@neta-art/generation 0.1.8 → 0.1.10

package/dist/{export-config-BuQ5Maoj.js → export-config-DolxdQdi.js}

@@ -1,7 +1,8 @@
-import { a as cloneJson, c as getBlockMeta, l as slugifyFileName, n as getBuiltinGenerationModel, o as compactArray, r as listBuiltinGenerationModels, s as compactObject, t as builtinGenerationModels } from "./builtins--2dnc9UT.js";
+import { a as compactArray, c as slugifyFileName, i as cloneJson, l as MODEL_SCHEMA, n as getBuiltinGenerationModel, o as compactObject, r as listBuiltinGenerationModels, s as getBlockMeta, t as builtinGenerationModels } from "./builtins-CcT0Ufp2.js";
 import { mkdir, readFile, readdir, writeFile } from "node:fs/promises";
 import { extname, join } from "node:path";
 import { parse, stringify } from "yaml";
+
 //#region src/errors.ts
 var GenerationError = class extends Error {
 	constructor(message) {
@@ -45,6 +46,7 @@ var GenerationTimeoutError = class extends GenerationProviderError {
 		this.name = "GenerationTimeoutError";
 	}
 };
+
 //#endregion
 //#region src/http.ts
 function headersToRecord(headers) {
@@ -132,6 +134,7 @@ async function fetchWithTimeout(fetchFn, url, init, timeoutMs) {
 function joinUrl(baseUrl, path) {
 	return `${baseUrl.replace(/\/+$/, "")}/${path.replace(/^\/+/, "")}`;
 }
+
 //#endregion
 //#region src/validation.ts
 function specsByType(specs) {
@@ -139,6 +142,9 @@ function specsByType(specs) {
 	for (const spec of specs) map.set(spec.type, spec);
 	return map;
 }
+function getRole(block) {
+	return block.meta?.role;
+}
 function validateGenerationContent(declaration, content) {
 	const inputSpecs = specsByType(declaration.content.input);
 	const counts = /* @__PURE__ */ new Map();
@@ -147,6 +153,11 @@ function validateGenerationContent(declaration, content) {
 		if (!spec) throw new GenerationValidationError(`Content block type is not supported by ${declaration.model}: ${block.type}`);
 		counts.set(block.type, (counts.get(block.type) ?? 0) + 1);
 		if ("source" in block && spec.sources && !spec.sources.includes(block.source.type)) throw new GenerationValidationError(`${block.type} source is not supported by ${declaration.model}: ${block.source.type}`);
+		const role = getRole(block);
+		if (spec.roleRequired && (typeof role !== "string" || role.length === 0)) throw new GenerationValidationError(`${block.type} role is required by ${declaration.model}`);
+		if (role !== void 0 && spec.roles) {
+			if (typeof role !== "string" || role.length === 0 || !spec.roles.includes(role)) throw new GenerationValidationError(`${block.type} role is not supported by ${declaration.model}: ${String(role)}`);
+		}
 	}
 	for (const spec of declaration.content.input) {
 		const count = counts.get(spec.type) ?? 0;
@@ -245,28 +256,12 @@ function mergeTextBlocks(declaration, content) {
 	const separator = textSpec?.merge === "space" ? " " : textSpec?.merge === "concat" ? "" : "\n";
 	return content.filter((block) => block.type === "text").map((block) => block.text).join(separator).trim();
 }
+
 //#endregion
 //#region src/adapters/ark-video-generations.ts
 const REQUEST_TIMEOUT_MS$5 = 186e4;
 const DEFAULT_POLL_INTERVAL_SEC$2 = 2;
 const DEFAULT_MAX_WAIT_SEC$2 = 600;
-const RESOLUTION_SHORT_EDGE = {
-	"480p": 480,
-	"720p": 720,
-	"1080p": 1080,
-	"2K": 1440
-};
-const ASPECT_RATIOS = {
-	"16:9": [16, 9],
-	"9:16": [9, 16],
-	"1:1": [1, 1],
-	"4:3": [4, 3],
-	"3:2": [3, 2],
-	"2:3": [2, 3],
-	"3:4": [3, 4],
-	"21:9": [21, 9],
-	adaptive: null
-};
 function sleep$2(ms) {
 	return new Promise((resolve) => setTimeout(resolve, ms));
 }
@@ -287,52 +282,73 @@ function getIntegerParameter(parameters, key, fallback) {
 	const value = parameters[key];
 	return typeof value === "number" && Number.isInteger(value) ? value : fallback;
 }
-function resolveSize(resolution, aspectRatio) {
-	const ratio = ASPECT_RATIOS[aspectRatio];
-	if (!ratio) return null;
-	const shortEdge = RESOLUTION_SHORT_EDGE[resolution];
-	if (!shortEdge) throw new GenerationValidationError(`Unsupported resolution: ${resolution}`);
-	const [wRatio, hRatio] = ratio;
-	const width = wRatio >= hRatio ? Math.round(shortEdge * wRatio / hRatio) : shortEdge;
-	const height = wRatio >= hRatio ? shortEdge : Math.round(shortEdge * hRatio / wRatio);
-	return {
-		width: width % 2 === 0 ? width : width + 1,
-		height: height % 2 === 0 ? height : height + 1
-	};
-}
-function getImageRole$1(block) {
+function getMediaRole(block) {
 	const role = getBlockMeta(block)?.role;
 	return typeof role === "string" && role ? role : void 0;
 }
-function classifyImages(images) {
-	if (images.length === 0) return null;
-	const hasFirstOrLast = images.some((image) => image.role === "first_frame" || image.role === "last_frame");
-	const hasReference = images.some((image) => image.role === "reference_image");
+function isFrameImage(item) {
+	return item.kind === "image" && (item.role === "first_frame" || item.role === "last_frame");
+}
+function isReferenceMedia(item) {
+	return item.kind === "image" && item.role === "reference_image" || item.kind === "video" && item.role === "reference_video";
+}
+function assertSingleRole(media, role, message) {
+	if (media.filter((item) => item.role === role).length > 1) throw new GenerationValidationError(message);
+}
+function classifyMedia(media) {
+	if (media.length === 0) return null;
+	for (const item of media) {
+		if (item.kind === "image" && item.role && item.role !== "first_frame" && item.role !== "last_frame" && item.role !== "reference_image") throw new GenerationValidationError("Image input must use meta.role first_frame, last_frame, or reference_image");
+		if (item.kind === "video" && item.role !== "reference_video") throw new GenerationValidationError("Video input must use meta.role reference_video");
+	}
+	const hasFrame = media.some(isFrameImage);
+	const hasReference = media.some(isReferenceMedia);
+	const hasPlain = media.some((item) => item.kind === "image" && !item.role);
 	if ([
-		images.some((image) => !image.role),
-		hasFirstOrLast,
+		hasPlain,
+		hasFrame,
 		hasReference
-	].filter(Boolean).length > 1) throw new GenerationValidationError("Cannot mix video image modes: use only plain image, first_frame/last_frame, or reference_image");
-	if (hasReference) return "reference";
-	if (hasFirstOrLast) return "frame";
-	return "image";
-}
-function buildMetadataContent(prompt, images, mode) {
-	const content = [{
-		type: "text",
-		text: prompt
-	}];
-	for (const image of images) {
-		if (mode === "frame" && image.role !== "first_frame" && image.role !== "last_frame") throw new GenerationValidationError("Frame mode images must use meta.role first_frame or last_frame");
-		if (mode === "reference" && image.role !== "reference_image") throw new GenerationValidationError("Reference mode images must use meta.role reference_image");
-		content.push({
+	].filter(Boolean).length > 1) throw new GenerationValidationError("Cannot mix video media modes: use only plain image, first_frame/last_frame, or reference_image/reference_video");
+	if (hasReference) {
+		assertSingleRole(media, "reference_video", "Reference mode supports at most one reference_video");
+		return "reference";
+	}
+	if (hasFrame) {
+		assertSingleRole(media, "first_frame", "Frame mode supports at most one first_frame image");
+		assertSingleRole(media, "last_frame", "Frame mode supports at most one last_frame image");
+		return "frame";
+	}
+	if (hasPlain) {
+		if (media.filter((item) => item.kind === "image" && !item.role).length > 1) throw new GenerationValidationError("Plain image mode supports at most one image");
+		return "image";
+	}
+	return null;
+}
+function buildMetadataContent(media, mode) {
+	const content = [];
+	for (const item of media) {
+		if (mode === "frame" && (item.kind !== "image" || item.role !== "first_frame" && item.role !== "last_frame")) throw new GenerationValidationError("Frame mode images must use meta.role first_frame or last_frame");
+		if (mode === "reference" && !(item.kind === "image" && item.role === "reference_image" || item.kind === "video" && item.role === "reference_video")) throw new GenerationValidationError("Reference mode media must use meta.role reference_image or reference_video");
+		if (item.kind === "image") content.push({
 			type: "image_url",
-			image_url: { url: image.url },
-			role: image.role
+			image_url: { url: item.url },
+			role: item.role
+		});
+		else content.push({
+			type: "video_url",
+			video_url: { url: item.url },
+			role: item.role
 		});
 	}
 	return content;
 }
+async function resolveMedia(input, media) {
+	return Promise.all(media.map(async (item) => ({
+		kind: item.kind,
+		role: item.role,
+		url: await input.context.resolveSource(item.source)
+	})));
+}
 function extractTaskId$2(response) {
 	const taskId = asString$2(response.task_id) ?? asString$2(response.id);
 	if (!taskId) throw new GenerationProviderError("Video generation provider did not return a task id", { details: { response } });
@@ -392,14 +408,15 @@ async function requestJson$2(input, path, init) {
 async function arkVideoGenerationsAdapter(input) {
 	const prompt = mergeTextBlocks(input.declaration, input.request.content);
 	if (!prompt) throw new GenerationValidationError("Prompt text is required");
-	const imageBlocks = input.request.content.filter((block) => block.type === "image");
-	const images = await Promise.all(imageBlocks.map(async (block) => ({
-		url: await input.context.resolveSource(block.source),
-		role: getImageRole$1(block)
-	})));
-	const mode = classifyImages(images);
+	const inputMedia = input.request.content.filter((block) => block.type === "image" || block.type === "video").map((block) => ({
+		kind: block.type,
+		source: block.source,
+		role: getMediaRole(block)
+	}));
+	const mode = classifyMedia(inputMedia);
+	const media = await resolveMedia(input, inputMedia);
 	const resolution = asString$2(input.parameters.resolution) ?? "720p";
-	const aspectRatio = asString$2(input.parameters.aspect_ratio) ?? "16:9";
+	const ratio = asString$2(input.request.parameters?.ratio) ?? asString$2(input.request.parameters?.aspect_ratio) ?? asString$2(input.parameters.ratio) ?? "16:9";
 	const duration = getIntegerParameter(input.parameters, "duration", 5);
 	const fps = getIntegerParameter(input.parameters, "fps", 30);
 	const pollIntervalSec = getIntegerParameter(input.parameters, "poll_interval", DEFAULT_POLL_INTERVAL_SEC$2);
@@ -416,23 +433,18 @@ async function arkVideoGenerationsAdapter(input) {
 	const metadata = {
 		duration,
 		fps,
-		generate_audio: generateAudio
+		generate_audio: generateAudio,
+		resolution,
+		ratio
 	};
 	if (seed !== void 0) metadata.seed = seed;
 	if (returnLastFrame) metadata.return_last_frame = true;
 	if (cameraFixed) metadata.camera_fixed = true;
 	if (watermark) metadata.watermark = true;
-	if (mode === "frame" || mode === "reference") {
-		metadata.content = buildMetadataContent(prompt, images, mode);
-		metadata.resolution = resolution;
-		metadata.ratio = aspectRatio;
-	} else {
-		const size = resolveSize(resolution, aspectRatio);
-		if (size) {
-			payload.width = size.width;
-			payload.height = size.height;
-		}
-		if (images[0]) payload.image = images[0].url;
+	if (mode === "frame" || mode === "reference") metadata.content = buildMetadataContent(media, mode);
+	else {
+		const firstImage = media.find((item) => item.kind === "image");
+		if (firstImage) payload.image = firstImage.url;
 	}
 	payload.metadata = metadata;
 	const taskId = extractTaskId$2(await requestJson$2(input, "/v1/video/generations", {
@@ -486,6 +498,7 @@ async function arkVideoGenerationsAdapter(input) {
 	}
 	throw new GenerationTimeoutError("Timed out waiting for video generation", { taskId });
 }
+
 //#endregion
 //#region src/adapters/gemini-generate-content.ts
 const REQUEST_TIMEOUT_MS$4 = 3e5;
@@ -590,7 +603,7 @@ async function geminiGenerateContentAdapter(input) {
 		const image = {};
 		if (typeof aspectRatio === "string") image.aspectRatio = aspectRatio;
 		if (typeof imageSize === "string") image.imageSize = imageSize;
-		generationConfig.responseFormat = { image };
+		generationConfig.imageConfig = image;
 	}
 	const payload = {
 		contents: [{ parts: [{ text: prompt }, ...imageParts] }],
@@ -617,6 +630,7 @@ async function geminiGenerateContentAdapter(input) {
 	if (output.length === 0) throw new GenerationProviderError("Gemini generation returned no output", { details: collectGeminiNoOutputDetails(raw) });
 	return output;
 }
+
 //#endregion
 //#region src/adapters/kling-video-generations.ts
 const REQUEST_TIMEOUT_MS$3 = 186e4;
@@ -855,6 +869,7 @@ async function klingVideoGenerationsAdapter(input) {
 	}
 	throw new GenerationTimeoutError("Timed out waiting for Kling video generation", { taskId });
 }
+
 //#endregion
 //#region src/adapters/openai-image-edits.ts
 const REQUEST_TIMEOUT_MS$2 = 3e5;
@@ -924,6 +939,7 @@ async function openAiImageEditsAdapter(input) {
 	if (output.length === 0) throw new GenerationProviderError("Image edit returned no output", { details: collectOpenAiImageEditsNoOutputDetails(raw) });
 	return output;
 }
+
 //#endregion
 //#region src/adapters/openai-images.ts
 const REQUEST_TIMEOUT_MS$1 = 3e5;
@@ -1000,6 +1016,7 @@ async function openAiImagesAdapter(input) {
 	if (output.length === 0) throw new GenerationProviderError("Image generation returned no output", { details: collectOpenAiImagesNoOutputDetails(raw) });
 	return output;
 }
+
 //#endregion
 //#region src/adapters/suno-tasks.ts
 const REQUEST_TIMEOUT_MS = 6e4;
@@ -1034,12 +1051,12 @@ const OPERATION_PATHS = {
 		poll: true
 	}
 };
-const FINAL_SUCCESS_STATUSES = /* @__PURE__ */ new Set([
+const FINAL_SUCCESS_STATUSES = new Set([
 	"success",
 	"succeeded",
 	"completed"
 ]);
-const FINAL_FAILURE_STATUSES = /* @__PURE__ */ new Set([
+const FINAL_FAILURE_STATUSES = new Set([
 	"failure",
 	"failed",
 	"error",
@@ -1358,6 +1375,7 @@ async function sunoTasksAdapter(input) {
 	if (!taskId) return buildImmediateResult(operation, data, raw);
 	return pollSunoTask(input, operation, taskId, asInteger(input.parameters.poll_interval, DEFAULT_POLL_INTERVAL_SEC), asInteger(input.parameters.max_wait, DEFAULT_MAX_WAIT_SEC));
 }
+
 //#endregion
 //#region src/adapters/index.ts
 const builtinGenerationAdapters = {
@@ -1373,9 +1391,10 @@ function getGenerationAdapter(type, adapters = {}) {
 	if (!adapter) throw new GenerationUnsupportedAdapterError(type);
 	return adapter;
 }
+
 //#endregion
 //#region src/config.ts
-const DECLARATION_EXTENSIONS = /* @__PURE__ */ new Set([
+const DECLARATION_EXTENSIONS = new Set([
 	".yaml",
 	".yml",
 	".json"
@@ -1411,7 +1430,7 @@ function isGenerationModelDeclaration(value) {
 	const parameters = value.parameters;
 	const meta = value.meta;
 	const examples = value.examples;
-	return value.schema === "neta.generation.model.v1" && typeof value.model === "string" && value.model.trim().length > 0 && (value.allowUnknownParameters === void 0 || typeof value.allowUnknownParameters === "boolean") && isRecord(adapter) && typeof adapter.type === "string" && isRecord(content) && Array.isArray(content.input) && (parameters === void 0 || isRecord(parameters) && Object.values(parameters).every(isParameterSpec)) && (meta === void 0 || isMetaSpec(meta)) && (examples === void 0 || Array.isArray(examples));
+	return value.schema === MODEL_SCHEMA && typeof value.model === "string" && value.model.trim().length > 0 && (value.allowUnknownParameters === void 0 || typeof value.allowUnknownParameters === "boolean") && isRecord(adapter) && typeof adapter.type === "string" && isRecord(content) && Array.isArray(content.input) && (parameters === void 0 || isRecord(parameters) && Object.values(parameters).every(isParameterSpec)) && (meta === void 0 || isMetaSpec(meta)) && (examples === void 0 || Array.isArray(examples));
 }
 function parseGenerationModelDeclaration(rawText, filePath = "model.yaml") {
 	const parsed = extname(filePath) === ".json" ? JSON.parse(rawText) : parse(rawText);
@@ -1445,6 +1464,7 @@ async function writeGenerationModelDeclarations(declarations, directory, options
 	const ext = options.format === "json" ? "json" : "yaml";
 	await Promise.all(declarations.map((declaration) => writeGenerationModelDeclaration(declaration, join(directory, `${slugifyFileName(declaration.model)}.${ext}`), options)));
 }
+
 //#endregion
 //#region src/source.ts
 const defaultGenerationSourceResolver = (source) => {
@@ -1453,13 +1473,14 @@ const defaultGenerationSourceResolver = (source) => {
 		case "base64": return `data:${source.mediaType};base64,${source.data}`;
 	}
 };
+
 //#endregion
 //#region src/client.ts
 const DEFAULT_BASE_URL = "https://router.neta.art";
 const REDACTED = "[REDACTED]";
 const SECRET_DEBUG_KEY_PATTERN = /^(authorization|api[-_]?key|token|thoughtSignature)$/i;
 const BASE64_DEBUG_KEY_PATTERN = /^(b64_json|data)$/i;
-const MEDIA_PAYLOAD_KEYS = /* @__PURE__ */ new Set([
+const MEDIA_PAYLOAD_KEYS = new Set([
 	"audio",
 	"audio_url",
 	"image",
@@ -1604,6 +1625,7 @@ async function createGenerationClientFromDirectory(directory, options = {}) {
 async function createGenerationClientFromFile(filePath, options = {}) {
 	return createGenerationClientFromFiles([filePath], options);
 }
+
 //#endregion
 //#region src/export-config.ts
 function stringifyBuiltinModelConfig(model, options = {}) {
@@ -1619,7 +1641,7 @@ async function exportBuiltinModelConfig(model, filePath) {
 async function exportBuiltinModelConfigs(directory) {
 	await writeGenerationModelDeclarations(listBuiltinGenerationModels(), directory);
 }
+
 //#endregion
 export { validateGenerationContent as A, klingVideoGenerationsAdapter as C, mergeTextBlocks as D, mergeGenerationMeta as E, GenerationUnsupportedAdapterError as F, GenerationValidationError as I, GenerationError as M, GenerationProviderError as N, resolveGenerationMeta as O, GenerationTimeoutError as P, openAiImageEditsAdapter as S, arkVideoGenerationsAdapter as T, writeGenerationModelDeclarations as _, createGenerationClientFromDirectory as a, sunoTasksAdapter as b, defaultGenerationSourceResolver as c, parseGenerationModelDeclaration as d, readGenerationModelDeclaration as f, writeGenerationModelDeclaration as g, stringifyGenerationModelDeclaration as h, createGenerationClient as i, GenerationConfigError as j, resolveGenerationParameters as k, isGenerationModelDeclaration as l, readGenerationModelDeclarationsFromFiles as m, exportBuiltinModelConfigs as n, createGenerationClientFromFile as o, readGenerationModelDeclarationsFromDirectory as p, stringifyBuiltinModelConfig as r, createGenerationClientFromFiles as s, exportBuiltinModelConfig as t, mergeGenerationModelDeclarations as u, builtinGenerationAdapters as v, geminiGenerateContentAdapter as w, openAiImagesAdapter as x, getGenerationAdapter as y };
-
-//# sourceMappingURL=export-config-BuQ5Maoj.js.map
+//# sourceMappingURL=export-config-DolxdQdi.js.map