@ai-sdk/baseten 2.1.37 → 2.1.39

package/CHANGELOG.md

@@ -1,5 +1,25 @@
 # @ai-sdk/baseten
 
+## 2.1.39
+
+### Patch Changes
+
+- Updated dependencies [8c65988]
+- Updated dependencies [527a163]
+  - @ai-sdk/provider@4.0.21
+  - @ai-sdk/provider-utils@5.0.53
+  - @ai-sdk/openai-compatible@3.0.61
+
+## 2.1.38
+
+### Patch Changes
+
+- ede5b89: chore: migrate package builds from tsup to tsdown
+- Updated dependencies [ede5b89]
+  - @ai-sdk/openai-compatible@3.0.60
+  - @ai-sdk/provider@4.0.20
+  - @ai-sdk/provider-utils@5.0.52
+
 ## 2.1.37
 
 ### Patch Changes

package/dist/index.d.ts

@@ -1,105 +1,109 @@
-import { ProviderV4, LanguageModelV4, EmbeddingModelV4 } from '@ai-sdk/provider';
-import { FetchFunction } from '@ai-sdk/provider-utils';
-import { z } from 'zod/v4';
-
+import { EmbeddingModelV4, LanguageModelV4, ProviderV4 } from "@ai-sdk/provider";
+import { FetchFunction } from "@ai-sdk/provider-utils";
+import { z } from "zod/v4";
+//#region src/baseten-chat-options.d.ts
 type BasetenChatModelId = 'deepseek-ai/DeepSeek-R1-0528' | 'deepseek-ai/DeepSeek-V3-0324' | 'deepseek-ai/DeepSeek-V3.1' | 'moonshotai/Kimi-K2-Instruct-0905' | 'moonshotai/Kimi-K2-Thinking' | 'Qwen/Qwen3-235B-A22B-Instruct-2507' | 'Qwen/Qwen3-Coder-480B-A35B-Instruct' | 'openai/gpt-oss-120b' | 'zai-org/GLM-4.6' | 'zai-org/GLM-4.7' | (string & {});
-
+//#endregion
+//#region src/baseten-embedding-options.d.ts
 type BasetenEmbeddingModelId = string & {};
 declare const basetenEmbeddingModelOptions: z.ZodObject<{}, z.core.$strip>;
 type BasetenEmbeddingModelOptions = z.infer<typeof basetenEmbeddingModelOptions>;
-
+//#endregion
+//#region src/baseten-provider.d.ts
 /**
  * The part of `@basetenlabs/performance-client` we use, declared structurally to
  * keep that native addon out of our dependency and type graph.
  */
 type BasetenPerformanceClient = {
-    embed(input: string[], model: string): Promise<{
-        data: {
-            embedding: number[];
-        }[];
-        usage?: {
-            total_tokens?: number;
-        };
-    }>;
+  embed(input: string[], model: string): Promise<{
+    data: {
+      embedding: number[];
+    }[];
+    usage?: {
+      total_tokens?: number;
+    };
+  }>;
 };
 type BasetenPerformanceClientConstructor = new (baseUrl: string, apiKey?: string) => BasetenPerformanceClient;
 type BasetenErrorData = z.infer<typeof basetenErrorSchema>;
 declare const basetenErrorSchema: z.ZodObject<{
-    error: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-        message: z.ZodString;
-        object: z.ZodOptional<z.ZodNullable<z.ZodString>>;
-        type: z.ZodOptional<z.ZodNullable<z.ZodString>>;
-        param: z.ZodOptional<z.ZodNullable<z.ZodAny>>;
-        code: z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodString, z.ZodNumber]>>>;
-    }, z.core.$strip>]>;
+  error: z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
+    message: z.ZodString;
+    object: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+    type: z.ZodOptional<z.ZodNullable<z.ZodString>>;
+    param: z.ZodOptional<z.ZodNullable<z.ZodAny>>;
+    code: z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodString, z.ZodNumber]>>>;
+  }, z.core.$strip>]>;
 }, z.core.$strip>;
 interface BasetenProviderSettings {
-    /**
-     * Baseten API key. Default value is taken from the `BASETEN_API_KEY`
-     * environment variable.
-     */
-    apiKey?: string;
-    /**
-     * Base URL for the Model APIs. Default: 'https://inference.baseten.co/v1'
-     */
-    baseURL?: string;
-    /**
-     * Model URL for custom models (chat or embeddings).
-     * If not supplied, the default Model APIs will be used.
-     */
-    modelURL?: string;
-    /**
-     * Custom headers to include in the requests.
-     */
-    headers?: Record<string, string>;
-    /**
-     * Custom fetch implementation. You can use it as a middleware to intercept requests,
-     * or to provide a custom fetch implementation for e.g. testing.
-     */
-    fetch?: FetchFunction;
-    /**
-     * Opt in to Baseten's native performance client for embeddings, for
-     * client-side batching and request hedging. Pass the `PerformanceClient`
-     * constructor from `@basetenlabs/performance-client`, which you install
-     * yourself:
-     *
-     * ```ts
-     * import { PerformanceClient } from '@basetenlabs/performance-client';
-     *
-     * const baseten = createBaseten({ modelURL, performanceClient: PerformanceClient });
-     * ```
-     *
-     * When omitted, embeddings go over plain HTTP to Baseten's OpenAI-compatible
-     * endpoint — the default, since this NAPI addon cannot load in edge runtimes
-     * and bundlers cannot resolve its platform binaries.
-     */
-    performanceClient?: BasetenPerformanceClientConstructor;
+  /**
+   * Baseten API key. Default value is taken from the `BASETEN_API_KEY`
+   * environment variable.
+   */
+  apiKey?: string;
+  /**
+   * Base URL for the Model APIs. Default: 'https://inference.baseten.co/v1'
+   */
+  baseURL?: string;
+  /**
+   * Model URL for custom models (chat or embeddings).
+   * If not supplied, the default Model APIs will be used.
+   */
+  modelURL?: string;
+  /**
+   * Custom headers to include in the requests.
+   */
+  headers?: Record<string, string>;
+  /**
+   * Custom fetch implementation. You can use it as a middleware to intercept requests,
+   * or to provide a custom fetch implementation for e.g. testing.
+   */
+  fetch?: FetchFunction;
+  /**
+   * Opt in to Baseten's native performance client for embeddings, for
+   * client-side batching and request hedging. Pass the `PerformanceClient`
+   * constructor from `@basetenlabs/performance-client`, which you install
+   * yourself:
+   *
+   * ```ts
+   * import { PerformanceClient } from '@basetenlabs/performance-client';
+   *
+   * const baseten = createBaseten({ modelURL, performanceClient: PerformanceClient });
+   * ```
+   *
+   * When omitted, embeddings go over plain HTTP to Baseten's OpenAI-compatible
+   * endpoint — the default, since this NAPI addon cannot load in edge runtimes
+   * and bundlers cannot resolve its platform binaries.
+   */
+  performanceClient?: BasetenPerformanceClientConstructor;
 }
 interface BasetenProvider extends ProviderV4 {
-    /**
-     * Creates a chat model for text generation.
-     */
-    (modelId?: BasetenChatModelId): LanguageModelV4;
-    /**
-     * Creates a chat model for text generation.
-     */
-    chatModel(modelId?: BasetenChatModelId): LanguageModelV4;
-    /**
-     * Creates a language model for text generation. Alias for chatModel.
-     */
-    languageModel(modelId?: BasetenChatModelId): LanguageModelV4;
-    /**
-     * Creates a embedding model for text generation.
-     */
-    embeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;
-    /**
-     * @deprecated Use `embeddingModel` instead.
-     */
-    textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;
+  /**
+   * Creates a chat model for text generation.
+   */
+  (modelId?: BasetenChatModelId): LanguageModelV4;
+  /**
+   * Creates a chat model for text generation.
+   */
+  chatModel(modelId?: BasetenChatModelId): LanguageModelV4;
+  /**
+   * Creates a language model for text generation. Alias for chatModel.
+   */
+  languageModel(modelId?: BasetenChatModelId): LanguageModelV4;
+  /**
+   * Creates a embedding model for text generation.
+   */
+  embeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;
+  /**
+   * @deprecated Use `embeddingModel` instead.
+   */
+  textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;
 }
-declare function createBaseten(options?: BasetenProviderSettings): BasetenProvider;
-declare const baseten: BasetenProvider;
-
-declare const VERSION: string;
-
-export { type BasetenChatModelId, type BasetenEmbeddingModelOptions, type BasetenErrorData, type BasetenPerformanceClient, type BasetenPerformanceClientConstructor, type BasetenProvider, type BasetenProviderSettings, VERSION, baseten, createBaseten };
+export declare function createBaseten(options?: BasetenProviderSettings): BasetenProvider;
+export declare const baseten: BasetenProvider;
+//#endregion
+//#region src/version.d.ts
+export declare const VERSION: string;
+//#endregion
+export type { BasetenChatModelId, BasetenEmbeddingModelOptions, BasetenErrorData, BasetenPerformanceClient, BasetenPerformanceClientConstructor, BasetenProvider, BasetenProviderSettings };
+//# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","names":[],"sources":["../src/baseten-chat-options.ts","../src/baseten-embedding-options.ts","../src/baseten-provider.ts","../src/version.ts"],"mappings":";;;;KAGY;;;KCEA;cAEC,8BAA4B,EAAA,cAAA,EAAA,KAAA;KAE7B,+BAA+B,EAAE,aACpC;;;;;;;KCwBG;EACV,MACE,iBACA,gBACC;IACD;MAAQ;;IACR;MAAU;;;;KAIF,2CACV,iBACA,oBACG;KAEO,mBAAmB,EAAE,aAAa;cAOxC,oBAAkB,EAAA;;;;;;;;GAWtB,EAAA,KAAA;UAQe;;;;;EAKf;;;;EAKA;;;;;EAMA;;;;EAIA,UAAU;;;;;EAMV,QAAQ;;;;;;;;;;;;;;;;;EAkBR,oBAAoB;;UAGL,wBAAwB;;;;GAItC,UAAU,qBAAqB;;;;EAKhC,UAAU,UAAU,qBAAqB;;;;EAKzC,cAAc,UAAU,qBAAqB;;;;EAK7C,eAAe,UAAU,0BAA0B;;;;EAKnD,mBAAmB,UAAU,0BAA0B;;wBAMzC,cACd,UAAS,0BACR;qBAsJU,SAAO;;;qBC9SP"}

package/dist/index.js

@@ -1,154 +1,112 @@
-// src/baseten-provider.ts
-import {
-  OpenAICompatibleChatLanguageModel,
-  OpenAICompatibleEmbeddingModel
-} from "@ai-sdk/openai-compatible";
-import {
-  NoSuchModelError
-} from "@ai-sdk/provider";
-import {
-  loadApiKey,
-  withoutTrailingSlash,
-  withUserAgentSuffix
-} from "@ai-sdk/provider-utils";
+import { OpenAICompatibleChatLanguageModel, OpenAICompatibleEmbeddingModel } from "@ai-sdk/openai-compatible";
+import { NoSuchModelError } from "@ai-sdk/provider";
+import { loadApiKey, withUserAgentSuffix, withoutTrailingSlash } from "@ai-sdk/provider-utils";
 import { z } from "zod/v4";
-
-// src/version.ts
-var VERSION = true ? "2.1.37" : "0.0.0-test";
-
-// src/baseten-provider.ts
-var MAX_EMBEDDINGS_PER_CALL = 128;
-var basetenErrorSchema = z.object({
-  error: z.union([
-    z.string(),
-    z.object({
-      message: z.string(),
-      object: z.string().nullish(),
-      type: z.string().nullish(),
-      param: z.any().nullish(),
-      code: z.union([z.string(), z.number()]).nullish()
-    })
-  ])
-});
-var basetenErrorStructure = {
-  errorSchema: basetenErrorSchema,
-  errorToMessage: (data) => typeof data.error === "string" ? data.error : data.error.message
+//#region src/version.ts
+const VERSION = "2.1.39";
+//#endregion
+//#region src/baseten-provider.ts
+/**
+* Baseten's per-request embedding input limit: larger batches are rejected with
+* `413 batch size N > maximum allowed batch size 128`. It is also the native
+* performance client's own default `batchSize`. `embedMany` splits larger
+* inputs into chunks of this size and runs them in parallel.
+*/
+const MAX_EMBEDDINGS_PER_CALL = 128;
+const basetenErrorStructure = {
+	errorSchema: z.object({ error: z.union([z.string(), z.object({
+		message: z.string(),
+		object: z.string().nullish(),
+		type: z.string().nullish(),
+		param: z.any().nullish(),
+		code: z.union([z.string(), z.number()]).nullish()
+	})]) }),
+	errorToMessage: (data) => typeof data.error === "string" ? data.error : data.error.message
 };
-var defaultBaseURL = "https://inference.baseten.co/v1";
+const defaultBaseURL = "https://inference.baseten.co/v1";
 function createBaseten(options = {}) {
-  const baseURL = withoutTrailingSlash(options.baseURL ?? defaultBaseURL);
-  const getHeaders = () => withUserAgentSuffix(
-    {
-      Authorization: `Bearer ${loadApiKey({
-        apiKey: options.apiKey,
-        environmentVariableName: "BASETEN_API_KEY",
-        description: "Baseten API key"
-      })}`,
-      ...options.headers
-    },
-    `ai-sdk-baseten/${VERSION}`
-  );
-  const getCommonModelConfig = (modelType, customURL) => ({
-    provider: `baseten.${modelType}`,
-    url: ({ path }) => {
-      if (modelType === "embedding" && customURL?.includes("/sync") && !customURL?.includes("/sync/v1")) {
-        return `${customURL}/v1${path}`;
-      }
-      return `${customURL || baseURL}${path}`;
-    },
-    headers: getHeaders,
-    fetch: options.fetch
-  });
-  const createChatModel = (modelId) => {
-    const customURL = options.modelURL;
-    if (customURL) {
-      if (customURL.includes("/sync/v1")) {
-        return new OpenAICompatibleChatLanguageModel(modelId ?? "placeholder", {
-          ...getCommonModelConfig("chat", customURL),
-          errorStructure: basetenErrorStructure,
-          // Or stream_options.include_usage is omitted and streams report no usage.
-          includeUsage: true,
-          supportsStructuredOutputs: true
-        });
-      } else if (customURL.includes("/predict")) {
-        throw new Error(
-          "Not supported. You must use a /sync/v1 endpoint for chat models."
-        );
-      }
-    }
-    return new OpenAICompatibleChatLanguageModel(modelId ?? "chat", {
-      ...getCommonModelConfig("chat"),
-      errorStructure: basetenErrorStructure,
-      includeUsage: true,
-      supportsStructuredOutputs: true
-    });
-  };
-  const createEmbeddingModel = (modelId) => {
-    const customURL = options.modelURL;
-    if (!customURL) {
-      throw new Error(
-        "No model URL provided for embeddings. Please set modelURL option for embeddings."
-      );
-    }
-    if (!customURL.includes("/sync")) {
-      throw new Error(
-        "Not supported. You must use a /sync or /sync/v1 endpoint for embeddings."
-      );
-    }
-    const model = new OpenAICompatibleEmbeddingModel(modelId ?? "embeddings", {
-      ...getCommonModelConfig("embedding", customURL),
-      errorStructure: basetenErrorStructure,
-      // Over HTTP, cap each request and let `embedMany` split and parallelise.
-      // The native client does its own batching, so let it take everything at
-      // once — `embedMany` treats Infinity as "one call".
-      maxEmbeddingsPerCall: options.performanceClient ? Number.POSITIVE_INFINITY : MAX_EMBEDDINGS_PER_CALL
-    });
-    if (!options.performanceClient) {
-      return model;
-    }
-    const performanceClient = new options.performanceClient(
-      customURL.replace("/sync/v1", "/sync"),
-      loadApiKey({
-        apiKey: options.apiKey,
-        environmentVariableName: "BASETEN_API_KEY",
-        description: "Baseten API key"
-      })
-    );
-    model.doEmbed = async (params) => {
-      if (!params.values || !Array.isArray(params.values)) {
-        throw new Error("params.values must be an array of strings");
-      }
-      const response = await performanceClient.embed(
-        params.values,
-        // model_id is for Model APIs; dedicated deployments ignore it.
-        modelId ?? "embeddings"
-      );
-      return {
-        embeddings: response.data.map((item) => item.embedding),
-        // The native client types its response as `any`; only report usage when
-        // a token count is actually present rather than `{ tokens: undefined }`.
-        usage: typeof response.usage?.total_tokens === "number" ? { tokens: response.usage.total_tokens } : void 0,
-        response: { headers: {}, body: response },
-        warnings: []
-      };
-    };
-    return model;
-  };
-  const provider = (modelId) => createChatModel(modelId);
-  provider.specificationVersion = "v4";
-  provider.chatModel = createChatModel;
-  provider.languageModel = createChatModel;
-  provider.imageModel = (modelId) => {
-    throw new NoSuchModelError({ modelId, modelType: "imageModel" });
-  };
-  provider.embeddingModel = createEmbeddingModel;
-  provider.textEmbeddingModel = createEmbeddingModel;
-  return provider;
+	const baseURL = withoutTrailingSlash(options.baseURL ?? defaultBaseURL);
+	const getHeaders = () => withUserAgentSuffix({
+		Authorization: `Bearer ${loadApiKey({
+			apiKey: options.apiKey,
+			environmentVariableName: "BASETEN_API_KEY",
+			description: "Baseten API key"
+		})}`,
+		...options.headers
+	}, `ai-sdk-baseten/${VERSION}`);
+	const getCommonModelConfig = (modelType, customURL) => ({
+		provider: `baseten.${modelType}`,
+		url: ({ path }) => {
+			if (modelType === "embedding" && customURL?.includes("/sync") && !customURL?.includes("/sync/v1")) return `${customURL}/v1${path}`;
+			return `${customURL || baseURL}${path}`;
+		},
+		headers: getHeaders,
+		fetch: options.fetch
+	});
+	const createChatModel = (modelId) => {
+		const customURL = options.modelURL;
+		if (customURL) {
+			if (customURL.includes("/sync/v1")) return new OpenAICompatibleChatLanguageModel(modelId ?? "placeholder", {
+				...getCommonModelConfig("chat", customURL),
+				errorStructure: basetenErrorStructure,
+				includeUsage: true,
+				supportsStructuredOutputs: true
+			});
+			else if (customURL.includes("/predict")) throw new Error("Not supported. You must use a /sync/v1 endpoint for chat models.");
+		}
+		return new OpenAICompatibleChatLanguageModel(modelId ?? "chat", {
+			...getCommonModelConfig("chat"),
+			errorStructure: basetenErrorStructure,
+			includeUsage: true,
+			supportsStructuredOutputs: true
+		});
+	};
+	const createEmbeddingModel = (modelId) => {
+		const customURL = options.modelURL;
+		if (!customURL) throw new Error("No model URL provided for embeddings. Please set modelURL option for embeddings.");
+		if (!customURL.includes("/sync")) throw new Error("Not supported. You must use a /sync or /sync/v1 endpoint for embeddings.");
+		const model = new OpenAICompatibleEmbeddingModel(modelId ?? "embeddings", {
+			...getCommonModelConfig("embedding", customURL),
+			errorStructure: basetenErrorStructure,
+			maxEmbeddingsPerCall: options.performanceClient ? Number.POSITIVE_INFINITY : MAX_EMBEDDINGS_PER_CALL
+		});
+		if (!options.performanceClient) return model;
+		const performanceClient = new options.performanceClient(customURL.replace("/sync/v1", "/sync"), loadApiKey({
+			apiKey: options.apiKey,
+			environmentVariableName: "BASETEN_API_KEY",
+			description: "Baseten API key"
+		}));
+		model.doEmbed = async (params) => {
+			if (!params.values || !Array.isArray(params.values)) throw new Error("params.values must be an array of strings");
+			const response = await performanceClient.embed(params.values, modelId ?? "embeddings");
+			return {
+				embeddings: response.data.map((item) => item.embedding),
+				usage: typeof response.usage?.total_tokens === "number" ? { tokens: response.usage.total_tokens } : void 0,
+				response: {
+					headers: {},
+					body: response
+				},
+				warnings: []
+			};
+		};
+		return model;
+	};
+	const provider = (modelId) => createChatModel(modelId);
+	provider.specificationVersion = "v4";
+	provider.chatModel = createChatModel;
+	provider.languageModel = createChatModel;
+	provider.imageModel = (modelId) => {
+		throw new NoSuchModelError({
+			modelId,
+			modelType: "imageModel"
+		});
+	};
+	provider.embeddingModel = createEmbeddingModel;
+	provider.textEmbeddingModel = createEmbeddingModel;
+	return provider;
 }
-var baseten = createBaseten();
-export {
-  VERSION,
-  baseten,
-  createBaseten
-};
+const baseten = createBaseten();
+//#endregion
+export { VERSION, baseten, createBaseten };
+
 //# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/baseten-provider.ts","../src/version.ts"],"sourcesContent":["import {\n  OpenAICompatibleChatLanguageModel,\n  OpenAICompatibleEmbeddingModel,\n  type ProviderErrorStructure,\n} from '@ai-sdk/openai-compatible';\nimport {\n  NoSuchModelError,\n  type EmbeddingModelV4,\n  type LanguageModelV4,\n  type ProviderV4,\n} from '@ai-sdk/provider';\nimport {\n  loadApiKey,\n  withoutTrailingSlash,\n  withUserAgentSuffix,\n  type FetchFunction,\n} from '@ai-sdk/provider-utils';\nimport { z } from 'zod/v4';\nimport type { BasetenChatModelId } from './baseten-chat-options';\nimport type { BasetenEmbeddingModelId } from './baseten-embedding-options';\nimport { VERSION } from './version';\n\n/**\n * Baseten's per-request embedding input limit: larger batches are rejected with\n * `413 batch size N > maximum allowed batch size 128`. It is also the native\n * performance client's own default `batchSize`. `embedMany` splits larger\n * inputs into chunks of this size and runs them in parallel.\n */\nconst MAX_EMBEDDINGS_PER_CALL = 128;\n\n/**\n * The part of `@basetenlabs/performance-client` we use, declared structurally to\n * keep that native addon out of our dependency and type graph.\n */\nexport type BasetenPerformanceClient = {\n  embed(\n    input: string[],\n    model: string,\n  ): Promise<{\n    data: { embedding: number[] }[];\n    usage?: { total_tokens?: number };\n  }>;\n};\n\nexport type BasetenPerformanceClientConstructor = new (\n  baseUrl: string,\n  apiKey?: string,\n) => BasetenPerformanceClient;\n\nexport type BasetenErrorData = z.infer<typeof basetenErrorSchema>;\n\n// Baseten returns two different envelopes. The Model APIs send a bare string\n// (`{\"error\":\"please check the model you provided\"}`), while dedicated\n// deployments pass through their server's OpenAI-shaped object. Parsing only\n// the string form left dedicated-deployment errors falling back to the HTTP\n// reason phrase — \"Not Found\", or nothing at all over HTTP/2.\nconst basetenErrorSchema = z.object({\n  error: z.union([\n    z.string(),\n    z.object({\n      message: z.string(),\n      object: z.string().nullish(),\n      type: z.string().nullish(),\n      param: z.any().nullish(),\n      code: z.union([z.string(), z.number()]).nullish(),\n    }),\n  ]),\n});\n\nconst basetenErrorStructure: ProviderErrorStructure<BasetenErrorData> = {\n  errorSchema: basetenErrorSchema,\n  errorToMessage: data =>\n    typeof data.error === 'string' ? data.error : data.error.message,\n};\n\nexport interface BasetenProviderSettings {\n  /**\n   * Baseten API key. Default value is taken from the `BASETEN_API_KEY`\n   * environment variable.\n   */\n  apiKey?: string;\n\n  /**\n   * Base URL for the Model APIs. Default: 'https://inference.baseten.co/v1'\n   */\n  baseURL?: string;\n\n  /**\n   * Model URL for custom models (chat or embeddings).\n   * If not supplied, the default Model APIs will be used.\n   */\n  modelURL?: string;\n  /**\n   * Custom headers to include in the requests.\n   */\n  headers?: Record<string, string>;\n\n  /**\n   * Custom fetch implementation. You can use it as a middleware to intercept requests,\n   * or to provide a custom fetch implementation for e.g. testing.\n   */\n  fetch?: FetchFunction;\n\n  /**\n   * Opt in to Baseten's native performance client for embeddings, for\n   * client-side batching and request hedging. Pass the `PerformanceClient`\n   * constructor from `@basetenlabs/performance-client`, which you install\n   * yourself:\n   *\n   * ```ts\n   * import { PerformanceClient } from '@basetenlabs/performance-client';\n   *\n   * const baseten = createBaseten({ modelURL, performanceClient: PerformanceClient });\n   * ```\n   *\n   * When omitted, embeddings go over plain HTTP to Baseten's OpenAI-compatible\n   * endpoint — the default, since this NAPI addon cannot load in edge runtimes\n   * and bundlers cannot resolve its platform binaries.\n   */\n  performanceClient?: BasetenPerformanceClientConstructor;\n}\n\nexport interface BasetenProvider extends ProviderV4 {\n  /**\n   * Creates a chat model for text generation.\n   */\n  (modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a chat model for text generation.\n   */\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a language model for text generation. Alias for chatModel.\n   */\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a embedding model for text generation.\n   */\n  embeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;\n\n  /**\n   * @deprecated Use `embeddingModel` instead.\n   */\n  textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;\n}\n\n// by default, we use the Model APIs\nconst defaultBaseURL = 'https://inference.baseten.co/v1';\n\nexport function createBaseten(\n  options: BasetenProviderSettings = {},\n): BasetenProvider {\n  const baseURL = withoutTrailingSlash(options.baseURL ?? defaultBaseURL);\n  const getHeaders = () =>\n    withUserAgentSuffix(\n      {\n        Authorization: `Bearer ${loadApiKey({\n          apiKey: options.apiKey,\n          environmentVariableName: 'BASETEN_API_KEY',\n          description: 'Baseten API key',\n        })}`,\n        ...options.headers,\n      },\n      `ai-sdk-baseten/${VERSION}`,\n    );\n\n  interface CommonModelConfig {\n    provider: string;\n    url: ({ path }: { path: string }) => string;\n    headers: () => Record<string, string>;\n    fetch?: FetchFunction;\n  }\n\n  const getCommonModelConfig = (\n    modelType: string,\n    customURL?: string,\n  ): CommonModelConfig => ({\n    provider: `baseten.${modelType}`,\n    url: ({ path }) => {\n      // For embeddings with /sync URLs (but not /sync/v1), we need to add /v1\n      if (\n        modelType === 'embedding' &&\n        customURL?.includes('/sync') &&\n        !customURL?.includes('/sync/v1')\n      ) {\n        return `${customURL}/v1${path}`;\n      }\n      return `${customURL || baseURL}${path}`;\n    },\n    headers: getHeaders,\n    fetch: options.fetch,\n  });\n\n  const createChatModel = (modelId?: BasetenChatModelId) => {\n    const customURL = options.modelURL;\n    if (customURL) {\n      if (customURL.includes('/sync/v1')) {\n        return new OpenAICompatibleChatLanguageModel(modelId ?? 'placeholder', {\n          ...getCommonModelConfig('chat', customURL),\n          errorStructure: basetenErrorStructure,\n          // Or stream_options.include_usage is omitted and streams report no usage.\n          includeUsage: true,\n          supportsStructuredOutputs: true,\n        });\n      } else if (customURL.includes('/predict')) {\n        throw new Error(\n          'Not supported. You must use a /sync/v1 endpoint for chat models.',\n        );\n      }\n    }\n\n    return new OpenAICompatibleChatLanguageModel(modelId ?? 'chat', {\n      ...getCommonModelConfig('chat'),\n      errorStructure: basetenErrorStructure,\n      includeUsage: true,\n      supportsStructuredOutputs: true,\n    });\n  };\n\n  const createEmbeddingModel = (modelId?: BasetenEmbeddingModelId) => {\n    const customURL = options.modelURL;\n    if (!customURL) {\n      throw new Error(\n        'No model URL provided for embeddings. Please set modelURL option for embeddings.',\n      );\n    }\n\n    if (!customURL.includes('/sync')) {\n      throw new Error(\n        'Not supported. You must use a /sync or /sync/v1 endpoint for embeddings.',\n      );\n    }\n\n    // BEI embedding deployments are OpenAI-compatible with no extra settings, so\n    // plain HTTP is the default and needs no override.\n    const model = new OpenAICompatibleEmbeddingModel(modelId ?? 'embeddings', {\n      ...getCommonModelConfig('embedding', customURL),\n      errorStructure: basetenErrorStructure,\n      // Over HTTP, cap each request and let `embedMany` split and parallelise.\n      // The native client does its own batching, so let it take everything at\n      // once — `embedMany` treats Infinity as \"one call\".\n      maxEmbeddingsPerCall: options.performanceClient\n        ? Number.POSITIVE_INFINITY\n        : MAX_EMBEDDINGS_PER_CALL,\n    });\n\n    if (!options.performanceClient) {\n      return model;\n    }\n\n    // Opted in to the native client. It appends /v1 itself, so hand it the bare\n    // /sync form.\n    const performanceClient = new options.performanceClient(\n      customURL.replace('/sync/v1', '/sync'),\n      loadApiKey({\n        apiKey: options.apiKey,\n        environmentVariableName: 'BASETEN_API_KEY',\n        description: 'Baseten API key',\n      }),\n    );\n\n    model.doEmbed = async params => {\n      if (!params.values || !Array.isArray(params.values)) {\n        throw new Error('params.values must be an array of strings');\n      }\n\n      const response = await performanceClient.embed(\n        params.values,\n        // model_id is for Model APIs; dedicated deployments ignore it.\n        modelId ?? 'embeddings',\n      );\n\n      return {\n        embeddings: response.data.map(item => item.embedding),\n        // The native client types its response as `any`; only report usage when\n        // a token count is actually present rather than `{ tokens: undefined }`.\n        usage:\n          typeof response.usage?.total_tokens === 'number'\n            ? { tokens: response.usage.total_tokens }\n            : undefined,\n        response: { headers: {}, body: response },\n        warnings: [],\n      };\n    };\n\n    return model;\n  };\n\n  const provider = (modelId?: BasetenChatModelId) => createChatModel(modelId);\n\n  provider.specificationVersion = 'v4' as const;\n  provider.chatModel = createChatModel;\n  provider.languageModel = createChatModel;\n  provider.imageModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });\n  };\n  provider.embeddingModel = createEmbeddingModel;\n  provider.textEmbeddingModel = createEmbeddingModel;\n  return provider;\n}\n\nexport const baseten = createBaseten();\n","// Version string of this package injected at build time.\ndeclare const __PACKAGE_VERSION__: string | undefined;\nexport const VERSION: string =\n  typeof __PACKAGE_VERSION__ !== 'undefined'\n    ? __PACKAGE_VERSION__\n    : '0.0.0-test';\n"],"mappings":";AAAA;AAAA,EACE;AAAA,EACA;AAAA,OAEK;AACP;AAAA,EACE;AAAA,OAIK;AACP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OAEK;AACP,SAAS,SAAS;;;ACfX,IAAM,UACX,OACI,WACA;;;ADuBN,IAAM,0BAA0B;AA4BhC,IAAM,qBAAqB,EAAE,OAAO;AAAA,EAClC,OAAO,EAAE,MAAM;AAAA,IACb,EAAE,OAAO;AAAA,IACT,EAAE,OAAO;AAAA,MACP,SAAS,EAAE,OAAO;AAAA,MAClB,QAAQ,EAAE,OAAO,EAAE,QAAQ;AAAA,MAC3B,MAAM,EAAE,OAAO,EAAE,QAAQ;AAAA,MACzB,OAAO,EAAE,IAAI,EAAE,QAAQ;AAAA,MACvB,MAAM,EAAE,MAAM,CAAC,EAAE,OAAO,GAAG,EAAE,OAAO,CAAC,CAAC,EAAE,QAAQ;AAAA,IAClD,CAAC;AAAA,EACH,CAAC;AACH,CAAC;AAED,IAAM,wBAAkE;AAAA,EACtE,aAAa;AAAA,EACb,gBAAgB,UACd,OAAO,KAAK,UAAU,WAAW,KAAK,QAAQ,KAAK,MAAM;AAC7D;AA6EA,IAAM,iBAAiB;AAEhB,SAAS,cACd,UAAmC,CAAC,GACnB;AACjB,QAAM,UAAU,qBAAqB,QAAQ,WAAW,cAAc;AACtE,QAAM,aAAa,MACjB;AAAA,IACE;AAAA,MACE,eAAe,UAAU,WAAW;AAAA,QAClC,QAAQ,QAAQ;AAAA,QAChB,yBAAyB;AAAA,QACzB,aAAa;AAAA,MACf,CAAC,CAAC;AAAA,MACF,GAAG,QAAQ;AAAA,IACb;AAAA,IACA,kBAAkB,OAAO;AAAA,EAC3B;AASF,QAAM,uBAAuB,CAC3B,WACA,eACuB;AAAA,IACvB,UAAU,WAAW,SAAS;AAAA,IAC9B,KAAK,CAAC,EAAE,KAAK,MAAM;AAEjB,UACE,cAAc,eACd,WAAW,SAAS,OAAO,KAC3B,CAAC,WAAW,SAAS,UAAU,GAC/B;AACA,eAAO,GAAG,SAAS,MAAM,IAAI;AAAA,MAC/B;AACA,aAAO,GAAG,aAAa,OAAO,GAAG,IAAI;AAAA,IACvC;AAAA,IACA,SAAS;AAAA,IACT,OAAO,QAAQ;AAAA,EACjB;AAEA,QAAM,kBAAkB,CAAC,YAAiC;AACxD,UAAM,YAAY,QAAQ;AAC1B,QAAI,WAAW;AACb,UAAI,UAAU,SAAS,UAAU,GAAG;AAClC,eAAO,IAAI,kCAAkC,WAAW,eAAe;AAAA,UACrE,GAAG,qBAAqB,QAAQ,SAAS;AAAA,UACzC,gBAAgB;AAAA;AAAA,UAEhB,cAAc;AAAA,UACd,2BAA2B;AAAA,QAC7B,CAAC;AAAA,MACH,WAAW,UAAU,SAAS,UAAU,GAAG;AACzC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAEA,WAAO,IAAI,kCAAkC,WAAW,QAAQ;AAAA,MAC9D,GAAG,qBAAqB,MAAM;AAAA,MAC9B,gBAAgB;AAAA,MAChB,cAAc;AAAA,MACd,2BAA2B;AAAA,IAC7B,CAAC;AAAA,EACH;AAEA,QAAM,uBAAuB,CAAC,YAAsC;AAClE,UAAM,YAAY,QAAQ;AAC1B,QAAI,CAAC,WAAW;AACd,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAEA,QAAI,CAAC,UAAU,SAAS,OAAO,GAAG;AAChC,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAIA,UAAM,QAAQ,IAAI,+BAA+B,WAAW,cAAc;AAAA,MACxE,GAAG,qBAAqB,aAAa,SAAS;AAAA,MAC9C,gBAAgB;AAAA;AAAA;AAAA;AAAA,MAIhB,sBAAsB,QAAQ,oBAC1B,OAAO,oBACP;AAAA,IACN,CAAC;AAED,QAAI,CAAC,QAAQ,mBAAmB;AAC9B,aAAO;AAAA,IACT;AAIA,UAAM,oBAAoB,IAAI,QAAQ;AAAA,MACpC,UAAU,QAAQ,YAAY,OAAO;AAAA,MACrC,WAAW;AAAA,QACT,QAAQ,QAAQ;AAAA,QAChB,yBAAyB;AAAA,QACzB,aAAa;AAAA,MACf,CAAC;AAAA,IACH;AAEA,UAAM,UAAU,OAAM,WAAU;AAC9B,UAAI,CAAC,OAAO,UAAU,CAAC,MAAM,QAAQ,OAAO,MAAM,GAAG;AACnD,cAAM,IAAI,MAAM,2CAA2C;AAAA,MAC7D;AAEA,YAAM,WAAW,MAAM,kBAAkB;AAAA,QACvC,OAAO;AAAA;AAAA,QAEP,WAAW;AAAA,MACb;AAEA,aAAO;AAAA,QACL,YAAY,SAAS,KAAK,IAAI,UAAQ,KAAK,SAAS;AAAA;AAAA;AAAA,QAGpD,OACE,OAAO,SAAS,OAAO,iBAAiB,WACpC,EAAE,QAAQ,SAAS,MAAM,aAAa,IACtC;AAAA,QACN,UAAU,EAAE,SAAS,CAAC,GAAG,MAAM,SAAS;AAAA,QACxC,UAAU,CAAC;AAAA,MACb;AAAA,IACF;AAEA,WAAO;AAAA,EACT;AAEA,QAAM,WAAW,CAAC,YAAiC,gBAAgB,OAAO;AAE1E,WAAS,uBAAuB;AAChC,WAAS,YAAY;AACrB,WAAS,gBAAgB;AACzB,WAAS,aAAa,CAAC,YAAoB;AACzC,UAAM,IAAI,iBAAiB,EAAE,SAAS,WAAW,aAAa,CAAC;AAAA,EACjE;AACA,WAAS,iBAAiB;AAC1B,WAAS,qBAAqB;AAC9B,SAAO;AACT;AAEO,IAAM,UAAU,cAAc;","names":[]}
+{"version":3,"file":"index.js","names":[],"sources":["../src/version.ts","../src/baseten-provider.ts"],"sourcesContent":["// Version string of this package injected at build time.\ndeclare const __PACKAGE_VERSION__: string | undefined;\nexport const VERSION: string =\n  typeof __PACKAGE_VERSION__ !== 'undefined'\n    ? __PACKAGE_VERSION__\n    : '0.0.0-test';\n","import {\n  OpenAICompatibleChatLanguageModel,\n  OpenAICompatibleEmbeddingModel,\n  type ProviderErrorStructure,\n} from '@ai-sdk/openai-compatible';\nimport {\n  NoSuchModelError,\n  type EmbeddingModelV4,\n  type LanguageModelV4,\n  type ProviderV4,\n} from '@ai-sdk/provider';\nimport {\n  loadApiKey,\n  withoutTrailingSlash,\n  withUserAgentSuffix,\n  type FetchFunction,\n} from '@ai-sdk/provider-utils';\nimport { z } from 'zod/v4';\nimport type { BasetenChatModelId } from './baseten-chat-options';\nimport type { BasetenEmbeddingModelId } from './baseten-embedding-options';\nimport { VERSION } from './version';\n\n/**\n * Baseten's per-request embedding input limit: larger batches are rejected with\n * `413 batch size N > maximum allowed batch size 128`. It is also the native\n * performance client's own default `batchSize`. `embedMany` splits larger\n * inputs into chunks of this size and runs them in parallel.\n */\nconst MAX_EMBEDDINGS_PER_CALL = 128;\n\n/**\n * The part of `@basetenlabs/performance-client` we use, declared structurally to\n * keep that native addon out of our dependency and type graph.\n */\nexport type BasetenPerformanceClient = {\n  embed(\n    input: string[],\n    model: string,\n  ): Promise<{\n    data: { embedding: number[] }[];\n    usage?: { total_tokens?: number };\n  }>;\n};\n\nexport type BasetenPerformanceClientConstructor = new (\n  baseUrl: string,\n  apiKey?: string,\n) => BasetenPerformanceClient;\n\nexport type BasetenErrorData = z.infer<typeof basetenErrorSchema>;\n\n// Baseten returns two different envelopes. The Model APIs send a bare string\n// (`{\"error\":\"please check the model you provided\"}`), while dedicated\n// deployments pass through their server's OpenAI-shaped object. Parsing only\n// the string form left dedicated-deployment errors falling back to the HTTP\n// reason phrase — \"Not Found\", or nothing at all over HTTP/2.\nconst basetenErrorSchema = z.object({\n  error: z.union([\n    z.string(),\n    z.object({\n      message: z.string(),\n      object: z.string().nullish(),\n      type: z.string().nullish(),\n      param: z.any().nullish(),\n      code: z.union([z.string(), z.number()]).nullish(),\n    }),\n  ]),\n});\n\nconst basetenErrorStructure: ProviderErrorStructure<BasetenErrorData> = {\n  errorSchema: basetenErrorSchema,\n  errorToMessage: data =>\n    typeof data.error === 'string' ? data.error : data.error.message,\n};\n\nexport interface BasetenProviderSettings {\n  /**\n   * Baseten API key. Default value is taken from the `BASETEN_API_KEY`\n   * environment variable.\n   */\n  apiKey?: string;\n\n  /**\n   * Base URL for the Model APIs. Default: 'https://inference.baseten.co/v1'\n   */\n  baseURL?: string;\n\n  /**\n   * Model URL for custom models (chat or embeddings).\n   * If not supplied, the default Model APIs will be used.\n   */\n  modelURL?: string;\n  /**\n   * Custom headers to include in the requests.\n   */\n  headers?: Record<string, string>;\n\n  /**\n   * Custom fetch implementation. You can use it as a middleware to intercept requests,\n   * or to provide a custom fetch implementation for e.g. testing.\n   */\n  fetch?: FetchFunction;\n\n  /**\n   * Opt in to Baseten's native performance client for embeddings, for\n   * client-side batching and request hedging. Pass the `PerformanceClient`\n   * constructor from `@basetenlabs/performance-client`, which you install\n   * yourself:\n   *\n   * ```ts\n   * import { PerformanceClient } from '@basetenlabs/performance-client';\n   *\n   * const baseten = createBaseten({ modelURL, performanceClient: PerformanceClient });\n   * ```\n   *\n   * When omitted, embeddings go over plain HTTP to Baseten's OpenAI-compatible\n   * endpoint — the default, since this NAPI addon cannot load in edge runtimes\n   * and bundlers cannot resolve its platform binaries.\n   */\n  performanceClient?: BasetenPerformanceClientConstructor;\n}\n\nexport interface BasetenProvider extends ProviderV4 {\n  /**\n   * Creates a chat model for text generation.\n   */\n  (modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a chat model for text generation.\n   */\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a language model for text generation. Alias for chatModel.\n   */\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a embedding model for text generation.\n   */\n  embeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;\n\n  /**\n   * @deprecated Use `embeddingModel` instead.\n   */\n  textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV4;\n}\n\n// by default, we use the Model APIs\nconst defaultBaseURL = 'https://inference.baseten.co/v1';\n\nexport function createBaseten(\n  options: BasetenProviderSettings = {},\n): BasetenProvider {\n  const baseURL = withoutTrailingSlash(options.baseURL ?? defaultBaseURL);\n  const getHeaders = () =>\n    withUserAgentSuffix(\n      {\n        Authorization: `Bearer ${loadApiKey({\n          apiKey: options.apiKey,\n          environmentVariableName: 'BASETEN_API_KEY',\n          description: 'Baseten API key',\n        })}`,\n        ...options.headers,\n      },\n      `ai-sdk-baseten/${VERSION}`,\n    );\n\n  interface CommonModelConfig {\n    provider: string;\n    url: ({ path }: { path: string }) => string;\n    headers: () => Record<string, string>;\n    fetch?: FetchFunction;\n  }\n\n  const getCommonModelConfig = (\n    modelType: string,\n    customURL?: string,\n  ): CommonModelConfig => ({\n    provider: `baseten.${modelType}`,\n    url: ({ path }) => {\n      // For embeddings with /sync URLs (but not /sync/v1), we need to add /v1\n      if (\n        modelType === 'embedding' &&\n        customURL?.includes('/sync') &&\n        !customURL?.includes('/sync/v1')\n      ) {\n        return `${customURL}/v1${path}`;\n      }\n      return `${customURL || baseURL}${path}`;\n    },\n    headers: getHeaders,\n    fetch: options.fetch,\n  });\n\n  const createChatModel = (modelId?: BasetenChatModelId) => {\n    const customURL = options.modelURL;\n    if (customURL) {\n      if (customURL.includes('/sync/v1')) {\n        return new OpenAICompatibleChatLanguageModel(modelId ?? 'placeholder', {\n          ...getCommonModelConfig('chat', customURL),\n          errorStructure: basetenErrorStructure,\n          // Or stream_options.include_usage is omitted and streams report no usage.\n          includeUsage: true,\n          supportsStructuredOutputs: true,\n        });\n      } else if (customURL.includes('/predict')) {\n        throw new Error(\n          'Not supported. You must use a /sync/v1 endpoint for chat models.',\n        );\n      }\n    }\n\n    return new OpenAICompatibleChatLanguageModel(modelId ?? 'chat', {\n      ...getCommonModelConfig('chat'),\n      errorStructure: basetenErrorStructure,\n      includeUsage: true,\n      supportsStructuredOutputs: true,\n    });\n  };\n\n  const createEmbeddingModel = (modelId?: BasetenEmbeddingModelId) => {\n    const customURL = options.modelURL;\n    if (!customURL) {\n      throw new Error(\n        'No model URL provided for embeddings. Please set modelURL option for embeddings.',\n      );\n    }\n\n    if (!customURL.includes('/sync')) {\n      throw new Error(\n        'Not supported. You must use a /sync or /sync/v1 endpoint for embeddings.',\n      );\n    }\n\n    // BEI embedding deployments are OpenAI-compatible with no extra settings, so\n    // plain HTTP is the default and needs no override.\n    const model = new OpenAICompatibleEmbeddingModel(modelId ?? 'embeddings', {\n      ...getCommonModelConfig('embedding', customURL),\n      errorStructure: basetenErrorStructure,\n      // Over HTTP, cap each request and let `embedMany` split and parallelise.\n      // The native client does its own batching, so let it take everything at\n      // once — `embedMany` treats Infinity as \"one call\".\n      maxEmbeddingsPerCall: options.performanceClient\n        ? Number.POSITIVE_INFINITY\n        : MAX_EMBEDDINGS_PER_CALL,\n    });\n\n    if (!options.performanceClient) {\n      return model;\n    }\n\n    // Opted in to the native client. It appends /v1 itself, so hand it the bare\n    // /sync form.\n    const performanceClient = new options.performanceClient(\n      customURL.replace('/sync/v1', '/sync'),\n      loadApiKey({\n        apiKey: options.apiKey,\n        environmentVariableName: 'BASETEN_API_KEY',\n        description: 'Baseten API key',\n      }),\n    );\n\n    model.doEmbed = async params => {\n      if (!params.values || !Array.isArray(params.values)) {\n        throw new Error('params.values must be an array of strings');\n      }\n\n      const response = await performanceClient.embed(\n        params.values,\n        // model_id is for Model APIs; dedicated deployments ignore it.\n        modelId ?? 'embeddings',\n      );\n\n      return {\n        embeddings: response.data.map(item => item.embedding),\n        // The native client types its response as `any`; only report usage when\n        // a token count is actually present rather than `{ tokens: undefined }`.\n        usage:\n          typeof response.usage?.total_tokens === 'number'\n            ? { tokens: response.usage.total_tokens }\n            : undefined,\n        response: { headers: {}, body: response },\n        warnings: [],\n      };\n    };\n\n    return model;\n  };\n\n  const provider = (modelId?: BasetenChatModelId) => createChatModel(modelId);\n\n  provider.specificationVersion = 'v4' as const;\n  provider.chatModel = createChatModel;\n  provider.languageModel = createChatModel;\n  provider.imageModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });\n  };\n  provider.embeddingModel = createEmbeddingModel;\n  provider.textEmbeddingModel = createEmbeddingModel;\n  return provider;\n}\n\nexport const baseten = createBaseten();\n"],"mappings":";;;;;AAEA,MAAa,UAAA;;;;;;;;;AC0Bb,MAAM,0BAA0B;AAyChC,MAAM,wBAAkE;CACtE,aAdyB,EAAE,OAAO,EAClC,OAAO,EAAE,MAAM,CACb,EAAE,OAAO,GACT,EAAE,OAAO;EACP,SAAS,EAAE,OAAO;EAClB,QAAQ,EAAE,OAAO,CAAC,CAAC,QAAQ;EAC3B,MAAM,EAAE,OAAO,CAAC,CAAC,QAAQ;EACzB,OAAO,EAAE,IAAI,CAAC,CAAC,QAAQ;EACvB,MAAM,EAAE,MAAM,CAAC,EAAE,OAAO,GAAG,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,QAAQ;CAClD,CAAC,CACH,CAAC,EACH,CAGgC;CAC9B,iBAAgB,SACd,OAAO,KAAK,UAAU,WAAW,KAAK,QAAQ,KAAK,MAAM;AAC7D;AA6EA,MAAM,iBAAiB;AAEvB,SAAgB,cACd,UAAmC,CAAC,GACnB;CACjB,MAAM,UAAU,qBAAqB,QAAQ,WAAW,cAAc;CACtE,MAAM,mBACJ,oBACE;EACE,eAAe,UAAU,WAAW;GAClC,QAAQ,QAAQ;GAChB,yBAAyB;GACzB,aAAa;EACf,CAAC;EACD,GAAG,QAAQ;CACb,GACA,kBAAkB,SACpB;CASF,MAAM,wBACJ,WACA,eACuB;EACvB,UAAU,WAAW;EACrB,MAAM,EAAE,WAAW;GAEjB,IACE,cAAc,eACd,WAAW,SAAS,OAAO,KAC3B,CAAC,WAAW,SAAS,UAAU,GAE/B,OAAO,GAAG,UAAU,KAAK;GAE3B,OAAO,GAAG,aAAa,UAAU;EACnC;EACA,SAAS;EACT,OAAO,QAAQ;CACjB;CAEA,MAAM,mBAAmB,YAAiC;EACxD,MAAM,YAAY,QAAQ;EAC1B,IAAI,WAAW;GACb,IAAI,UAAU,SAAS,UAAU,GAC/B,OAAO,IAAI,kCAAkC,WAAW,eAAe;IACrE,GAAG,qBAAqB,QAAQ,SAAS;IACzC,gBAAgB;IAEhB,cAAc;IACd,2BAA2B;GAC7B,CAAC;QACI,IAAI,UAAU,SAAS,UAAU,GACtC,MAAM,IAAI,MACR,kEACF;EAEJ;EAEA,OAAO,IAAI,kCAAkC,WAAW,QAAQ;GAC9D,GAAG,qBAAqB,MAAM;GAC9B,gBAAgB;GAChB,cAAc;GACd,2BAA2B;EAC7B,CAAC;CACH;CAEA,MAAM,wBAAwB,YAAsC;EAClE,MAAM,YAAY,QAAQ;EAC1B,IAAI,CAAC,WACH,MAAM,IAAI,MACR,kFACF;EAGF,IAAI,CAAC,UAAU,SAAS,OAAO,GAC7B,MAAM,IAAI,MACR,0EACF;EAKF,MAAM,QAAQ,IAAI,+BAA+B,WAAW,cAAc;GACxE,GAAG,qBAAqB,aAAa,SAAS;GAC9C,gBAAgB;GAIhB,sBAAsB,QAAQ,oBAC1B,OAAO,oBACP;EACN,CAAC;EAED,IAAI,CAAC,QAAQ,mBACX,OAAO;EAKT,MAAM,oBAAoB,IAAI,QAAQ,kBACpC,UAAU,QAAQ,YAAY,OAAO,GACrC,WAAW;GACT,QAAQ,QAAQ;GAChB,yBAAyB;GACzB,aAAa;EACf,CAAC,CACH;EAEA,MAAM,UAAU,OAAM,WAAU;GAC9B,IAAI,CAAC,OAAO,UAAU,CAAC,MAAM,QAAQ,OAAO,MAAM,GAChD,MAAM,IAAI,MAAM,2CAA2C;GAG7D,MAAM,WAAW,MAAM,kBAAkB,MACvC,OAAO,QAEP,WAAW,YACb;GAEA,OAAO;IACL,YAAY,SAAS,KAAK,KAAI,SAAQ,KAAK,SAAS;IAGpD,OACE,OAAO,SAAS,OAAO,iBAAiB,WACpC,EAAE,QAAQ,SAAS,MAAM,aAAa,IACtC,KAAA;IACN,UAAU;KAAE,SAAS,CAAC;KAAG,MAAM;IAAS;IACxC,UAAU,CAAC;GACb;EACF;EAEA,OAAO;CACT;CAEA,MAAM,YAAY,YAAiC,gBAAgB,OAAO;CAE1E,SAAS,uBAAuB;CAChC,SAAS,YAAY;CACrB,SAAS,gBAAgB;CACzB,SAAS,cAAc,YAAoB;EACzC,MAAM,IAAI,iBAAiB;GAAE;GAAS,WAAW;EAAa,CAAC;CACjE;CACA,SAAS,iBAAiB;CAC1B,SAAS,qBAAqB;CAC9B,OAAO;AACT;AAEA,MAAa,UAAU,cAAc"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/baseten",
-  "version": "2.1.37",
+  "version": "2.1.39",
   "type": "module",
   "license": "Apache-2.0",
   "sideEffects": false,
@@ -29,15 +29,15 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/openai-compatible": "3.0.59",
-    "@ai-sdk/provider": "4.0.19",
-    "@ai-sdk/provider-utils": "5.0.51"
+    "@ai-sdk/openai-compatible": "3.0.61",
+    "@ai-sdk/provider": "4.0.21",
+    "@ai-sdk/provider-utils": "5.0.53"
   },
   "devDependencies": {
-    "@ai-sdk/test-server": "2.0.2",
+    "@ai-sdk/test-server": "2.0.3",
     "@types/node": "22.19.19",
     "@vercel/ai-tsconfig": "0.0.0",
-    "tsup": "^8.5.1",
+    "tsdown": "^0.23.0",
     "typescript": "5.8.3",
     "zod": "3.25.76"
   },
@@ -64,9 +64,9 @@
     "ai"
   ],
   "scripts": {
-    "build": "pnpm clean && tsup --tsconfig tsconfig.build.json",
-    "build:watch": "pnpm clean && tsup --watch",
-    "clean": "del-cli dist docs *.tsbuildinfo",
+    "build": "pnpm clean && tsdown",
+    "build:watch": "pnpm clean && tsdown --watch",
+    "clean": "del-cli docs *.tsbuildinfo",
     "type-check": "tsc --build",
     "test": "pnpm test:node && pnpm test:edge",
     "test:update": "pnpm test:node -u",