@agentor/dashscope 0.0.0 → 0.0.2

package/dist/index.mjs

@@ -1,23 +1,287 @@
+import { OpenAICompatibleEmbeddingModel } from "@ai-sdk/openai-compatible";
+import { combineHeaders, convertToBase64, createEventSourceResponseHandler, createJsonErrorResponseHandler, createJsonResponseHandler, createProviderToolFactoryWithOutputSchema, delay, generateId, getFromApi, isParsableJson, lazySchema, parseProviderOptions, postJsonToApi, zodSchema } from "@ai-sdk/provider-utils";
 import { z } from "zod/v4";
-import { combineHeaders, convertToBase64, createEventSourceResponseHandler, createJsonErrorResponseHandler, createJsonResponseHandler, createProviderToolFactoryWithOutputSchema, generateId, isParsableJson, lazySchema, parseProviderOptions, postJsonToApi, zodSchema } from "@ai-sdk/provider-utils";
-import { mapOpenAICompatibleFinishReason, prepareTools } from "@ai-sdk/openai-compatible/internal";
-//#region src/types.ts
-const DASHSCOPE_REGION_BASE_URLS = {
-	beijing: {
-		baseURL: "https://dashscope.aliyuncs.com/compatible-mode/v1",
-		videoBaseURL: "https://dashscope.aliyuncs.com"
-	},
-	singapore: {
-		baseURL: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
-		videoBaseURL: "https://dashscope-intl.aliyuncs.com"
-	},
-	us: {
-		baseURL: "https://dashscope-us.aliyuncs.com/compatible-mode/v1",
-		videoBaseURL: "https://dashscope-us.aliyuncs.com"
-	},
-	germany: {
-		baseURL: "https://{workspaceId}.eu-central-1.maas.aliyuncs.com/compatible-mode/v1",
-		videoBaseURL: "https://{workspaceId}.eu-central-1.maas.aliyuncs.com"
+import { AISDKError } from "@ai-sdk/provider";
+import { convertOpenAICompatibleChatUsage, getResponseMetadata, mapOpenAICompatibleFinishReason, prepareTools } from "@ai-sdk/openai-compatible/internal";
+//#region src/embedding.ts
+var DashScopeEmbeddingModel = class extends OpenAICompatibleEmbeddingModel {
+	constructor(modelId, config) {
+		super(modelId, {
+			provider: config.provider,
+			url: () => `${config.baseURL}/compatible-mode/v1/embeddings`,
+			headers: config.headers,
+			fetch: config.fetch
+		});
+	}
+};
+const failedResponseHandler = createJsonErrorResponseHandler({
+	errorSchema: zodSchema(z.object({ error: z.object({
+		message: z.string(),
+		code: z.string().nullish(),
+		type: z.string().nullish()
+	}) })),
+	errorToMessage: (data) => data.error.message
+});
+const nativeFailedHandler = createJsonErrorResponseHandler({
+	errorSchema: zodSchema(z.object({
+		code: z.string().nullish(),
+		message: z.string(),
+		request_id: z.string().nullish()
+	})),
+	errorToMessage: (data) => data.message
+});
+function convertResponsesUsage(usage) {
+	if (!usage) return {
+		inputTokens: {
+			total: 0,
+			noCache: void 0,
+			cacheRead: void 0,
+			cacheWrite: void 0
+		},
+		outputTokens: {
+			total: 0,
+			text: void 0,
+			reasoning: void 0
+		}
+	};
+	return {
+		inputTokens: {
+			total: usage.input_tokens ?? 0,
+			noCache: void 0,
+			cacheRead: usage.input_tokens_details?.cached_tokens ?? void 0,
+			cacheWrite: void 0
+		},
+		outputTokens: {
+			total: usage.output_tokens ?? 0,
+			text: void 0,
+			reasoning: usage.output_tokens_details?.reasoning_tokens ?? void 0
+		},
+		raw: usage
+	};
+}
+function uint8ArrayToBase64(data) {
+	let binary = "";
+	for (let i = 0; i < data.length; i++) binary += String.fromCharCode(data[i]);
+	return btoa(binary);
+}
+//#endregion
+//#region src/image.ts
+const imageOptionsSchema = z.object({
+	size: z.string().optional(),
+	negativePrompt: z.string().optional(),
+	promptExtend: z.boolean().optional(),
+	watermark: z.boolean().optional(),
+	n: z.number().optional()
+});
+const imageResponseSchema = zodSchema(z.object({
+	output: z.object({ choices: z.array(z.object({ message: z.object({ content: z.array(z.object({ image: z.string().optional() })) }) })).optional() }).nullish(),
+	usage: z.object({
+		image_count: z.number().optional(),
+		width: z.number().optional(),
+		height: z.number().optional()
+	}).nullish(),
+	request_id: z.string().nullish()
+}));
+var DashScopeImageModel = class {
+	specificationVersion = "v3";
+	modelId;
+	config;
+	constructor(modelId, config) {
+		this.modelId = modelId;
+		this.config = config;
+	}
+	get provider() {
+		return this.config.provider;
+	}
+	get maxImagesPerCall() {
+		return 1;
+	}
+	async doGenerate(options) {
+		const warnings = [];
+		const dsOptions = await parseProviderOptions({
+			provider: "dashscope",
+			providerOptions: options.providerOptions,
+			schema: imageOptionsSchema
+		});
+		const body = {
+			model: this.modelId,
+			input: { messages: [{
+				role: "user",
+				content: [{ text: options.prompt }]
+			}] },
+			parameters: {
+				...dsOptions?.size != null && { size: dsOptions.size },
+				...dsOptions?.negativePrompt != null && { negative_prompt: dsOptions.negativePrompt },
+				...dsOptions?.promptExtend != null && { prompt_extend: dsOptions.promptExtend },
+				...dsOptions?.watermark != null && { watermark: dsOptions.watermark },
+				...dsOptions?.n != null && { n: dsOptions.n }
+			}
+		};
+		const { responseHeaders, value: response } = await postJsonToApi({
+			url: `${this.config.baseURL}/api/v1/services/aigc/multimodal-generation/generation`,
+			headers: combineHeaders(this.config.headers(), options.headers),
+			body,
+			failedResponseHandler: nativeFailedHandler,
+			successfulResponseHandler: createJsonResponseHandler(imageResponseSchema),
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		const imageUrls = response.output?.choices?.flatMap((c) => c.message.content.filter((p) => p.image != null).map((p) => p.image)) ?? [];
+		const images = [];
+		for (const url of imageUrls) {
+			const buffer = await (await (this.config.fetch ?? fetch)(url, { headers: this.config.headers() })).arrayBuffer();
+			images.push(uint8ArrayToBase64(new Uint8Array(buffer)));
+		}
+		return {
+			images,
+			warnings,
+			response: {
+				timestamp: /* @__PURE__ */ new Date(),
+				modelId: this.modelId,
+				headers: responseHeaders
+			}
+		};
+	}
+};
+//#endregion
+//#region src/rerank.ts
+const rerankResponseSchema = zodSchema(z.object({
+	id: z.string().optional(),
+	model: z.string().optional(),
+	results: z.array(z.object({
+		index: z.number(),
+		relevance_score: z.number()
+	})).optional()
+}));
+var DashScopeRerankingModel = class {
+	specificationVersion = "v3";
+	modelId;
+	config;
+	constructor(modelId, config) {
+		this.modelId = modelId;
+		this.config = config;
+	}
+	get provider() {
+		return this.config.provider;
+	}
+	async doRerank(options) {
+		const warnings = [];
+		const documents = options.documents.type === "text" ? options.documents.values : options.documents.values.map((d) => JSON.stringify(d));
+		const body = {
+			model: this.modelId,
+			query: options.query,
+			documents,
+			...options.topN != null && { top_n: options.topN }
+		};
+		const { responseHeaders, value: response } = await postJsonToApi({
+			url: `${this.config.baseURL}/compatible-api/v1/reranks`,
+			headers: combineHeaders(this.config.headers(), options.headers),
+			body,
+			failedResponseHandler,
+			successfulResponseHandler: createJsonResponseHandler(rerankResponseSchema),
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		return {
+			ranking: (response.results ?? []).map((r) => ({
+				index: r.index,
+				relevanceScore: r.relevance_score
+			})),
+			warnings,
+			response: {
+				id: response.id ?? void 0,
+				modelId: response.model ?? void 0,
+				headers: responseHeaders
+			}
+		};
+	}
+};
+//#endregion
+//#region src/speech.ts
+const speechOptionsSchema = z.object({
+	voice: z.string().optional(),
+	format: z.string().optional(),
+	sampleRate: z.number().optional(),
+	languageType: z.string().optional(),
+	speed: z.number().optional(),
+	volume: z.number().optional(),
+	pitch: z.number().optional()
+});
+const cosyvoiceResponseSchema = zodSchema(z.object({
+	output: z.object({ audio: z.object({ url: z.string().optional() }).nullish() }).nullish(),
+	request_id: z.string().nullish()
+}));
+var DashScopeSpeechModel = class {
+	specificationVersion = "v3";
+	modelId;
+	config;
+	constructor(modelId, config) {
+		this.modelId = modelId;
+		this.config = config;
+	}
+	get provider() {
+		return this.config.provider;
+	}
+	async doGenerate(options) {
+		const warnings = [];
+		const dsOptions = await parseProviderOptions({
+			provider: "dashscope",
+			providerOptions: options.providerOptions,
+			schema: speechOptionsSchema
+		});
+		const voice = dsOptions?.voice ?? "longanyang";
+		const format = dsOptions?.format ?? "wav";
+		const sampleRate = dsOptions?.sampleRate ?? 24e3;
+		const isCosyVoice = this.modelId.startsWith("cosyvoice");
+		let url;
+		let body;
+		if (isCosyVoice) {
+			url = `${this.config.baseURL}/api/v1/services/audio/tts/SpeechSynthesizer`;
+			body = {
+				model: this.modelId,
+				input: {
+					text: options.text,
+					voice,
+					format,
+					sample_rate: sampleRate,
+					...dsOptions?.speed != null && { speech_rate: dsOptions.speed },
+					...dsOptions?.volume != null && { volume: dsOptions.volume },
+					...dsOptions?.pitch != null && { pitch_rate: dsOptions.pitch }
+				}
+			};
+		} else {
+			url = `${this.config.baseURL}/api/v1/services/aigc/multimodal-generation/generation`;
+			body = {
+				model: this.modelId,
+				input: {
+					text: options.text,
+					voice,
+					...dsOptions?.languageType != null && { language_type: dsOptions.languageType }
+				}
+			};
+		}
+		const { responseHeaders, value: response } = await postJsonToApi({
+			url,
+			headers: combineHeaders(this.config.headers(), options.headers),
+			body,
+			failedResponseHandler: nativeFailedHandler,
+			successfulResponseHandler: createJsonResponseHandler(cosyvoiceResponseSchema),
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		const audioUrl = response.output?.audio?.url;
+		if (!audioUrl) throw new Error("No audio URL returned from TTS API");
+		const audioBuffer = await (await (this.config.fetch ?? fetch)(audioUrl, { headers: this.config.headers() })).arrayBuffer();
+		return {
+			audio: new Uint8Array(audioBuffer),
+			warnings,
+			request: { body },
+			response: {
+				timestamp: /* @__PURE__ */ new Date(),
+				modelId: this.modelId,
+				headers: responseHeaders
+			}
+		};
 	}
 };
 //#endregion
@@ -95,6 +359,342 @@ const responsesTools = {
 	mcp: (args) => mcpToolFactory(args)
 };
 //#endregion
+//#region src/transcription.ts
+const transcriptionOptionsSchema = z.object({
+	fileUrl: z.string().optional(),
+	languageHints: z.array(z.string()).optional(),
+	enableItn: z.boolean().optional(),
+	enableWords: z.boolean().optional(),
+	channelId: z.array(z.number()).optional(),
+	pollIntervalMs: z.number().positive().optional(),
+	pollTimeoutMs: z.number().positive().optional()
+});
+const syncResponseSchema = zodSchema(z.object({
+	output: z.object({ choices: z.array(z.object({ message: z.object({ content: z.array(z.object({ text: z.string().optional() })) }) })).optional() }).nullish(),
+	request_id: z.string().nullish()
+}));
+const createTaskSchema$1 = zodSchema(z.object({
+	output: z.object({
+		task_id: z.string(),
+		task_status: z.string()
+	}).nullish(),
+	request_id: z.string().nullish()
+}));
+const taskStatusSchema$1 = zodSchema(z.object({
+	output: z.object({
+		task_id: z.string(),
+		task_status: z.string(),
+		result: z.object({ transcription_url: z.string().nullish() }).nullish(),
+		results: z.array(z.object({
+			subtask_status: z.string().nullish(),
+			transcription_url: z.string().nullish()
+		})).nullish(),
+		code: z.string().nullish(),
+		message: z.string().nullish()
+	}).nullish(),
+	request_id: z.string().nullish()
+}));
+function isAsyncModel(modelId) {
+	return modelId.includes("filetrans") || modelId.startsWith("fun-asr") || modelId.startsWith("paraformer");
+}
+function buildAudioUrl(audio, mediaType) {
+	if (typeof audio === "string") {
+		if (audio.startsWith("http")) return audio;
+		return `data:${mediaType};base64,${audio}`;
+	}
+	return `data:${mediaType};base64,${uint8ArrayToBase64(audio)}`;
+}
+var DashScopeTranscriptionModel = class {
+	specificationVersion = "v3";
+	modelId;
+	config;
+	constructor(modelId, config) {
+		this.modelId = modelId;
+		this.config = config;
+	}
+	get provider() {
+		return this.config.provider;
+	}
+	async doGenerate(options) {
+		const warnings = [];
+		const dsOptions = await parseProviderOptions({
+			provider: "dashscope",
+			providerOptions: options.providerOptions,
+			schema: transcriptionOptionsSchema
+		}) ?? null;
+		if (isAsyncModel(this.modelId) && dsOptions?.fileUrl) return this.doAsync(options, dsOptions, warnings);
+		return this.doSync(options, dsOptions, warnings);
+	}
+	async doSync(options, dsOptions, warnings) {
+		const audioUrl = buildAudioUrl(options.audio, options.mediaType);
+		const body = {
+			model: this.modelId,
+			input: { messages: [{
+				role: "user",
+				content: [{ audio: audioUrl }]
+			}] },
+			parameters: {
+				result_format: "message",
+				...dsOptions?.enableItn != null && { asr_options: { enable_itn: dsOptions.enableItn } }
+			}
+		};
+		const { responseHeaders, value: response } = await postJsonToApi({
+			url: `${this.config.baseURL}/api/v1/services/aigc/multimodal-generation/generation`,
+			headers: combineHeaders(this.config.headers(), options.headers),
+			body,
+			failedResponseHandler: nativeFailedHandler,
+			successfulResponseHandler: createJsonResponseHandler(syncResponseSchema),
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		return {
+			text: response.output?.choices?.[0]?.message.content.filter((p) => p.text != null).map((p) => p.text).join("") ?? "",
+			segments: [],
+			language: void 0,
+			durationInSeconds: void 0,
+			warnings,
+			request: { body },
+			response: {
+				timestamp: /* @__PURE__ */ new Date(),
+				modelId: this.modelId,
+				headers: responseHeaders
+			}
+		};
+	}
+	async doAsync(options, dsOptions, warnings) {
+		const audioUrl = dsOptions?.fileUrl;
+		if (!audioUrl) throw new AISDKError({
+			name: "DASHSCOPE_TRANSCRIPTION_ERROR",
+			message: "Async transcription requires providerOptions.dashscope.fileUrl with a publicly accessible audio URL."
+		});
+		const parameters = {};
+		if (dsOptions?.channelId != null) parameters.channel_id = dsOptions.channelId;
+		if (dsOptions?.enableItn != null) parameters.enable_itn = dsOptions.enableItn;
+		if (dsOptions?.enableWords != null) parameters.enable_words = dsOptions.enableWords;
+		if (dsOptions?.languageHints?.length) parameters.language_hints = dsOptions.languageHints;
+		const { value: createResponse } = await postJsonToApi({
+			url: `${this.config.baseURL}/api/v1/services/audio/asr/transcription`,
+			headers: combineHeaders(this.config.headers(), options.headers, { "X-DashScope-Async": "enable" }),
+			body: {
+				model: this.modelId,
+				input: { file_url: audioUrl },
+				...Object.keys(parameters).length > 0 && { parameters }
+			},
+			successfulResponseHandler: createJsonResponseHandler(createTaskSchema$1),
+			failedResponseHandler: nativeFailedHandler,
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		const taskId = createResponse.output?.task_id;
+		if (!taskId) throw new AISDKError({
+			name: "DASHSCOPE_TRANSCRIPTION_ERROR",
+			message: `No task_id returned. Response: ${JSON.stringify(createResponse)}`
+		});
+		const pollInterval = dsOptions?.pollIntervalMs ?? 5e3;
+		const pollTimeout = dsOptions?.pollTimeoutMs ?? 6e5;
+		const startTime = Date.now();
+		while (true) {
+			await delay(pollInterval, { abortSignal: options.abortSignal });
+			if (Date.now() - startTime > pollTimeout) throw new AISDKError({
+				name: "DASHSCOPE_TRANSCRIPTION_TIMEOUT",
+				message: `Transcription timed out after ${pollTimeout}ms`
+			});
+			const { value: status, responseHeaders } = await getFromApi({
+				url: `${this.config.baseURL}/api/v1/tasks/${taskId}`,
+				headers: combineHeaders(this.config.headers(), options.headers, { "X-DashScope-Async": "enable" }),
+				successfulResponseHandler: createJsonResponseHandler(taskStatusSchema$1),
+				failedResponseHandler: nativeFailedHandler,
+				abortSignal: options.abortSignal,
+				fetch: this.config.fetch
+			});
+			const taskStatus = status.output?.task_status;
+			if (taskStatus === "SUCCEEDED") {
+				let transcriptionUrl = status.output?.result?.transcription_url;
+				if (!transcriptionUrl) transcriptionUrl = ((status.output?.results)?.find((r) => r.subtask_status === "SUCCEEDED"))?.transcription_url;
+				if (!transcriptionUrl) throw new AISDKError({
+					name: "DASHSCOPE_TRANSCRIPTION_ERROR",
+					message: `No transcription URL in response. Task ID: ${taskId}`
+				});
+				const resultData = await (await (this.config.fetch ?? fetch)(transcriptionUrl)).json();
+				let text = "";
+				const segments = [];
+				if (resultData.transcripts) for (const transcript of resultData.transcripts) {
+					text += transcript.text;
+					if (transcript.sentences) {
+						for (const sentence of transcript.sentences) if (sentence.begin_time != null && sentence.end_time != null) segments.push({
+							text: sentence.text,
+							startSecond: sentence.begin_time / 1e3,
+							endSecond: sentence.end_time / 1e3
+						});
+					}
+				}
+				return {
+					text,
+					segments,
+					language: void 0,
+					durationInSeconds: void 0,
+					warnings,
+					response: {
+						timestamp: /* @__PURE__ */ new Date(),
+						modelId: this.modelId,
+						headers: responseHeaders
+					}
+				};
+			}
+			if (taskStatus === "FAILED" || taskStatus === "CANCELED") throw new AISDKError({
+				name: "DASHSCOPE_TRANSCRIPTION_FAILED",
+				message: `Transcription ${taskStatus.toLowerCase()}. ${status.output?.message ?? ""}`
+			});
+		}
+	}
+};
+//#endregion
+//#region src/types.ts
+const DASHSCOPE_REGION_URLS = {
+	beijing: "https://dashscope.aliyuncs.com",
+	singapore: "https://dashscope-intl.aliyuncs.com",
+	us: "https://dashscope-us.aliyuncs.com",
+	germany: "https://{workspaceId}.eu-central-1.maas.aliyuncs.com"
+};
+//#endregion
+//#region src/video.ts
+const videoOptionsSchema = z.object({
+	negativePrompt: z.string().optional(),
+	promptExtend: z.boolean().optional(),
+	watermark: z.boolean().optional(),
+	resolution: z.string().optional(),
+	size: z.string().optional(),
+	duration: z.number().optional(),
+	pollIntervalMs: z.number().positive().optional(),
+	pollTimeoutMs: z.number().positive().optional()
+});
+const createTaskSchema = zodSchema(z.object({
+	output: z.object({
+		task_id: z.string(),
+		task_status: z.string()
+	}).nullish(),
+	request_id: z.string().nullish()
+}));
+const taskStatusSchema = zodSchema(z.object({
+	output: z.object({
+		task_id: z.string(),
+		task_status: z.string(),
+		video_url: z.string().nullish(),
+		submit_time: z.string().nullish(),
+		scheduled_time: z.string().nullish(),
+		end_time: z.string().nullish(),
+		code: z.string().nullish(),
+		message: z.string().nullish()
+	}).nullish(),
+	usage: z.object({
+		duration: z.number().nullish(),
+		output_video_duration: z.number().nullish(),
+		size: z.string().nullish()
+	}).nullish(),
+	request_id: z.string().nullish()
+}));
+function detectMode(modelId) {
+	return modelId.includes("-i2v") ? "i2v" : "t2v";
+}
+var DashScopeVideoModel = class {
+	specificationVersion = "v3";
+	modelId;
+	config;
+	constructor(modelId, config) {
+		this.modelId = modelId;
+		this.config = config;
+	}
+	get provider() {
+		return this.config.provider;
+	}
+	get maxVideosPerCall() {
+		return 1;
+	}
+	async doGenerate(options) {
+		const warnings = [];
+		const mode = detectMode(this.modelId);
+		const dsOptions = await parseProviderOptions({
+			provider: "dashscope",
+			providerOptions: options.providerOptions,
+			schema: videoOptionsSchema
+		});
+		const input = {};
+		if (options.prompt != null) input.prompt = options.prompt;
+		if (dsOptions?.negativePrompt != null) input.negative_prompt = dsOptions.negativePrompt;
+		if (mode === "i2v" && options.image != null) if (options.image.type === "url") input.img_url = options.image.url;
+		else input.img_url = typeof options.image.data === "string" ? options.image.data : uint8ArrayToBase64(options.image.data);
+		const parameters = {};
+		if (dsOptions?.duration != null) parameters.duration = dsOptions.duration;
+		if (options.seed != null) parameters.seed = options.seed;
+		if (dsOptions?.promptExtend != null) parameters.prompt_extend = dsOptions.promptExtend;
+		if (dsOptions?.watermark != null) parameters.watermark = dsOptions.watermark;
+		if (mode === "i2v" && dsOptions?.resolution != null) parameters.resolution = dsOptions.resolution;
+		else if (options.resolution != null) parameters.size = options.resolution.replace("x", "*");
+		else if (dsOptions?.size != null) parameters.size = dsOptions.size;
+		const { value: createResponse } = await postJsonToApi({
+			url: `${this.config.baseURL}/api/v1/services/aigc/video-generation/video-synthesis`,
+			headers: combineHeaders(this.config.headers(), options.headers, { "X-DashScope-Async": "enable" }),
+			body: {
+				model: this.modelId,
+				input,
+				parameters
+			},
+			successfulResponseHandler: createJsonResponseHandler(createTaskSchema),
+			failedResponseHandler: nativeFailedHandler,
+			abortSignal: options.abortSignal,
+			fetch: this.config.fetch
+		});
+		const taskId = createResponse.output?.task_id;
+		if (!taskId) throw new AISDKError({
+			name: "DASHSCOPE_VIDEO_ERROR",
+			message: `No task_id returned. Response: ${JSON.stringify(createResponse)}`
+		});
+		const pollInterval = dsOptions?.pollIntervalMs ?? 5e3;
+		const pollTimeout = dsOptions?.pollTimeoutMs ?? 6e5;
+		const startTime = Date.now();
+		while (true) {
+			await delay(pollInterval, { abortSignal: options.abortSignal });
+			if (Date.now() - startTime > pollTimeout) throw new AISDKError({
+				name: "DASHSCOPE_VIDEO_TIMEOUT",
+				message: `Video generation timed out after ${pollTimeout}ms`
+			});
+			const { value: status, responseHeaders } = await getFromApi({
+				url: `${this.config.baseURL}/api/v1/tasks/${taskId}`,
+				headers: combineHeaders(this.config.headers(), options.headers),
+				successfulResponseHandler: createJsonResponseHandler(taskStatusSchema),
+				failedResponseHandler: nativeFailedHandler,
+				abortSignal: options.abortSignal,
+				fetch: this.config.fetch
+			});
+			const taskStatus = status.output?.task_status;
+			if (taskStatus === "SUCCEEDED") {
+				const videoUrl = status.output?.video_url;
+				if (!videoUrl) throw new AISDKError({
+					name: "DASHSCOPE_VIDEO_ERROR",
+					message: `No video URL in response. Task ID: ${taskId}`
+				});
+				return {
+					videos: [{
+						type: "url",
+						url: videoUrl,
+						mediaType: "video/mp4"
+					}],
+					warnings,
+					response: {
+						timestamp: /* @__PURE__ */ new Date(),
+						modelId: this.modelId,
+						headers: responseHeaders
+					}
+				};
+			}
+			if (taskStatus === "FAILED" || taskStatus === "CANCELED") throw new AISDKError({
+				name: "DASHSCOPE_VIDEO_FAILED",
+				message: `Video generation ${taskStatus.toLowerCase()}. ${status.output?.message ?? ""}`
+			});
+		}
+	}
+};
+//#endregion
 //#region src/chat.ts
 const chatOptionsSchema = z.object({
 	enableThinking: z.boolean().optional(),
@@ -165,45 +765,6 @@ const chatChunkSchema = z.object({
 	})),
 	usage: usageSchema.nullish()
 });
-const failedResponseHandler$1 = createJsonErrorResponseHandler({
-	errorSchema: z.object({ error: z.object({
-		message: z.string(),
-		code: z.string().nullish(),
-		type: z.string().nullish()
-	}) }),
-	errorToMessage: (data) => data.error.message
-});
-function convertUsage$1(usage) {
-	if (!usage) return {
-		inputTokens: {
-			total: 0,
-			noCache: void 0,
-			cacheRead: void 0,
-			cacheWrite: void 0
-		},
-		outputTokens: {
-			total: 0,
-			text: void 0,
-			reasoning: void 0
-		}
-	};
-	const cacheRead = usage.prompt_tokens_details?.cached_tokens ?? void 0;
-	const cacheWrite = usage.prompt_tokens_details?.cache_creation_input_tokens ?? void 0;
-	const noCache = cacheRead != null || cacheWrite != null ? (usage.prompt_tokens ?? 0) - (cacheRead ?? 0) - (cacheWrite ?? 0) : void 0;
-	return {
-		inputTokens: {
-			total: usage.prompt_tokens ?? 0,
-			noCache,
-			cacheRead,
-			cacheWrite
-		},
-		outputTokens: {
-			total: usage.completion_tokens ?? 0,
-			text: void 0,
-			reasoning: usage.completion_tokens_details?.reasoning_tokens ?? void 0
-		}
-	};
-}
 function convertMessages(prompt) {
 	const messages = [];
 	for (const { role, content } of prompt) switch (role) {
@@ -333,14 +894,7 @@ var DashScopeChatLanguageModel = class {
 				...options.presencePenalty != null && { presence_penalty: options.presencePenalty },
 				...options.stopSequences?.length && { stop: options.stopSequences },
 				...options.seed != null && { seed: options.seed },
-				...options.responseFormat?.type === "json" && { response_format: options.responseFormat.schema != null ? {
-					type: "json_schema",
-					json_schema: {
-						schema: options.responseFormat.schema,
-						name: options.responseFormat.name ?? "response",
-						description: options.responseFormat.description
-					}
-				} : { type: "json_object" } },
+				...options.responseFormat?.type === "json" && { response_format: { type: "json_object" } },
 				...apiTools != null && {
 					tools: apiTools,
 					tool_choice: toolChoice
@@ -358,10 +912,10 @@ var DashScopeChatLanguageModel = class {
 	async doGenerate(options) {
 		const { args, warnings } = await this.getArgs(options);
 		const { responseHeaders, value: response } = await postJsonToApi({
-			url: `${this.config.baseURL}/chat/completions`,
+			url: `${this.config.baseURL}/compatible-mode/v1/chat/completions`,
 			headers: combineHeaders(this.config.headers(), options.headers),
 			body: args,
-			failedResponseHandler: failedResponseHandler$1,
+			failedResponseHandler,
 			successfulResponseHandler: createJsonResponseHandler(chatResponseSchema),
 			abortSignal: options.abortSignal,
 			fetch: this.config.fetch
@@ -388,12 +942,10 @@ var DashScopeChatLanguageModel = class {
 				unified: mapOpenAICompatibleFinishReason(choice.finish_reason),
 				raw: choice.finish_reason ?? void 0
 			},
-			usage: convertUsage$1(response.usage),
+			usage: convertOpenAICompatibleChatUsage(response.usage),
 			request: { body: JSON.stringify(args) },
 			response: {
-				id: response.id ?? void 0,
-				modelId: response.model ?? void 0,
-				timestamp: response.created ? /* @__PURE__ */ new Date(response.created * 1e3) : void 0,
+				...getResponseMetadata(response),
 				headers: responseHeaders
 			},
 			warnings
@@ -406,10 +958,10 @@ var DashScopeChatLanguageModel = class {
 			stream: true
 		};
 		const { responseHeaders, value: response } = await postJsonToApi({
-			url: `${this.config.baseURL}/chat/completions`,
+			url: `${this.config.baseURL}/compatible-mode/v1/chat/completions`,
 			headers: combineHeaders(this.config.headers(), options.headers),
 			body,
-			failedResponseHandler: failedResponseHandler$1,
+			failedResponseHandler,
 			successfulResponseHandler: createEventSourceResponseHandler(chatChunkSchema),
 			abortSignal: options.abortSignal,
 			fetch: this.config.fetch
@@ -448,9 +1000,7 @@ var DashScopeChatLanguageModel = class {
 						isFirstChunk = false;
 						controller.enqueue({
 							type: "response-metadata",
-							id: value.id ?? void 0,
-							modelId: value.model ?? void 0,
-							timestamp: value.created ? /* @__PURE__ */ new Date(value.created * 1e3) : void 0
+							...getResponseMetadata(value)
 						});
 					}
 					if (value.usage != null) usage = value.usage;
@@ -594,7 +1144,7 @@ var DashScopeChatLanguageModel = class {
 					controller.enqueue({
 						type: "finish",
 						finishReason,
-						usage: convertUsage$1(usage)
+						usage: convertOpenAICompatibleChatUsage(usage)
 					});
 				}
 			})),
@@ -619,11 +1169,6 @@ const responsesOptionsSchema = zodSchema(z.object({
 	instructions: z.string().optional(),
 	includeUsage: z.boolean().optional()
 }));
-const errorSchema = zodSchema(z.object({ error: z.object({
-	message: z.string(),
-	type: z.string().optional(),
-	code: z.string().optional()
-}) }));
 const responseSchema = zodSchema(z.object({
 	id: z.string(),
 	created_at: z.number().optional(),
@@ -641,41 +1186,6 @@ const responseSchema = zodSchema(z.object({
 	error: z.object({ message: z.string() }).nullable().optional()
 }).loose());
 const streamChunkSchema = zodSchema(z.object({ type: z.string() }).loose());
-const failedResponseHandler = createJsonErrorResponseHandler({
-	errorSchema,
-	errorToMessage: (data) => data.error.message
-});
-function convertUsage(usage) {
-	if (!usage) return {
-		inputTokens: {
-			total: 0,
-			noCache: void 0,
-			cacheRead: void 0,
-			cacheWrite: void 0
-		},
-		outputTokens: {
-			total: 0,
-			text: void 0,
-			reasoning: void 0
-		}
-	};
-	const details = usage.output_tokens_details;
-	const inputDetails = usage.input_tokens_details;
-	return {
-		inputTokens: {
-			total: usage.input_tokens ?? 0,
-			noCache: void 0,
-			cacheRead: inputDetails?.cached_tokens ?? void 0,
-			cacheWrite: void 0
-		},
-		outputTokens: {
-			total: usage.output_tokens ?? 0,
-			text: void 0,
-			reasoning: details?.reasoning_tokens ?? void 0
-		},
-		raw: usage
-	};
-}
 function mapFinishReason(status) {
 	switch (status) {
 		case "completed": return {
@@ -1047,7 +1557,7 @@ var DashScopeResponsesLanguageModel = class {
 	async doGenerate(options) {
 		const { args: body, warnings } = await this.getArgs(options);
 		const { responseHeaders, value: response } = await postJsonToApi({
-			url: `${this.config.baseURL}/responses`,
+			url: `${this.config.baseURL}/compatible-mode/v1/responses`,
 			headers: combineHeaders(this.config.headers(), options.headers),
 			body,
 			failedResponseHandler,
@@ -1066,7 +1576,7 @@ var DashScopeResponsesLanguageModel = class {
 		return {
 			content,
 			finishReason,
-			usage: convertUsage(response.usage),
+			usage: convertResponsesUsage(response.usage),
 			request: { body },
 			response: {
 				id: response.id ?? void 0,
@@ -1083,7 +1593,7 @@ var DashScopeResponsesLanguageModel = class {
 	async doStream(options) {
 		const { args: body, warnings } = await this.getArgs(options);
 		const { responseHeaders, value: response } = await postJsonToApi({
-			url: `${this.config.baseURL}/responses`,
+			url: `${this.config.baseURL}/compatible-mode/v1/responses`,
 			headers: combineHeaders(this.config.headers(), options.headers),
 			body: {
 				...body,
@@ -1151,7 +1661,7 @@ var DashScopeResponsesLanguageModel = class {
 						case "response.completed": {
 							const resp = val.response;
 							finishReason = mapFinishReason(resp?.status);
-							if (resp?.usage) usage = convertUsage(resp.usage);
+							if (resp?.usage) usage = convertResponsesUsage(resp.usage);
 							if (resp?.id) controller.enqueue({
 								type: "response-metadata",
 								id: resp.id,
@@ -1192,10 +1702,9 @@ var DashScopeResponsesLanguageModel = class {
 //#endregion
 //#region src/provider.ts
 function createDashScope(options = {}) {
-	const { region = "beijing", workspaceId, baseURL: explicitBaseURL, videoBaseURL: _explicitVideoBaseURL, includeUsage, ...rest } = options;
-	const regionUrls = DASHSCOPE_REGION_BASE_URLS[region];
-	const baseURL = (explicitBaseURL ?? regionUrls.baseURL).replace("{workspaceId}", workspaceId ?? "");
+	const { region = "beijing", workspaceId, baseURL: explicitBaseURL, includeUsage, ...rest } = options;
 	if (region === "germany" && !explicitBaseURL && !workspaceId) throw new Error("workspaceId is required when region is 'germany'. See https://help.aliyun.com/zh/model-studio/obtain-the-app-id-and-workspace-id");
+	const baseURL = (explicitBaseURL ?? DASHSCOPE_REGION_URLS[region]).replace("{workspaceId}", workspaceId ?? "");
 	const apiKey = rest.apiKey ?? process.env.DASHSCOPE_API_KEY;
 	const getHeaders = () => {
 		const headers = {};
@@ -1203,23 +1712,53 @@ function createDashScope(options = {}) {
 		if (rest.headers) Object.assign(headers, rest.headers);
 		return headers;
 	};
-	const chatConfig = {
+	const baseConfig = {
 		provider: "dashscope",
 		baseURL,
 		headers: getHeaders,
-		fetch: rest.fetch,
-		includeUsage
+		fetch: rest.fetch
 	};
-	const createChatModel = (modelId) => new DashScopeChatLanguageModel(modelId, chatConfig);
+	const createChatModel = (modelId) => new DashScopeChatLanguageModel(modelId, {
+		...baseConfig,
+		includeUsage
+	});
+	const createEmbeddingModel = (modelId) => new DashScopeEmbeddingModel(modelId, {
+		...baseConfig,
+		includeUsage
+	});
+	const createRerankingModel = (modelId) => new DashScopeRerankingModel(modelId, {
+		...baseConfig,
+		provider: "dashscope.rerank"
+	});
 	const createResponsesModel = (modelId) => new DashScopeResponsesLanguageModel(modelId, {
-		provider: "dashscope.responses",
-		baseURL,
-		headers: getHeaders,
-		fetch: rest.fetch
+		...baseConfig,
+		provider: "dashscope.responses"
 	});
 	const responses = Object.assign(createResponsesModel, { tools: responsesTools });
+	const createImageModel = (modelId) => new DashScopeImageModel(modelId, {
+		...baseConfig,
+		provider: "dashscope.image"
+	});
+	const createVideoModel = (modelId) => new DashScopeVideoModel(modelId, {
+		...baseConfig,
+		provider: "dashscope.video"
+	});
+	const createSpeechModel = (modelId) => new DashScopeSpeechModel(modelId, {
+		...baseConfig,
+		provider: "dashscope.speech"
+	});
+	const createTranscriptionModel = (modelId) => new DashScopeTranscriptionModel(modelId, {
+		...baseConfig,
+		provider: "dashscope.transcription"
+	});
 	return Object.assign(createChatModel, {
 		languageModel: createChatModel,
+		embeddingModel: createEmbeddingModel,
+		rerankingModel: createRerankingModel,
+		imageModel: createImageModel,
+		videoModel: createVideoModel,
+		speechModel: createSpeechModel,
+		transcriptionModel: createTranscriptionModel,
 		chatOptions: (chatOpts) => ({ providerOptions: { dashscope: chatOpts } }),
 		responsesOptions: (responsesOpts) => ({ providerOptions: { dashscope: responsesOpts } }),
 		responses
@@ -1229,4 +1768,4 @@ function createDashScope(options = {}) {
 //#region src/index.ts
 const dashscope = createDashScope();
 //#endregion
-export { DASHSCOPE_REGION_BASE_URLS, createDashScope, dashscope, responsesTools };
+export { DASHSCOPE_REGION_URLS, DashScopeEmbeddingModel, DashScopeImageModel, DashScopeRerankingModel, DashScopeSpeechModel, DashScopeTranscriptionModel, DashScopeVideoModel, createDashScope, dashscope, responsesTools };