@ai-sdk/baseten 0.0.1

package/CHANGELOG.md

@@ -0,0 +1,7 @@
+# @ai-sdk/baseten
+
+## 0.0.1
+
+### Patch Changes
+
+- 9694b94: initial stable release

package/LICENSE

@@ -0,0 +1,13 @@
+Copyright 2023 Vercel, Inc.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

package/README.md

@@ -0,0 +1,35 @@
+# AI SDK - Baseten Provider
+
+The **[Baseten provider](https://ai-sdk.dev/providers/ai-sdk-providers/baseten)** for the [AI SDK](https://ai-sdk.dev/docs) contains language model and embedding model support for the [Baseten](https://baseten.co) platform.
+
+## Setup
+
+The Baseten provider is available in the `@ai-sdk/baseten` module. You can install it with
+
+```bash
+npm i @ai-sdk/baseten
+```
+
+## Provider Instance
+
+You can import the default provider instance `baseten` from `@ai-sdk/baseten`:
+
+```ts
+import { baseten } from '@ai-sdk/baseten';
+```
+
+## Language Model Example (Model APIs)
+
+```ts
+import { baseten } from '@ai-sdk/baseten';
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: baseten('deepseek-ai/DeepSeek-V3-0324'),
+  prompt: 'What is the meaning of life?',
+});
+```
+
+## Documentation
+
+Please check out the **[Baseten provider](https://ai-sdk.dev/providers/ai-sdk-providers/baseten)** for more information.

package/dist/index.d.mts

@@ -0,0 +1,61 @@
+import { ProviderV2, LanguageModelV2, EmbeddingModelV2 } from '@ai-sdk/provider';
+import { FetchFunction } from '@ai-sdk/provider-utils';
+import { z } from 'zod/v4';
+
+type BasetenChatModelId = 'deepseek-ai/DeepSeek-R1-0528' | 'deepseek-ai/DeepSeek-V3-0324' | 'meta-llama/Llama-4-Maverick-17B-128E-Instruct' | 'meta-llama/Llama-4-Scout-17B-16E-Instruct' | 'moonshotai/Kimi-K2-Instruct' | 'Qwen/Qwen3-235B-A22B-Instruct-2507' | 'Qwen/Qwen3-Coder-480B-A35B-Instruct' | 'openai/gpt-oss-120b' | (string & {});
+
+type BasetenEmbeddingModelId = string & {};
+
+type BasetenErrorData = z.infer<typeof basetenErrorSchema>;
+declare const basetenErrorSchema: z.ZodObject<{
+    error: z.ZodString;
+}, 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;
+}
+interface BasetenProvider extends ProviderV2 {
+    /**
+  Creates a chat model for text generation.
+  */
+    (modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a chat model for text generation.
+  */
+    chatModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a language model for text generation. Alias for chatModel.
+  */
+    languageModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a text embedding model for text generation.
+  */
+    textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV2<string>;
+}
+declare function createBaseten(options?: BasetenProviderSettings): BasetenProvider;
+declare const baseten: BasetenProvider;
+
+declare const VERSION: string;
+
+export { type BasetenChatModelId, type BasetenErrorData, type BasetenProvider, type BasetenProviderSettings, VERSION, baseten, createBaseten };

package/dist/index.d.ts

@@ -0,0 +1,61 @@
+import { ProviderV2, LanguageModelV2, EmbeddingModelV2 } from '@ai-sdk/provider';
+import { FetchFunction } from '@ai-sdk/provider-utils';
+import { z } from 'zod/v4';
+
+type BasetenChatModelId = 'deepseek-ai/DeepSeek-R1-0528' | 'deepseek-ai/DeepSeek-V3-0324' | 'meta-llama/Llama-4-Maverick-17B-128E-Instruct' | 'meta-llama/Llama-4-Scout-17B-16E-Instruct' | 'moonshotai/Kimi-K2-Instruct' | 'Qwen/Qwen3-235B-A22B-Instruct-2507' | 'Qwen/Qwen3-Coder-480B-A35B-Instruct' | 'openai/gpt-oss-120b' | (string & {});
+
+type BasetenEmbeddingModelId = string & {};
+
+type BasetenErrorData = z.infer<typeof basetenErrorSchema>;
+declare const basetenErrorSchema: z.ZodObject<{
+    error: z.ZodString;
+}, 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;
+}
+interface BasetenProvider extends ProviderV2 {
+    /**
+  Creates a chat model for text generation.
+  */
+    (modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a chat model for text generation.
+  */
+    chatModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a language model for text generation. Alias for chatModel.
+  */
+    languageModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    /**
+  Creates a text embedding model for text generation.
+  */
+    textEmbeddingModel(modelId?: BasetenEmbeddingModelId): EmbeddingModelV2<string>;
+}
+declare function createBaseten(options?: BasetenProviderSettings): BasetenProvider;
+declare const baseten: BasetenProvider;
+
+declare const VERSION: string;
+
+export { type BasetenChatModelId, type BasetenErrorData, type BasetenProvider, type BasetenProviderSettings, VERSION, baseten, createBaseten };

package/dist/index.js

@@ -0,0 +1,157 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/index.ts
+var src_exports = {};
+__export(src_exports, {
+  VERSION: () => VERSION,
+  baseten: () => baseten,
+  createBaseten: () => createBaseten
+});
+module.exports = __toCommonJS(src_exports);
+
+// src/baseten-provider.ts
+var import_openai_compatible = require("@ai-sdk/openai-compatible");
+var import_provider = require("@ai-sdk/provider");
+var import_provider_utils = require("@ai-sdk/provider-utils");
+var import_v4 = require("zod/v4");
+var import_performance_client = require("@basetenlabs/performance-client");
+
+// src/version.ts
+var VERSION = true ? "0.0.1" : "0.0.0-test";
+
+// src/baseten-provider.ts
+var basetenErrorSchema = import_v4.z.object({
+  error: import_v4.z.string()
+});
+var basetenErrorStructure = {
+  errorSchema: basetenErrorSchema,
+  errorToMessage: (data) => data.error
+};
+var defaultBaseURL = "https://inference.baseten.co/v1";
+function createBaseten(options = {}) {
+  var _a;
+  const baseURL = (0, import_provider_utils.withoutTrailingSlash)((_a = options.baseURL) != null ? _a : defaultBaseURL);
+  const getHeaders = () => (0, import_provider_utils.withUserAgentSuffix)(
+    {
+      Authorization: `Bearer ${(0, import_provider_utils.loadApiKey)({
+        apiKey: options.apiKey,
+        environmentVariableName: "BASETEN_API_KEY",
+        description: "Baseten API key"
+      })}`,
+      ...options.headers
+    },
+    `ai-sdk/baseten/${VERSION}`
+  );
+  const getCommonModelConfig = (modelType, customURL) => ({
+    provider: `baseten.${modelType}`,
+    url: ({ path }) => {
+      if (modelType === "embedding" && (customURL == null ? void 0 : customURL.includes("/sync")) && !(customURL == null ? void 0 : customURL.includes("/sync/v1"))) {
+        return `${customURL}/v1${path}`;
+      }
+      return `${customURL || baseURL}${path}`;
+    },
+    headers: getHeaders,
+    fetch: options.fetch
+  });
+  const createChatModel = (modelId) => {
+    const customURL = options.modelURL;
+    if (customURL) {
+      const isOpenAICompatible = customURL.includes("/sync/v1");
+      if (isOpenAICompatible) {
+        return new import_openai_compatible.OpenAICompatibleChatLanguageModel(modelId != null ? modelId : "placeholder", {
+          ...getCommonModelConfig("chat", customURL),
+          errorStructure: basetenErrorStructure
+        });
+      } else if (customURL.includes("/predict")) {
+        throw new Error(
+          "Not supported. You must use a /sync/v1 endpoint for chat models."
+        );
+      }
+    }
+    return new import_openai_compatible.OpenAICompatibleChatLanguageModel(modelId != null ? modelId : "chat", {
+      ...getCommonModelConfig("chat"),
+      errorStructure: basetenErrorStructure
+    });
+  };
+  const createTextEmbeddingModel = (modelId) => {
+    const customURL = options.modelURL;
+    if (!customURL) {
+      throw new Error(
+        "No model URL provided for embeddings. Please set modelURL option for embeddings."
+      );
+    }
+    const isOpenAICompatible = customURL.includes("/sync");
+    if (isOpenAICompatible) {
+      const model = new import_openai_compatible.OpenAICompatibleEmbeddingModel(
+        modelId != null ? modelId : "embeddings",
+        {
+          ...getCommonModelConfig("embedding", customURL),
+          errorStructure: basetenErrorStructure
+        }
+      );
+      const performanceClientURL = customURL.replace("/sync/v1", "/sync");
+      const performanceClient = new import_performance_client.PerformanceClient(
+        performanceClientURL,
+        (0, import_provider_utils.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 != null ? modelId : "embeddings"
+          // model_id is for Model APIs, we don't use it here for dedicated
+        );
+        const embeddings = response.data.map((item) => item.embedding);
+        return {
+          embeddings,
+          usage: response.usage ? { tokens: response.usage.total_tokens } : void 0,
+          response: { headers: {}, body: response }
+        };
+      };
+      return model;
+    } else {
+      throw new Error(
+        "Not supported. You must use a /sync or /sync/v1 endpoint for embeddings."
+      );
+    }
+  };
+  const provider = (modelId) => createChatModel(modelId);
+  provider.chatModel = createChatModel;
+  provider.languageModel = createChatModel;
+  provider.imageModel = (modelId) => {
+    throw new import_provider.NoSuchModelError({ modelId, modelType: "imageModel" });
+  };
+  provider.textEmbeddingModel = createTextEmbeddingModel;
+  return provider;
+}
+var baseten = createBaseten();
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  VERSION,
+  baseten,
+  createBaseten
+});
+//# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/index.ts","../src/baseten-provider.ts","../src/version.ts"],"sourcesContent":["export type { BasetenChatModelId } from './baseten-chat-options';\nexport { baseten, createBaseten } from './baseten-provider';\nexport type {\n  BasetenProvider,\n  BasetenProviderSettings,\n  BasetenErrorData,\n} from './baseten-provider';\nexport { VERSION } from './version';\n","import {\n  OpenAICompatibleChatLanguageModel,\n  OpenAICompatibleEmbeddingModel,\n  ProviderErrorStructure,\n} from '@ai-sdk/openai-compatible';\nimport {\n  EmbeddingModelV2,\n  LanguageModelV2,\n  NoSuchModelError,\n  ProviderV2,\n} from '@ai-sdk/provider';\nimport {\n  FetchFunction,\n  loadApiKey,\n  withoutTrailingSlash,\n  withUserAgentSuffix,\n} from '@ai-sdk/provider-utils';\nimport { z } from 'zod/v4';\nimport { BasetenChatModelId } from './baseten-chat-options';\nimport { BasetenEmbeddingModelId } from './baseten-embedding-options';\nimport { PerformanceClient } from '@basetenlabs/performance-client';\nimport { VERSION } from './version';\n\nexport type BasetenErrorData = z.infer<typeof basetenErrorSchema>;\n\nconst basetenErrorSchema = z.object({\n  error: z.string(),\n});\n\nconst basetenErrorStructure: ProviderErrorStructure<BasetenErrorData> = {\n  errorSchema: basetenErrorSchema,\n  errorToMessage: data => data.error,\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\nexport interface BasetenProvider extends ProviderV2 {\n  /**\nCreates a chat model for text generation. \n*/\n  (modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a chat model for text generation. \n*/\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a language model for text generation. Alias for chatModel.\n*/\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a text embedding model for text generation.\n*/\n  textEmbeddingModel(\n    modelId?: BasetenEmbeddingModelId,\n  ): EmbeddingModelV2<string>;\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    // Use modelURL if provided, otherwise use default Model APIs\n    const customURL = options.modelURL;\n\n    if (customURL) {\n      // Check if this is a /sync/v1 endpoint (OpenAI-compatible) or /predict endpoint (custom)\n      const isOpenAICompatible = customURL.includes('/sync/v1');\n\n      if (isOpenAICompatible) {\n        // For /sync/v1 endpoints, use standard OpenAI-compatible format\n        return new OpenAICompatibleChatLanguageModel(modelId ?? 'placeholder', {\n          ...getCommonModelConfig('chat', customURL),\n          errorStructure: basetenErrorStructure,\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    // Use default OpenAI-compatible format for Model APIs\n    return new OpenAICompatibleChatLanguageModel(modelId ?? 'chat', {\n      ...getCommonModelConfig('chat'),\n      errorStructure: basetenErrorStructure,\n    });\n  };\n\n  const createTextEmbeddingModel = (modelId?: BasetenEmbeddingModelId) => {\n    // Use modelURL if provided\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    // Check if this is a /sync or /sync/v1 endpoint (OpenAI-compatible)\n    // We support both /sync and /sync/v1, stripping /v1 before passing to Performance Client, as Performance Client adds /v1 itself\n    const isOpenAICompatible = customURL.includes('/sync');\n\n    if (isOpenAICompatible) {\n      // Create the model using OpenAICompatibleEmbeddingModel and override doEmbed\n      const model = new OpenAICompatibleEmbeddingModel(\n        modelId ?? 'embeddings',\n        {\n          ...getCommonModelConfig('embedding', customURL),\n          errorStructure: basetenErrorStructure,\n        },\n      );\n\n      // Strip /v1 from URL if present before passing to Performance Client to avoid double /v1\n      const performanceClientURL = customURL.replace('/sync/v1', '/sync');\n\n      // Initialize the B10 Performance Client once for reuse\n      const performanceClient = new PerformanceClient(\n        performanceClientURL,\n        loadApiKey({\n          apiKey: options.apiKey,\n          environmentVariableName: 'BASETEN_API_KEY',\n          description: 'Baseten API key',\n        }),\n      );\n\n      // Override the doEmbed method to use the pre-created Performance Client\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        // Performance Client handles batching internally, so we don't need to limit in 128 here\n        const response = await performanceClient.embed(\n          params.values,\n          modelId ?? 'embeddings', // model_id is for Model APIs, we don't use it here for dedicated\n        );\n        // Transform the response to match the expected format\n        const embeddings = response.data.map((item: any) => item.embedding);\n\n        return {\n          embeddings: embeddings,\n          usage: response.usage\n            ? { tokens: response.usage.total_tokens }\n            : undefined,\n          response: { headers: {}, body: response },\n        };\n      };\n\n      return model;\n    } else {\n      throw new Error(\n        'Not supported. You must use a /sync or /sync/v1 endpoint for embeddings.',\n      );\n    }\n  };\n\n  const provider = (modelId?: BasetenChatModelId) => createChatModel(modelId);\n  provider.chatModel = createChatModel;\n  provider.languageModel = createChatModel;\n  provider.imageModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });\n  };\n  provider.textEmbeddingModel = createTextEmbeddingModel;\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;AAAA;AAAA;AAAA;AAAA;AAAA;;;ACAA,+BAIO;AACP,sBAKO;AACP,4BAKO;AACP,gBAAkB;AAGlB,gCAAkC;;;AClB3B,IAAM,UACX,OACI,UACA;;;ADoBN,IAAM,qBAAqB,YAAE,OAAO;AAAA,EAClC,OAAO,YAAE,OAAO;AAClB,CAAC;AAED,IAAM,wBAAkE;AAAA,EACtE,aAAa;AAAA,EACb,gBAAgB,UAAQ,KAAK;AAC/B;AAwDA,IAAM,iBAAiB;AAEhB,SAAS,cACd,UAAmC,CAAC,GACnB;AA5FnB;AA6FE,QAAM,cAAU,6CAAqB,aAAQ,YAAR,YAAmB,cAAc;AACtE,QAAM,aAAa,UACjB;AAAA,IACE;AAAA,MACE,eAAe,cAAU,kCAAW;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,gBACd,uCAAW,SAAS,aACpB,EAAC,uCAAW,SAAS,cACrB;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;AAExD,UAAM,YAAY,QAAQ;AAE1B,QAAI,WAAW;AAEb,YAAM,qBAAqB,UAAU,SAAS,UAAU;AAExD,UAAI,oBAAoB;AAEtB,eAAO,IAAI,2DAAkC,4BAAW,eAAe;AAAA,UACrE,GAAG,qBAAqB,QAAQ,SAAS;AAAA,UACzC,gBAAgB;AAAA,QAClB,CAAC;AAAA,MACH,WAAW,UAAU,SAAS,UAAU,GAAG;AACzC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,WAAO,IAAI,2DAAkC,4BAAW,QAAQ;AAAA,MAC9D,GAAG,qBAAqB,MAAM;AAAA,MAC9B,gBAAgB;AAAA,IAClB,CAAC;AAAA,EACH;AAEA,QAAM,2BAA2B,CAAC,YAAsC;AAEtE,UAAM,YAAY,QAAQ;AAC1B,QAAI,CAAC,WAAW;AACd,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAIA,UAAM,qBAAqB,UAAU,SAAS,OAAO;AAErD,QAAI,oBAAoB;AAEtB,YAAM,QAAQ,IAAI;AAAA,QAChB,4BAAW;AAAA,QACX;AAAA,UACE,GAAG,qBAAqB,aAAa,SAAS;AAAA,UAC9C,gBAAgB;AAAA,QAClB;AAAA,MACF;AAGA,YAAM,uBAAuB,UAAU,QAAQ,YAAY,OAAO;AAGlE,YAAM,oBAAoB,IAAI;AAAA,QAC5B;AAAA,YACA,kCAAW;AAAA,UACT,QAAQ,QAAQ;AAAA,UAChB,yBAAyB;AAAA,UACzB,aAAa;AAAA,QACf,CAAC;AAAA,MACH;AAGA,YAAM,UAAU,OAAM,WAAU;AAC9B,YAAI,CAAC,OAAO,UAAU,CAAC,MAAM,QAAQ,OAAO,MAAM,GAAG;AACnD,gBAAM,IAAI,MAAM,2CAA2C;AAAA,QAC7D;AAGA,cAAM,WAAW,MAAM,kBAAkB;AAAA,UACvC,OAAO;AAAA,UACP,4BAAW;AAAA;AAAA,QACb;AAEA,cAAM,aAAa,SAAS,KAAK,IAAI,CAAC,SAAc,KAAK,SAAS;AAElE,eAAO;AAAA,UACL;AAAA,UACA,OAAO,SAAS,QACZ,EAAE,QAAQ,SAAS,MAAM,aAAa,IACtC;AAAA,UACJ,UAAU,EAAE,SAAS,CAAC,GAAG,MAAM,SAAS;AAAA,QAC1C;AAAA,MACF;AAEA,aAAO;AAAA,IACT,OAAO;AACL,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,QAAM,WAAW,CAAC,YAAiC,gBAAgB,OAAO;AAC1E,WAAS,YAAY;AACrB,WAAS,gBAAgB;AACzB,WAAS,aAAa,CAAC,YAAoB;AACzC,UAAM,IAAI,iCAAiB,EAAE,SAAS,WAAW,aAAa,CAAC;AAAA,EACjE;AACA,WAAS,qBAAqB;AAC9B,SAAO;AACT;AAEO,IAAM,UAAU,cAAc;","names":[]}

package/dist/index.mjs

@@ -0,0 +1,137 @@
+// 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 { z } from "zod/v4";
+import { PerformanceClient } from "@basetenlabs/performance-client";
+
+// src/version.ts
+var VERSION = true ? "0.0.1" : "0.0.0-test";
+
+// src/baseten-provider.ts
+var basetenErrorSchema = z.object({
+  error: z.string()
+});
+var basetenErrorStructure = {
+  errorSchema: basetenErrorSchema,
+  errorToMessage: (data) => data.error
+};
+var defaultBaseURL = "https://inference.baseten.co/v1";
+function createBaseten(options = {}) {
+  var _a;
+  const baseURL = withoutTrailingSlash((_a = options.baseURL) != null ? _a : 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.${modelType}`,
+    url: ({ path }) => {
+      if (modelType === "embedding" && (customURL == null ? void 0 : customURL.includes("/sync")) && !(customURL == null ? void 0 : customURL.includes("/sync/v1"))) {
+        return `${customURL}/v1${path}`;
+      }
+      return `${customURL || baseURL}${path}`;
+    },
+    headers: getHeaders,
+    fetch: options.fetch
+  });
+  const createChatModel = (modelId) => {
+    const customURL = options.modelURL;
+    if (customURL) {
+      const isOpenAICompatible = customURL.includes("/sync/v1");
+      if (isOpenAICompatible) {
+        return new OpenAICompatibleChatLanguageModel(modelId != null ? modelId : "placeholder", {
+          ...getCommonModelConfig("chat", customURL),
+          errorStructure: basetenErrorStructure
+        });
+      } else if (customURL.includes("/predict")) {
+        throw new Error(
+          "Not supported. You must use a /sync/v1 endpoint for chat models."
+        );
+      }
+    }
+    return new OpenAICompatibleChatLanguageModel(modelId != null ? modelId : "chat", {
+      ...getCommonModelConfig("chat"),
+      errorStructure: basetenErrorStructure
+    });
+  };
+  const createTextEmbeddingModel = (modelId) => {
+    const customURL = options.modelURL;
+    if (!customURL) {
+      throw new Error(
+        "No model URL provided for embeddings. Please set modelURL option for embeddings."
+      );
+    }
+    const isOpenAICompatible = customURL.includes("/sync");
+    if (isOpenAICompatible) {
+      const model = new OpenAICompatibleEmbeddingModel(
+        modelId != null ? modelId : "embeddings",
+        {
+          ...getCommonModelConfig("embedding", customURL),
+          errorStructure: basetenErrorStructure
+        }
+      );
+      const performanceClientURL = customURL.replace("/sync/v1", "/sync");
+      const performanceClient = new PerformanceClient(
+        performanceClientURL,
+        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 != null ? modelId : "embeddings"
+          // model_id is for Model APIs, we don't use it here for dedicated
+        );
+        const embeddings = response.data.map((item) => item.embedding);
+        return {
+          embeddings,
+          usage: response.usage ? { tokens: response.usage.total_tokens } : void 0,
+          response: { headers: {}, body: response }
+        };
+      };
+      return model;
+    } else {
+      throw new Error(
+        "Not supported. You must use a /sync or /sync/v1 endpoint for embeddings."
+      );
+    }
+  };
+  const provider = (modelId) => createChatModel(modelId);
+  provider.chatModel = createChatModel;
+  provider.languageModel = createChatModel;
+  provider.imageModel = (modelId) => {
+    throw new NoSuchModelError({ modelId, modelType: "imageModel" });
+  };
+  provider.textEmbeddingModel = createTextEmbeddingModel;
+  return provider;
+}
+var baseten = createBaseten();
+export {
+  VERSION,
+  baseten,
+  createBaseten
+};
+//# sourceMappingURL=index.mjs.map

package/dist/index.mjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/baseten-provider.ts","../src/version.ts"],"sourcesContent":["import {\n  OpenAICompatibleChatLanguageModel,\n  OpenAICompatibleEmbeddingModel,\n  ProviderErrorStructure,\n} from '@ai-sdk/openai-compatible';\nimport {\n  EmbeddingModelV2,\n  LanguageModelV2,\n  NoSuchModelError,\n  ProviderV2,\n} from '@ai-sdk/provider';\nimport {\n  FetchFunction,\n  loadApiKey,\n  withoutTrailingSlash,\n  withUserAgentSuffix,\n} from '@ai-sdk/provider-utils';\nimport { z } from 'zod/v4';\nimport { BasetenChatModelId } from './baseten-chat-options';\nimport { BasetenEmbeddingModelId } from './baseten-embedding-options';\nimport { PerformanceClient } from '@basetenlabs/performance-client';\nimport { VERSION } from './version';\n\nexport type BasetenErrorData = z.infer<typeof basetenErrorSchema>;\n\nconst basetenErrorSchema = z.object({\n  error: z.string(),\n});\n\nconst basetenErrorStructure: ProviderErrorStructure<BasetenErrorData> = {\n  errorSchema: basetenErrorSchema,\n  errorToMessage: data => data.error,\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\nexport interface BasetenProvider extends ProviderV2 {\n  /**\nCreates a chat model for text generation. \n*/\n  (modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a chat model for text generation. \n*/\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a language model for text generation. Alias for chatModel.\n*/\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV2;\n\n  /**\nCreates a text embedding model for text generation.\n*/\n  textEmbeddingModel(\n    modelId?: BasetenEmbeddingModelId,\n  ): EmbeddingModelV2<string>;\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    // Use modelURL if provided, otherwise use default Model APIs\n    const customURL = options.modelURL;\n\n    if (customURL) {\n      // Check if this is a /sync/v1 endpoint (OpenAI-compatible) or /predict endpoint (custom)\n      const isOpenAICompatible = customURL.includes('/sync/v1');\n\n      if (isOpenAICompatible) {\n        // For /sync/v1 endpoints, use standard OpenAI-compatible format\n        return new OpenAICompatibleChatLanguageModel(modelId ?? 'placeholder', {\n          ...getCommonModelConfig('chat', customURL),\n          errorStructure: basetenErrorStructure,\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    // Use default OpenAI-compatible format for Model APIs\n    return new OpenAICompatibleChatLanguageModel(modelId ?? 'chat', {\n      ...getCommonModelConfig('chat'),\n      errorStructure: basetenErrorStructure,\n    });\n  };\n\n  const createTextEmbeddingModel = (modelId?: BasetenEmbeddingModelId) => {\n    // Use modelURL if provided\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    // Check if this is a /sync or /sync/v1 endpoint (OpenAI-compatible)\n    // We support both /sync and /sync/v1, stripping /v1 before passing to Performance Client, as Performance Client adds /v1 itself\n    const isOpenAICompatible = customURL.includes('/sync');\n\n    if (isOpenAICompatible) {\n      // Create the model using OpenAICompatibleEmbeddingModel and override doEmbed\n      const model = new OpenAICompatibleEmbeddingModel(\n        modelId ?? 'embeddings',\n        {\n          ...getCommonModelConfig('embedding', customURL),\n          errorStructure: basetenErrorStructure,\n        },\n      );\n\n      // Strip /v1 from URL if present before passing to Performance Client to avoid double /v1\n      const performanceClientURL = customURL.replace('/sync/v1', '/sync');\n\n      // Initialize the B10 Performance Client once for reuse\n      const performanceClient = new PerformanceClient(\n        performanceClientURL,\n        loadApiKey({\n          apiKey: options.apiKey,\n          environmentVariableName: 'BASETEN_API_KEY',\n          description: 'Baseten API key',\n        }),\n      );\n\n      // Override the doEmbed method to use the pre-created Performance Client\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        // Performance Client handles batching internally, so we don't need to limit in 128 here\n        const response = await performanceClient.embed(\n          params.values,\n          modelId ?? 'embeddings', // model_id is for Model APIs, we don't use it here for dedicated\n        );\n        // Transform the response to match the expected format\n        const embeddings = response.data.map((item: any) => item.embedding);\n\n        return {\n          embeddings: embeddings,\n          usage: response.usage\n            ? { tokens: response.usage.total_tokens }\n            : undefined,\n          response: { headers: {}, body: response },\n        };\n      };\n\n      return model;\n    } else {\n      throw new Error(\n        'Not supported. You must use a /sync or /sync/v1 endpoint for embeddings.',\n      );\n    }\n  };\n\n  const provider = (modelId?: BasetenChatModelId) => createChatModel(modelId);\n  provider.chatModel = createChatModel;\n  provider.languageModel = createChatModel;\n  provider.imageModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });\n  };\n  provider.textEmbeddingModel = createTextEmbeddingModel;\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,EAGE;AAAA,OAEK;AACP;AAAA,EAEE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,SAAS;AAGlB,SAAS,yBAAyB;;;AClB3B,IAAM,UACX,OACI,UACA;;;ADoBN,IAAM,qBAAqB,EAAE,OAAO;AAAA,EAClC,OAAO,EAAE,OAAO;AAClB,CAAC;AAED,IAAM,wBAAkE;AAAA,EACtE,aAAa;AAAA,EACb,gBAAgB,UAAQ,KAAK;AAC/B;AAwDA,IAAM,iBAAiB;AAEhB,SAAS,cACd,UAAmC,CAAC,GACnB;AA5FnB;AA6FE,QAAM,UAAU,sBAAqB,aAAQ,YAAR,YAAmB,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,gBACd,uCAAW,SAAS,aACpB,EAAC,uCAAW,SAAS,cACrB;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;AAExD,UAAM,YAAY,QAAQ;AAE1B,QAAI,WAAW;AAEb,YAAM,qBAAqB,UAAU,SAAS,UAAU;AAExD,UAAI,oBAAoB;AAEtB,eAAO,IAAI,kCAAkC,4BAAW,eAAe;AAAA,UACrE,GAAG,qBAAqB,QAAQ,SAAS;AAAA,UACzC,gBAAgB;AAAA,QAClB,CAAC;AAAA,MACH,WAAW,UAAU,SAAS,UAAU,GAAG;AACzC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAGA,WAAO,IAAI,kCAAkC,4BAAW,QAAQ;AAAA,MAC9D,GAAG,qBAAqB,MAAM;AAAA,MAC9B,gBAAgB;AAAA,IAClB,CAAC;AAAA,EACH;AAEA,QAAM,2BAA2B,CAAC,YAAsC;AAEtE,UAAM,YAAY,QAAQ;AAC1B,QAAI,CAAC,WAAW;AACd,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAIA,UAAM,qBAAqB,UAAU,SAAS,OAAO;AAErD,QAAI,oBAAoB;AAEtB,YAAM,QAAQ,IAAI;AAAA,QAChB,4BAAW;AAAA,QACX;AAAA,UACE,GAAG,qBAAqB,aAAa,SAAS;AAAA,UAC9C,gBAAgB;AAAA,QAClB;AAAA,MACF;AAGA,YAAM,uBAAuB,UAAU,QAAQ,YAAY,OAAO;AAGlE,YAAM,oBAAoB,IAAI;AAAA,QAC5B;AAAA,QACA,WAAW;AAAA,UACT,QAAQ,QAAQ;AAAA,UAChB,yBAAyB;AAAA,UACzB,aAAa;AAAA,QACf,CAAC;AAAA,MACH;AAGA,YAAM,UAAU,OAAM,WAAU;AAC9B,YAAI,CAAC,OAAO,UAAU,CAAC,MAAM,QAAQ,OAAO,MAAM,GAAG;AACnD,gBAAM,IAAI,MAAM,2CAA2C;AAAA,QAC7D;AAGA,cAAM,WAAW,MAAM,kBAAkB;AAAA,UACvC,OAAO;AAAA,UACP,4BAAW;AAAA;AAAA,QACb;AAEA,cAAM,aAAa,SAAS,KAAK,IAAI,CAAC,SAAc,KAAK,SAAS;AAElE,eAAO;AAAA,UACL;AAAA,UACA,OAAO,SAAS,QACZ,EAAE,QAAQ,SAAS,MAAM,aAAa,IACtC;AAAA,UACJ,UAAU,EAAE,SAAS,CAAC,GAAG,MAAM,SAAS;AAAA,QAC1C;AAAA,MACF;AAEA,aAAO;AAAA,IACT,OAAO;AACL,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,QAAM,WAAW,CAAC,YAAiC,gBAAgB,OAAO;AAC1E,WAAS,YAAY;AACrB,WAAS,gBAAgB;AACzB,WAAS,aAAa,CAAC,YAAoB;AACzC,UAAM,IAAI,iBAAiB,EAAE,SAAS,WAAW,aAAa,CAAC;AAAA,EACjE;AACA,WAAS,qBAAqB;AAC9B,SAAO;AACT;AAEO,IAAM,UAAU,cAAc;","names":[]}

package/package.json

@@ -0,0 +1,67 @@
+{
+  "name": "@ai-sdk/baseten",
+  "version": "0.0.1",
+  "license": "Apache-2.0",
+  "sideEffects": false,
+  "main": "./dist/index.js",
+  "module": "./dist/index.mjs",
+  "types": "./dist/index.d.ts",
+  "files": [
+    "dist/**/*",
+    "CHANGELOG.md"
+  ],
+  "exports": {
+    "./package.json": "./package.json",
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.mjs",
+      "require": "./dist/index.js"
+    }
+  },
+  "dependencies": {
+    "@basetenlabs/performance-client": "^0.0.10",
+    "@ai-sdk/openai-compatible": "1.0.19",
+    "@ai-sdk/provider": "2.0.0",
+    "@ai-sdk/provider-utils": "3.0.10"
+  },
+  "devDependencies": {
+    "@types/node": "20.17.24",
+    "tsup": "^8",
+    "typescript": "5.8.3",
+    "zod": "3.25.76",
+    "@vercel/ai-tsconfig": "0.0.0"
+  },
+  "peerDependencies": {
+    "zod": "^3.25.76 || ^4.1.8"
+  },
+  "engines": {
+    "node": ">=18"
+  },
+  "publishConfig": {
+    "access": "public"
+  },
+  "homepage": "https://ai-sdk.dev/docs",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/vercel/ai.git"
+  },
+  "bugs": {
+    "url": "https://github.com/vercel/ai/issues"
+  },
+  "keywords": [
+    "ai"
+  ],
+  "scripts": {
+    "build": "pnpm clean && tsup --tsconfig tsconfig.build.json",
+    "build:watch": "pnpm clean && tsup --watch",
+    "clean": "rm -rf dist *.tsbuildinfo",
+    "lint": "eslint \"./**/*.ts*\"",
+    "type-check": "tsc --build",
+    "prettier-check": "prettier --check \"./**/*.ts*\"",
+    "test": "pnpm test:node && pnpm test:edge",
+    "test:update": "pnpm test:node -u",
+    "test:watch": "vitest --config vitest.node.config.js",
+    "test:edge": "vitest --config vitest.edge.config.js --run",
+    "test:node": "vitest --config vitest.node.config.js --run"
+  }
+}