
+    lora_apply_mode: z.enum(["auto", "immediately", "at_runtime"]).optional()
+        .describe("How LoRA adapters passed via diffusion({ lora }) are applied. " +
+        "'auto' (default): picked based on weight type — 'at_runtime' for " +
+        "quantized weights, 'immediately' for full-precision. " +
+        "'immediately': adapter is fused into the model on first use and " +
+        "persists across subsequent diffusion() calls until the model is " +
+        "unloaded. " +
+        "'at_runtime': adapter is applied per-call and not persisted."),
     verbosity: z.number().optional(),
     clipLModelSrc: modelSrcInputSchema.optional()
         .describe("CLIP-L text encoder model — required for SD3"),
@@ -34,6 +50,39 @@ export const sdcppConfigSchema = z
         .describe("LLM text encoder model (e.g. Qwen3) — required for FLUX.2 [klein]"),
     vaeModelSrc: modelSrcInputSchema.optional()
         .describe("VAE decoder model — required for FLUX.2 [klein], optional for SDXL"),
+    upscaler: z.object({
+        type: z.literal("esrgan").optional()
+            .describe("Type of upscaler to use for post-generation upscaling when requested in diffusion({ upscale })."),
+        model_src: modelSrcInputSchema.optional()
+            .describe("ESRGAN upscaler model (e.g. RealESRGAN_x4plus_anime_6B.pth). " +
+            "Required in diffusion mode when this `upscaler` block is set — " +
+            "configures the post-generation upscaler invoked via diffusion({ upscale }). " +
+            "In `mode: 'upscale'` the primary modelSrc itself is the ESRGAN model, " +
+            "so this field is ignored."),
+        tile_size: z.number().int().positive().optional()
+            .describe("ESRGAN upscaler tile size in pixels. Smaller tiles use less VRAM " +
+            "at the cost of more passes."),
+        direct: z.boolean().optional()
+            .describe("Use direct convolution in the ESRGAN upscaler instead of im2col + " +
+            "GEMM. Faster on some backends, slower on others."),
+        offload_params_to_cpu: z.boolean().optional()
+            .describe("Keep ESRGAN upscaler weights on CPU and offload them during compute. " +
+            "Trades latency for VRAM headroom on memory-constrained GPUs."),
+        threads: z.union([
+            z.literal(-1),
+            z.number().int().positive(),
+        ])
+            .optional()
+            .describe("Number of CPU threads dedicated to the ESRGAN upscaler. -1 = auto."),
+    }).strict().optional()
+        .describe("ESRGAN upscaler configuration. In diffusion mode this enables the " +
+        "post-generation upscale path invoked via diffusion({ upscale }) and " +
+        "requires `model_src`. In `mode: 'upscale'` only the tuning fields " +
+        "(tile_size, direct, offload_params_to_cpu, threads) are honored — " +
+        "the primary modelSrc IS the ESRGAN model in that mode and " +
+        "`model_src` here is ignored. Mode-dependent constraints (e.g. " +
+        "`model_src` required in diffusion mode) are enforced by the " +
+        "sdcpp-generation plugin at load time, not at the schema layer."),
 });
 export const diffusionStatsSchema = z.object({
     modelLoadMs: z
@@ -182,14 +231,127 @@ export const diffusionRequestSchema = z.object({
         .min(1)
         .regex(BASE64_PATTERN)
         .optional()
-        .describe("Base64-encoded image for img2img generation"),
+        .describe("Base64-encoded image for img2img generation. Mutually exclusive with init_images."),
+    init_images: z.array(z.string().min(1).regex(BASE64_PATTERN))
+        .min(1)
+        .optional()
+        .describe("FLUX.2-only multi-reference fusion: array of base64-encoded PNG/JPEG buffers. " +
+        "Each buffer becomes a separate reference image that the FLUX.2 transformer attends to. " +
+        "Mutually exclusive with init_image; requires the model to be loaded with " +
+        "config.prediction='flux2_flow' and a Qwen3 text encoder via llmModelSrc."),
+    increase_ref_index: z.boolean().optional()
+        .describe("FLUX.2 fusion only. When omitted, the addon default (false) is used. When false, all " +
+        "reference latents share one RoPE index slot and blend via attention (recommended for " +
+        "FLUX.2-klein). When true, each reference gets its own RoPE index slot — use only with " +
+        "text encoders that receive per-image vision tokens."),
+    auto_resize_ref_image: z.boolean().optional()
+        .describe("FLUX.2 only. When omitted, the addon default (true) is used. When true, every reference " +
+        "image (single or fusion) is auto-resized to the target width/height before VAE-encoding. " +
+        "Disable only if the buffers are already at the exact target dimensions."),
+    lora: z
+        .string()
+        .min(1)
+        .regex(ABSOLUTE_PATH_PATTERN, {
+        message: "lora must be an absolute path",
+    })
+        .optional()
+        .describe("Optional local LoRA adapter path to apply for this generation. " +
+        "Must be an absolute filesystem path. " +
+        "Whether the adapter persists across subsequent diffusion() calls is controlled " +
+        "by sdcppConfigSchema.lora_apply_mode (set at loadModel time)."),
     strength: z
         .number()
         .min(0)
         .max(1)
         .optional()
         .describe("img2img denoising strength (0.0 = keep source, 1.0 = ignore source); used by the SD/SDXL SDEdit path. No-op for FLUX.2, which uses in-context conditioning and ignores this field."),
+    upscale: z
+        .union([
+        z.boolean(),
+        z.object({
+            repeats: z.number().int().positive().optional(),
+        }).strict(),
+    ])
+        .optional()
+        .describe("Post-generation ESRGAN upscale. " +
+        "`true` (or `{}` / `{ repeats: 1 }`) runs a single upscale pass at the " +
+        "model's native scale factor (e.g. x4 for RealESRGAN_x4plus). " +
+        "`false` is a no-op (same as omitting the field). " +
+        "`{ repeats: N }` runs the upscaler N times sequentially — each pass " +
+        "multiplies the output dimensions by the model's scale factor. When " +
+        "`batch_count > 1`, every output image is upscaled independently. " +
+        "Requires the model to be loaded with `upscaler.model_src` set in modelConfig."),
+}).refine((d) => d.init_image === undefined || d.init_images === undefined, {
+    message: "init_image and init_images are mutually exclusive — pass one or the other, not both.",
 });
 export const diffusionStreamRequestSchema = diffusionRequestSchema.extend({
     type: z.literal("diffusionStream"),
 });
+// ============================================
+// Standalone ESRGAN upscale (mode: "upscale")
+// ============================================
+export const upscaleStatsSchema = z.object({
+    modelLoadMs: z
+        .number()
+        .optional()
+        .describe("Wall-clock time in milliseconds spent loading the upscaler model."),
+    upscaleMs: z
+        .number()
+        .optional()
+        .describe("Wall-clock time in milliseconds for the most recent upscale job."),
+    totalUpscaleMs: z
+        .number()
+        .optional()
+        .describe("Cumulative upscale time in milliseconds across all jobs."),
+    totalWallMs: z
+        .number()
+        .optional()
+        .describe("Total wall-clock time in milliseconds including model load and upscaling."),
+    totalUpscales: z
+        .number()
+        .optional()
+        .describe("Cumulative number of upscale calls."),
+    totalImages: z
+        .number()
+        .optional()
+        .describe("Cumulative number of images produced."),
+    totalPixels: z
+        .number()
+        .optional()
+        .describe("Cumulative number of pixels produced across all images."),
+    width: z.number().optional().describe("Width of the most recent emitted PNG."),
+    height: z.number().optional().describe("Height of the most recent emitted PNG."),
+    repeats: z
+        .number()
+        .optional()
+        .describe("Number of ESRGAN passes used by the most recent upscale job."),
+});
+export const upscaleRequestSchema = z.object({
+    modelId: z
+        .string()
+        .describe("Identifier of the loaded upscaler model. The model must have been loaded " +
+        "with `modelType: 'diffusion'` and `modelConfig.mode: 'upscale'`."),
+    image: z
+        .string()
+        .min(1)
+        .regex(BASE64_PATTERN)
+        .describe("Base64-encoded PNG/JPEG bytes of the source image."),
+    repeats: z
+        .number()
+        .int()
+        .positive()
+        .optional()
+        .describe("Number of ESRGAN passes to run sequentially. Each pass multiplies " +
+        "dimensions by the model's native scale factor; only the final image " +
+        "is emitted (`outputs.length === 1`). Defaults to 1."),
+});
+export const upscaleStreamRequestSchema = upscaleRequestSchema.extend({
+    type: z.literal("upscaleStream"),
+});
+export const upscaleStreamResponseSchema = z.object({
+    type: z.literal("upscaleStream"),
+    data: z.string().optional(),
+    outputIndex: z.number().optional(),
+    done: z.boolean().optional(),
+    stats: upscaleStatsSchema.optional(),
+});