@neta-art/generation 0.1.17 → 0.1.18

package/dist/{export-config-CIZi8HJ8.js → export-config-DJdXXiMl.js}

@@ -1,4 +1,4 @@
-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-BqqqQ7Ib.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-B4Agk9nt.js";
 import { mkdir, readFile, readdir, writeFile } from "node:fs/promises";
 import { extname, join } from "node:path";
 import { parse, stringify } from "yaml";
@@ -270,7 +270,7 @@ function mergeTextBlocks(declaration, content) {
 
 //#endregion
 //#region src/adapters/ark-video-generations.ts
-const REQUEST_TIMEOUT_MS$5 = 186e4;
+const REQUEST_TIMEOUT_MS$6 = 186e4;
 const DEFAULT_POLL_INTERVAL_SEC$2 = 2;
 const DEFAULT_MAX_WAIT_SEC$2 = 600;
 function sleep$2(ms) {
@@ -406,7 +406,7 @@ async function requestJson$2(input, path, init) {
 			"Content-Type": "application/json",
 			...init.headers
 		}
-	}, REQUEST_TIMEOUT_MS$5);
+	}, REQUEST_TIMEOUT_MS$6);
 	if (!response.ok) {
 		const body = await response.text().catch(() => response.statusText);
 		throw new GenerationProviderError("Video generation provider request failed", {
@@ -510,6 +510,202 @@ async function arkVideoGenerationsAdapter(input) {
 	throw new GenerationTimeoutError("Timed out waiting for video generation", { taskId });
 }
 
+//#endregion
+//#region src/adapters/audio-speech.ts
+const REQUEST_TIMEOUT_MS$5 = 21e4;
+const QWEN_MODELS = new Set([
+	"qwen-tts",
+	"qwen-audio-3.0-tts-plus",
+	"qwen-audio-3.0-tts-flash"
+]);
+const QWEN_AUDIO_3_MODELS = new Set(["qwen-audio-3.0-tts-plus", "qwen-audio-3.0-tts-flash"]);
+const HIGGS_MODEL = "higgs-tts";
+function isRecord$4(value) {
+	return !!value && typeof value === "object" && !Array.isArray(value);
+}
+function textBlocks(input) {
+	return input.request.content.filter((block) => block.type === "text");
+}
+function audioBlocks(input) {
+	return input.request.content.filter((block) => block.type === "audio");
+}
+function validateMetaKeys(label, meta, allowed) {
+	for (const key of Object.keys(meta ?? {})) if (!allowed.has(key)) throw new GenerationValidationError(`Unknown ${label} field: ${key}`);
+}
+function validateHttpUrl(value, label) {
+	if (typeof value !== "string") throw new GenerationValidationError(`${label} must be a valid HTTP(S) URL`);
+	const trimmed = value.trim();
+	if (!trimmed) throw new GenerationValidationError(`${label} must be a non-empty HTTP(S) URL`);
+	try {
+		const url = new URL(trimmed);
+		if (url.protocol !== "http:" && url.protocol !== "https:") throw new Error("unsupported protocol");
+	} catch {
+		throw new GenerationValidationError(`${label} must be a valid HTTP(S) URL`);
+	}
+}
+function validateCommonContent(input) {
+	const texts = textBlocks(input);
+	if (texts.length !== 1 || typeof texts[0]?.text !== "string" || !texts[0].text.trim()) throw new GenerationValidationError(`${input.declaration.model} requires exactly one non-empty text content block as input`);
+	const audio = audioBlocks(input);
+	for (const [index, block] of audio.entries()) {
+		if (block.source.type !== "url") throw new GenerationValidationError(`${input.declaration.model} reference audio must use a URL source`);
+		validateHttpUrl(block.source.url, `audio content block ${index} URL`);
+	}
+	return {
+		text: texts[0],
+		audio
+	};
+}
+function validateQwen(input, text, audio) {
+	if (audio.length > 1) throw new GenerationValidationError(`${input.declaration.model} supports at most one reference audio`);
+	const requestMetaKeys = new Set(["voice_prompt", "preview_text"]);
+	validateMetaKeys("request.metadata", input.request.metadata, requestMetaKeys);
+	validateMetaKeys("request.meta", input.request.meta, requestMetaKeys);
+	validateMetaKeys("text content meta", text.meta, /* @__PURE__ */ new Set());
+	for (const block of audio) validateMetaKeys("audio content meta", block.meta, /* @__PURE__ */ new Set());
+	const voicePrompt = input.meta.voice_prompt;
+	const hasVoicePrompt = typeof voicePrompt === "string" && voicePrompt.trim().length > 0;
+	if (voicePrompt !== void 0 && !hasVoicePrompt) throw new GenerationValidationError(`${input.declaration.model} meta.voice_prompt must be a non-empty string`);
+	if (audio.length === 0 && !hasVoicePrompt) throw new GenerationValidationError(`${input.declaration.model} requires one reference audio or meta.voice_prompt`);
+	if (audio.length > 0 && hasVoicePrompt) throw new GenerationValidationError(`${input.declaration.model} reference audio and meta.voice_prompt are mutually exclusive`);
+	if (QWEN_AUDIO_3_MODELS.has(input.declaration.model) && Array.from(text.text.trim()).length < 15) throw new GenerationValidationError(`${input.declaration.model} requires input of at least 15 Unicode code points`);
+}
+function hasOwnWeight(block) {
+	return Object.hasOwn(block.meta ?? {}, "weight");
+}
+function isPositiveFiniteNumber(value) {
+	return typeof value === "number" && Number.isFinite(value) && value > 0;
+}
+function validateHiggs(input, text, audio) {
+	if (audio.length > 16) throw new GenerationValidationError(`${input.declaration.model} supports at most 16 references`);
+	validateMetaKeys("request.metadata", input.request.metadata, /* @__PURE__ */ new Set());
+	validateMetaKeys("request.meta", input.request.meta, /* @__PURE__ */ new Set());
+	validateMetaKeys("text content meta", text.meta, /* @__PURE__ */ new Set());
+	for (const [index, block] of audio.entries()) {
+		validateMetaKeys(`audio content block ${index} meta`, block.meta, new Set(["weight"]));
+		const weight = block.meta?.weight;
+		if (audio.length === 1 && (!hasOwnWeight(block) || weight === void 0)) continue;
+		if (!isPositiveFiniteNumber(weight)) throw new GenerationValidationError(`audio content block ${index} meta.weight must be a finite positive number`);
+	}
+}
+function validateAudioSpeechRequest(input) {
+	const { text, audio } = validateCommonContent(input);
+	if (QWEN_MODELS.has(input.declaration.model)) {
+		validateQwen(input, text, audio);
+		return;
+	}
+	if (input.declaration.model === HIGGS_MODEL) {
+		validateHiggs(input, text, audio);
+		return;
+	}
+	throw new GenerationValidationError(`Unsupported audio speech model: ${input.declaration.model}`);
+}
+function requestIdFromHeaders(headers) {
+	return headers.get("x-request-id")?.trim() || headers.get("x-oneapi-request-id")?.trim() || void 0;
+}
+function buildPayload$2(input) {
+	const text = textBlocks(input)[0];
+	if (!text) throw new GenerationValidationError("Audio speech input was not validated");
+	const audio = audioBlocks(input);
+	const payload = {
+		model: input.declaration.model,
+		input: text.text
+	};
+	if (QWEN_MODELS.has(input.declaration.model)) {
+		if (audio[0]?.source.type === "url") payload.ref_audio = audio[0].source.url.trim();
+		else payload.metadata = { voice_prompt: input.meta.voice_prompt };
+		return payload;
+	}
+	const firstAudio = audio[0];
+	if (audio.length === 1 && firstAudio && (!hasOwnWeight(firstAudio) || firstAudio.meta?.weight === void 0)) {
+		if (firstAudio.source.type === "url") payload.ref_audio = firstAudio.source.url.trim();
+		return payload;
+	}
+	if (audio.length > 0) payload.metadata = { references: audio.map((block) => ({
+		url: block.source.type === "url" ? block.source.url.trim() : "",
+		weight: block.meta?.weight
+	})) };
+	return payload;
+}
+function parseJson(rawBody) {
+	try {
+		return JSON.parse(rawBody);
+	} catch {
+		return;
+	}
+}
+function nonEmptyString(value) {
+	return typeof value === "string" && value.trim() ? value : void 0;
+}
+function providerError(response, rawBody) {
+	const requestId = requestIdFromHeaders(response.headers);
+	const contentType = (response.headers.get("content-type") ?? "").toLowerCase();
+	const parsed = contentType.includes("application/json") || contentType.includes("+json") ? parseJson(rawBody) : void 0;
+	const parsedRecord = isRecord$4(parsed) ? parsed : void 0;
+	const errorRecord = isRecord$4(parsedRecord?.error) ? parsedRecord.error : void 0;
+	const message = nonEmptyString(errorRecord?.message) ?? (rawBody.trim() ? rawBody : void 0) ?? (response.statusText.trim() ? response.statusText : void 0) ?? "Audio speech provider request failed";
+	const code = nonEmptyString(errorRecord?.code);
+	return new GenerationProviderError(message, {
+		status: response.status,
+		body: rawBody,
+		details: {
+			requestId,
+			code,
+			...parsedRecord ? { error: parsedRecord } : {}
+		}
+	});
+}
+function unsupportedSuccessResponse(response, rawBody) {
+	const statusText = response.statusText.trim();
+	return new GenerationProviderError(response.status === 204 && statusText ? statusText : "Audio speech provider returned an unsupported success response", {
+		status: response.status,
+		body: rawBody,
+		details: { requestId: requestIdFromHeaders(response.headers) }
+	});
+}
+function parseSuccessResponse(response, rawBody) {
+	const parsed = parseJson(rawBody);
+	if (!isRecord$4(parsed)) throw unsupportedSuccessResponse(response, rawBody);
+	const urlValue = nonEmptyString(parsed.url);
+	const contentType = nonEmptyString(parsed.content_type);
+	if (!urlValue || parsed.media_type !== "audio" || !contentType || !/^audio\/[^\s/;]+(?:\s*;.*)?$/i.test(contentType)) throw unsupportedSuccessResponse(response, rawBody);
+	try {
+		const url = new URL(urlValue.trim());
+		if (url.protocol !== "http:" && url.protocol !== "https:") throw new Error("unsupported protocol");
+	} catch {
+		throw unsupportedSuccessResponse(response, rawBody);
+	}
+	const size = parsed.size;
+	if (size !== void 0 && (typeof size !== "number" || !Number.isFinite(size) || !Number.isInteger(size) || size < 0)) throw unsupportedSuccessResponse(response, rawBody);
+	const requestId = requestIdFromHeaders(response.headers);
+	return {
+		type: "audio",
+		source: {
+			type: "url",
+			url: urlValue.trim()
+		},
+		meta: {
+			content_type: contentType,
+			...size !== void 0 ? { size } : {},
+			...requestId ? { request_id: requestId } : {}
+		}
+	};
+}
+async function generateAudioSpeech(input) {
+	const response = await fetchWithTimeout(input.context.fetch, joinUrl(input.context.baseUrl, "/v1/audio/speech"), {
+		method: "POST",
+		headers: {
+			Authorization: `Bearer ${input.context.apiKey}`,
+			"Content-Type": "application/json"
+		},
+		body: JSON.stringify(buildPayload$2(input))
+	}, REQUEST_TIMEOUT_MS$5);
+	const rawBody = await response.text().catch(() => "");
+	if (!response.ok) throw providerError(response, rawBody);
+	return [parseSuccessResponse(response, rawBody)];
+}
+const audioSpeechAdapter = Object.assign(generateAudioSpeech, { validate: validateAudioSpeechRequest });
+
 //#endregion
 //#region src/adapters/gemini-generate-content.ts
 const REQUEST_TIMEOUT_MS$4 = 3e5;
@@ -1391,14 +1587,18 @@ async function sunoTasksAdapter(input) {
 //#region src/adapters/index.ts
 const builtinGenerationAdapters = {
 	"ark.videoGenerations": arkVideoGenerationsAdapter,
+	"openai.audioSpeech": audioSpeechAdapter,
 	"gemini.generateContent": geminiGenerateContentAdapter,
 	"kling.videoGenerations": klingVideoGenerationsAdapter,
 	"openai.imageEdits": openAiImageEditsAdapter,
 	"openai.images": openAiImagesAdapter,
 	"suno.tasks": sunoTasksAdapter
 };
+function tryGetGenerationAdapter(type, adapters = {}) {
+	return adapters[type] ?? builtinGenerationAdapters[type];
+}
 function getGenerationAdapter(type, adapters = {}) {
-	const adapter = adapters[type] ?? builtinGenerationAdapters[type];
+	const adapter = tryGetGenerationAdapter(type, adapters);
 	if (!adapter) throw new GenerationUnsupportedAdapterError(type);
 	return adapter;
 }
@@ -1499,6 +1699,14 @@ function stringValue(value) {
 function isRecord(value) {
 	return !!value && typeof value === "object" && !Array.isArray(value);
 }
+function extractGenerationResultHeaderFields(headers) {
+	const requestId = headers.get("x-request-id")?.trim() || void 0;
+	const oneApiRequestId = headers.get("x-oneapi-request-id")?.trim() || void 0;
+	return requestId || oneApiRequestId ? {
+		...requestId ? { requestId } : {},
+		...oneApiRequestId ? { oneApiRequestId } : {}
+	} : void 0;
+}
 function extractGenerationResultFields(raw) {
 	if (!isRecord(raw)) return void 0;
 	const usage = isRecord(raw.usage) ? raw.usage : void 0;
@@ -1594,11 +1802,13 @@ function resolveModels(options) {
 function createResultFieldCaptureFetch(fetchFn, onFields) {
 	return (async (input, init) => {
 		const response = await fetchFn(input, init);
+		const headerFields = extractGenerationResultHeaderFields(response.headers);
 		const contentType = response.headers.get("content-type") ?? "";
+		let fields;
 		if (contentType.includes("application/json") || contentType.includes("+json")) try {
-			const fields = extractGenerationResultFields(await response.clone().json());
-			if (fields) onFields(fields);
+			fields = extractGenerationResultFields(await response.clone().json());
 		} catch {}
+		if (fields || headerFields) onFields(fields, headerFields);
 		return response;
 	});
 }
@@ -1622,12 +1832,17 @@ function createGenerationClient(options = {}) {
 			...request.metadata ?? {},
 			...request.meta ?? {}
 		}, request.content), request.content);
-		return {
+		const resolved = {
 			declaration: cloneJson(declaration),
-			request: cloneJson(request),
+			request,
 			parameters,
 			meta
 		};
+		tryGetGenerationAdapter(declaration.adapter.type, options.adapters)?.validate?.(resolved);
+		return {
+			...resolved,
+			request: cloneJson(request)
+		};
 	}
 	async function runAdapter(request, fetch) {
 		const resolved = validateRequest(request);
@@ -1652,12 +1867,23 @@ function createGenerationClient(options = {}) {
 		},
 		async generateResult(request) {
 			let capturedFields;
+			let capturedBodyRequestId;
+			let capturedRequestId;
+			let capturedOneApiRequestId;
 			return {
-				content: await runAdapter(request, createResultFieldCaptureFetch(adapterFetch, (fields) => {
-					capturedFields = {
+				content: await runAdapter(request, createResultFieldCaptureFetch(adapterFetch, (fields, headerFields) => {
+					if (fields) capturedFields = {
 						...capturedFields,
 						...fields
 					};
+					if (fields?.requestId) capturedBodyRequestId = fields.requestId;
+					if (headerFields?.requestId) capturedRequestId = headerFields.requestId;
+					if (headerFields?.oneApiRequestId) capturedOneApiRequestId = headerFields.oneApiRequestId;
+					const requestId = capturedBodyRequestId ?? capturedRequestId ?? capturedOneApiRequestId;
+					if (requestId) capturedFields = {
+						...capturedFields,
+						requestId
+					};
 				})),
 				...capturedFields ?? {}
 			};
@@ -1717,5 +1943,5 @@ async function exportBuiltinModelConfigs(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-CIZi8HJ8.js.map
+export { resolveGenerationParameters as A, klingVideoGenerationsAdapter as C, mergeGenerationMeta as D, arkVideoGenerationsAdapter as E, GenerationTimeoutError as F, GenerationUnsupportedAdapterError as I, GenerationValidationError as L, GenerationConfigError as M, GenerationError as N, mergeTextBlocks as O, GenerationProviderError as P, openAiImageEditsAdapter as S, audioSpeechAdapter 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, validateGenerationContent as j, resolveGenerationMeta 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-DJdXXiMl.js.map