@qvac/sdk 0.12.2 → 0.13.0

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

@@ -75,6 +75,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -86,6 +87,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -110,6 +112,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -121,6 +124,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -145,6 +149,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -156,6 +161,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -180,6 +186,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -191,6 +198,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -215,6 +223,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -226,6 +235,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -250,6 +260,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             vla: "vla";
             "llamacpp-completion": "llamacpp-completion";
             "whispercpp-transcription": "whispercpp-transcription";
+            "bci-whispercpp-transcription": "bci-whispercpp-transcription";
             "llamacpp-embedding": "llamacpp-embedding";
             "nmtcpp-translation": "nmtcpp-translation";
             "onnx-tts": "onnx-tts";
@@ -261,6 +272,44 @@ export declare const sdcppConfigSchema: z.ZodObject<{
             "ggml-classification": "ggml-classification";
             llm: "llm";
             whisper: "whisper";
+            bci: "bci";
+            nmt: "nmt";
+            parakeet: "parakeet";
+            classification: "classification";
+        }>, z.ZodLiteral<"vad">]>>;
+    }, z.core.$strip>]>>;
+    clipVisionModelSrc: 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";
+            "bci-whispercpp-transcription": "bci-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";
+            bci: "bci";
             nmt: "nmt";
             parakeet: "parakeet";
             classification: "classification";
@@ -287,6 +336,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
                 vla: "vla";
                 "llamacpp-completion": "llamacpp-completion";
                 "whispercpp-transcription": "whispercpp-transcription";
+                "bci-whispercpp-transcription": "bci-whispercpp-transcription";
                 "llamacpp-embedding": "llamacpp-embedding";
                 "nmtcpp-translation": "nmtcpp-translation";
                 "onnx-tts": "onnx-tts";
@@ -298,6 +348,7 @@ export declare const sdcppConfigSchema: z.ZodObject<{
                 "ggml-classification": "ggml-classification";
                 llm: "llm";
                 whisper: "whisper";
+                bci: "bci";
                 nmt: "nmt";
                 parakeet: "parakeet";
                 classification: "classification";
@@ -512,9 +563,6 @@ export type DiffusionClientParams = DiffusionClientParamsBase & ({
 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>;
@@ -602,14 +650,17 @@ export declare const videoRequestSchema: z.ZodObject<{
     }>>;
     cache_preset: z.ZodOptional<z.ZodString>;
     cache_threshold: z.ZodOptional<z.ZodNumber>;
+    mode: z.ZodEnum<{
+        txt2vid: "txt2vid";
+        img2vid: "img2vid";
+    }>;
+    init_image: z.ZodOptional<z.ZodString>;
+    strength: 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>;
@@ -697,13 +748,29 @@ export declare const videoStreamRequestSchema: z.ZodObject<{
     }>>;
     cache_preset: z.ZodOptional<z.ZodString>;
     cache_threshold: z.ZodOptional<z.ZodNumber>;
+    mode: z.ZodEnum<{
+        txt2vid: "txt2vid";
+        img2vid: "img2vid";
+    }>;
+    init_image: z.ZodOptional<z.ZodString>;
+    strength: 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 & {
+type VideoClientParamsCommon = Omit<VideoRequest, "requestId" | "mode" | "init_image" | "strength" | "control_frames"> & {
     control_frames?: Uint8Array[];
 };
+export type VideoTxt2vidClientParams = VideoClientParamsCommon & {
+    mode: "txt2vid";
+    init_image?: never;
+    strength?: never;
+};
+export type VideoImg2vidClientParams = VideoClientParamsCommon & {
+    mode: "img2vid";
+    init_image: Uint8Array;
+    strength?: number;
+};
+export type VideoClientParams = VideoTxt2vidClientParams | VideoImg2vidClientParams;
 export declare const upscaleStatsSchema: z.ZodObject<{
     modelLoadMs: z.ZodOptional<z.ZodNumber>;
     upscaleMs: z.ZodOptional<z.ZodNumber>;

package/dist/schemas/sdcpp-config.js

@@ -89,6 +89,9 @@ export const sdcppConfigSchema = z
         .describe("VAE decoder model — required for FLUX.2 [klein], optional for SDXL"),
     highNoiseDiffusionModelSrc: modelSrcInputSchema.optional()
         .describe("High-noise diffusion expert — required for Wan 2.2 mixture-of-experts video models"),
+    clipVisionModelSrc: modelSrcInputSchema.optional()
+        .describe("OpenCLIP ViT-H/14 weights (`clip_vision_h.safetensors`). Required for " +
+        "Wan image-to-video (`img2vid`); omit for text-to-video-only pipelines."),
     upscaler: z.object({
         type: z.literal("esrgan").optional()
             .describe("Type of upscaler to use for post-generation upscaling when requested in diffusion({ upscale })."),
@@ -344,7 +347,7 @@ export const diffusionRequestSchema = z.object({
 export const diffusionStreamRequestSchema = diffusionRequestSchema.extend({
     type: z.literal("diffusionStream"),
 });
-export const videoRequestSchema = z.object({
+const videoGenerationBaseSchema = z.object({
     modelId: z
         .string()
         .describe("The identifier of the loaded video model to use for generation. " +
@@ -356,9 +359,6 @@ export const videoRequestSchema = z.object({
         .min(1)
         .optional()
         .describe("Stable identifier for this in-flight video generation. Optional on the wire — the server falls back to a server-generated id when the field is missing."),
-    mode: z
-        .enum(["txt2vid"])
-        .describe("Video generation mode. Only `txt2vid` is supported today."),
     prompt: z.string().describe("Positive prompt describing the video to generate."),
     negative_prompt: z
         .string()
@@ -368,16 +368,16 @@ export const videoRequestSchema = z.object({
         .number()
         .int()
         .positive()
-        .multipleOf(8)
+        .multipleOf(16)
         .optional()
-        .describe("Video width in pixels (must be a multiple of 8)."),
+        .describe("Video width in pixels (must be a multiple of 16)."),
     height: z
         .number()
         .int()
         .positive()
-        .multipleOf(8)
+        .multipleOf(16)
         .optional()
-        .describe("Video height in pixels (must be a multiple of 8)."),
+        .describe("Video height in pixels (must be a multiple of 16)."),
     video_frames: z
         .number()
         .int()
@@ -477,9 +477,55 @@ export const videoRequestSchema = z.object({
         .optional()
         .describe("Direct cache reuse threshold override."),
 });
-export const videoStreamRequestSchema = videoRequestSchema.extend({
-    type: z.literal("videoStream"),
+// Single wire object with mode-dependent rules expressed via a shared refine.
+// Keeping the wire schema a plain object (instead of a discriminated union)
+// lets the client builder construct the request without an `as` cast that
+// would otherwise disable field-level type-checking. The compile-time
+// "img2vid requires init_image" guarantee lives on the client-facing
+// discriminated union types below.
+const videoRequestObjectSchema = videoGenerationBaseSchema.extend({
+    mode: z
+        .enum(["txt2vid", "img2vid"])
+        .describe("Generation mode: 'txt2vid' (no source frame) or 'img2vid' (first-frame image)."),
+    init_image: base64StringSchema
+        .optional()
+        .describe("Base64-encoded first-frame image (PNG/JPEG). Required for img2vid; rejected for txt2vid."),
+    strength: z
+        .number()
+        .min(0)
+        .max(1)
+        .optional()
+        .describe("img2vid denoise strength in [0, 1]; rejected for txt2vid."),
 });
+function refineVideoMode(data, ctx) {
+    if (data.mode === "img2vid" && data.init_image === undefined) {
+        ctx.addIssue({
+            code: "custom",
+            path: ["init_image"],
+            message: "init_image is required when mode is 'img2vid'.",
+        });
+    }
+    if (data.mode === "txt2vid") {
+        if (data.init_image !== undefined) {
+            ctx.addIssue({
+                code: "custom",
+                path: ["init_image"],
+                message: "init_image is only valid for img2vid.",
+            });
+        }
+        if (data.strength !== undefined) {
+            ctx.addIssue({
+                code: "custom",
+                path: ["strength"],
+                message: "strength is only valid for img2vid.",
+            });
+        }
+    }
+}
+export const videoRequestSchema = videoRequestObjectSchema.superRefine(refineVideoMode);
+export const videoStreamRequestSchema = videoRequestObjectSchema
+    .extend({ type: z.literal("videoStream") })
+    .superRefine(refineVideoMode);
 // ============================================
 // Standalone ESRGAN upscale (mode: "upscale")
 // ============================================