@ai-sdk/baseten 1.0.0-beta.4 → 1.0.0-beta.6

package/CHANGELOG.md

@@ -1,5 +1,22 @@
 # @ai-sdk/baseten
 
+## 1.0.0-beta.6
+
+### Patch Changes
+
+- 916bc46: bumped performance client to 0.0.10
+
+## 1.0.0-beta.5
+
+### Patch Changes
+
+- 8dac895: feat: `LanguageModelV3`
+- Updated dependencies [8dac895]
+- Updated dependencies [10c1322]
+  - @ai-sdk/openai-compatible@1.1.0-beta.5
+  - @ai-sdk/provider-utils@3.1.0-beta.5
+  - @ai-sdk/provider@2.1.0-beta.3
+
 ## 1.0.0-beta.4
 
 ### Patch Changes

package/dist/index.d.mts

@@ -1,4 +1,4 @@
-import { ProviderV3, LanguageModelV2, EmbeddingModelV3 } from '@ai-sdk/provider';
+import { ProviderV3, LanguageModelV3, EmbeddingModelV3 } from '@ai-sdk/provider';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
 
@@ -39,15 +39,15 @@ interface BasetenProvider extends ProviderV3 {
     /**
   Creates a chat model for text generation.
   */
-    (modelId?: BasetenChatModelId): LanguageModelV2;
+    (modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a chat model for text generation.
   */
-    chatModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    chatModel(modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a language model for text generation. Alias for chatModel.
   */
-    languageModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    languageModel(modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a text embedding model for text generation.
   */

package/dist/index.d.ts

@@ -1,4 +1,4 @@
-import { ProviderV3, LanguageModelV2, EmbeddingModelV3 } from '@ai-sdk/provider';
+import { ProviderV3, LanguageModelV3, EmbeddingModelV3 } from '@ai-sdk/provider';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
 
@@ -39,15 +39,15 @@ interface BasetenProvider extends ProviderV3 {
     /**
   Creates a chat model for text generation.
   */
-    (modelId?: BasetenChatModelId): LanguageModelV2;
+    (modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a chat model for text generation.
   */
-    chatModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    chatModel(modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a language model for text generation. Alias for chatModel.
   */
-    languageModel(modelId?: BasetenChatModelId): LanguageModelV2;
+    languageModel(modelId?: BasetenChatModelId): LanguageModelV3;
     /**
   Creates a text embedding model for text generation.
   */

package/dist/index.js

@@ -34,7 +34,7 @@ var import_v4 = require("zod/v4");
 var import_performance_client = require("@basetenlabs/performance-client");
 
 // src/version.ts
-var VERSION = true ? "1.0.0-beta.4" : "0.0.0-test";
+var VERSION = true ? "1.0.0-beta.6" : "0.0.0-test";
 
 // src/baseten-provider.ts
 var basetenErrorSchema = import_v4.z.object({

package/dist/index.js.map

@@ -1 +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  EmbeddingModelV3,\n  LanguageModelV2,\n  NoSuchModelError,\n  ProviderV3,\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 ProviderV3 {\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  ): EmbeddingModelV3<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,iBACA;;;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":[]}
+{"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  EmbeddingModelV3,\n  LanguageModelV3,\n  NoSuchModelError,\n  ProviderV3,\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 ProviderV3 {\n  /**\nCreates a chat model for text generation. \n*/\n  (modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a chat model for text generation. \n*/\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a language model for text generation. Alias for chatModel.\n*/\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a text embedding model for text generation.\n*/\n  textEmbeddingModel(\n    modelId?: BasetenEmbeddingModelId,\n  ): EmbeddingModelV3<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,iBACA;;;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

@@ -15,7 +15,7 @@ import { z } from "zod/v4";
 import { PerformanceClient } from "@basetenlabs/performance-client";
 
 // src/version.ts
-var VERSION = true ? "1.0.0-beta.4" : "0.0.0-test";
+var VERSION = true ? "1.0.0-beta.6" : "0.0.0-test";
 
 // src/baseten-provider.ts
 var basetenErrorSchema = z.object({

package/dist/index.mjs.map

@@ -1 +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  EmbeddingModelV3,\n  LanguageModelV2,\n  NoSuchModelError,\n  ProviderV3,\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 ProviderV3 {\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  ): EmbeddingModelV3<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,iBACA;;;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":[]}
+{"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  EmbeddingModelV3,\n  LanguageModelV3,\n  NoSuchModelError,\n  ProviderV3,\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 ProviderV3 {\n  /**\nCreates a chat model for text generation. \n*/\n  (modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a chat model for text generation. \n*/\n  chatModel(modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a language model for text generation. Alias for chatModel.\n*/\n  languageModel(modelId?: BasetenChatModelId): LanguageModelV3;\n\n  /**\nCreates a text embedding model for text generation.\n*/\n  textEmbeddingModel(\n    modelId?: BasetenEmbeddingModelId,\n  ): EmbeddingModelV3<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,iBACA;;;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

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/baseten",
-  "version": "1.0.0-beta.4",
+  "version": "1.0.0-beta.6",
   "license": "Apache-2.0",
   "sideEffects": false,
   "main": "./dist/index.js",
@@ -19,10 +19,10 @@
     }
   },
   "dependencies": {
-    "@basetenlabs/performance-client": "^0.0.8",
-    "@ai-sdk/openai-compatible": "1.1.0-beta.4",
-    "@ai-sdk/provider": "2.1.0-beta.2",
-    "@ai-sdk/provider-utils": "3.1.0-beta.4"
+    "@basetenlabs/performance-client": "^0.0.10",
+    "@ai-sdk/openai-compatible": "1.1.0-beta.5",
+    "@ai-sdk/provider": "2.1.0-beta.3",
+    "@ai-sdk/provider-utils": "3.1.0-beta.5"
   },
   "devDependencies": {
     "@types/node": "20.17.24",