@ai-sdk/zai 0.0.0 → 3.0.1

package/CHANGELOG.md

@@ -0,0 +1,19 @@
+# @ai-sdk/zai
+
+## 3.0.1
+
+### Patch Changes
+
+- Updated dependencies [3e125ba]
+  - @ai-sdk/provider-utils@5.0.32
+  - @ai-sdk/openai-compatible@3.0.39
+
+## 3.0.0
+
+### Major Changes
+
+- b580642: feat(zai): add the Z.AI provider with GLM chat completions, streaming, reasoning, tools, and multimodal inputs
+
+### Patch Changes
+
+- 81bebaf: Add GLM-5.3-Flash model support to the Z.AI provider and AI Gateway.

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

@@ -1,5 +1,47 @@
-# @ai-sdk/zai
+# AI SDK - Z.AI Provider
 
-AI SDK provider for Z.ai.
+The **Z.AI provider** for the [AI SDK](https://ai-sdk.dev/docs) contains language model support for [Z.AI](https://z.ai/) GLM models.
 
-The implementation is maintained in the `vercel/ai` repository.
+> **Deploying to Vercel?** With Vercel's AI Gateway you can access Z.AI (and hundreds of models from other providers) without installing an additional provider package. [Get started with AI Gateway](https://vercel.com/ai-gateway).
+
+## Setup
+
+Install the Z.AI provider with:
+
+```bash
+npm i @ai-sdk/zai
+```
+
+## Provider Instance
+
+Import the default provider instance from `@ai-sdk/zai`:
+
+```ts
+import { zai } from '@ai-sdk/zai';
+```
+
+The provider reads the API key from `ZAI_API_KEY` by default. To configure it explicitly, use `createZai`:
+
+```ts
+import { createZai } from '@ai-sdk/zai';
+
+const zai = createZai({
+  apiKey: process.env.ZAI_API_KEY,
+});
+```
+
+## Language Models
+
+```ts
+import { zai } from '@ai-sdk/zai';
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: zai('glm-5.3'),
+  prompt: 'Explain why the sky is blue.',
+});
+
+console.log(text);
+```
+
+The provider supports streaming, reasoning, function tools, JSON object output, and URL-based image and video input on compatible GLM models.

package/dist/index.d.ts

@@ -0,0 +1,88 @@
+import { z } from 'zod/v4';
+import { ProviderV4, LanguageModelV4 } from '@ai-sdk/provider';
+import { FetchFunction } from '@ai-sdk/provider-utils';
+
+declare const zaiLanguageModelChatOptions: z.ZodObject<{
+    doSample: z.ZodOptional<z.ZodBoolean>;
+    thinking: z.ZodOptional<z.ZodObject<{
+        type: z.ZodOptional<z.ZodEnum<{
+            enabled: "enabled";
+            disabled: "disabled";
+        }>>;
+        clearThinking: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    reasoningEffort: z.ZodOptional<z.ZodEnum<{
+        none: "none";
+        minimal: "minimal";
+        low: "low";
+        medium: "medium";
+        high: "high";
+        xhigh: "xhigh";
+        max: "max";
+    }>>;
+    toolStream: z.ZodOptional<z.ZodBoolean>;
+    requestId: z.ZodOptional<z.ZodString>;
+    userId: z.ZodOptional<z.ZodString>;
+}, z.core.$strip>;
+type ZaiLanguageModelChatOptions = z.infer<typeof zaiLanguageModelChatOptions>;
+
+/**
+ * Z.AI chat model ids from the official OpenAPI 1.0.0 specification,
+ * retrieved from https://docs.z.ai/openapi.json on 2026-08-26.
+ */
+type ZaiChatModelId = 'glm-5.3' | 'glm-5.2' | 'glm-5.1' | 'glm-5-turbo' | 'glm-5' | 'glm-4.7' | 'glm-4.7-flash' | 'glm-4.7-flashx' | 'glm-4.6' | 'glm-4.5' | 'glm-4.5-air' | 'glm-4.5-x' | 'glm-4.5-airx' | 'glm-4.5-flash' | 'glm-4-32b-0414-128k' | 'glm-5.3-flash' | 'glm-5v-turbo' | 'glm-4.6v' | 'glm-4.6v-flash' | 'glm-4.6v-flashx' | 'glm-4.5v' | 'autoglm-phone-multilingual' | (string & {});
+
+declare const zaiErrorSchema: z.ZodUnion<readonly [z.ZodObject<{
+    code: z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
+    message: z.ZodString;
+}, z.core.$strip>, z.ZodObject<{
+    error: z.ZodObject<{
+        code: z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
+        message: z.ZodString;
+    }, z.core.$strip>;
+}, z.core.$strip>]>;
+type ZaiErrorData = z.infer<typeof zaiErrorSchema>;
+
+interface ZaiProviderSettings {
+    /**
+     * Z.AI API key. Defaults to the `ZAI_API_KEY` environment variable.
+     */
+    apiKey?: string;
+    /**
+     * Base URL for API calls. Defaults to
+     * `https://api.z.ai/api/paas/v4`.
+     */
+    baseURL?: string;
+    /**
+     * Custom headers to include in requests.
+     */
+    headers?: Record<string, string>;
+    /**
+     * Custom fetch implementation.
+     */
+    fetch?: FetchFunction;
+}
+interface ZaiProvider extends ProviderV4 {
+    /**
+     * Creates a Z.AI chat model for text generation.
+     */
+    (modelId: ZaiChatModelId): LanguageModelV4;
+    /**
+     * Creates a Z.AI language model.
+     */
+    languageModel(modelId: ZaiChatModelId): LanguageModelV4;
+    /**
+     * Creates a Z.AI chat model.
+     */
+    chat(modelId: ZaiChatModelId): LanguageModelV4;
+    /**
+     * @deprecated Use `embeddingModel` instead.
+     */
+    textEmbeddingModel(modelId: string): never;
+}
+declare function createZai(options?: ZaiProviderSettings): ZaiProvider;
+declare const zai: ZaiProvider;
+
+declare const VERSION: string;
+
+export { VERSION, type ZaiChatModelId, type ZaiErrorData, type ZaiLanguageModelChatOptions, type ZaiProvider, type ZaiProviderSettings, createZai, zai };

package/dist/index.js

@@ -0,0 +1,265 @@
+// src/zai-provider.ts
+import {
+  NoSuchModelError
+} from "@ai-sdk/provider";
+import {
+  loadApiKey,
+  withoutTrailingSlash,
+  withUserAgentSuffix
+} from "@ai-sdk/provider-utils";
+
+// src/version.ts
+var VERSION = true ? "3.0.1" : "0.0.0-test";
+
+// src/zai-chat-language-model.ts
+import { OpenAICompatibleChatLanguageModel } from "@ai-sdk/openai-compatible";
+import {
+  parseProviderOptions,
+  serializeModelOptions,
+  WORKFLOW_DESERIALIZE,
+  WORKFLOW_SERIALIZE
+} from "@ai-sdk/provider-utils";
+
+// src/zai-chat-language-model-options.ts
+import { z } from "zod/v4";
+var zaiLanguageModelChatOptions = z.object({
+  /**
+   * Enables or disables sampling. When disabled, temperature and topP do not
+   * take effect.
+   */
+  doSample: z.boolean().optional(),
+  /**
+   * Controls model thinking and whether reasoning from earlier turns is kept.
+   */
+  thinking: z.object({
+    type: z.enum(["enabled", "disabled"]).optional(),
+    clearThinking: z.boolean().optional()
+  }).optional(),
+  /**
+   * Controls reasoning effort for GLM-5.2 and later models.
+   */
+  reasoningEffort: z.enum(["none", "minimal", "low", "medium", "high", "xhigh", "max"]).optional(),
+  /**
+   * Enables incremental function-call argument streaming on supported models.
+   */
+  toolStream: z.boolean().optional(),
+  /**
+   * A caller-provided request identifier between 6 and 64 characters.
+   */
+  requestId: z.string().min(6).max(64).optional(),
+  /**
+   * A non-sensitive end-user identifier between 6 and 128 characters.
+   */
+  userId: z.string().min(6).max(128).optional()
+});
+
+// src/zai-error.ts
+import { z as z2 } from "zod/v4";
+var zaiErrorDetailsSchema = z2.object({
+  code: z2.union([z2.number(), z2.string()]).nullish(),
+  message: z2.string()
+});
+var zaiErrorSchema = z2.union([
+  zaiErrorDetailsSchema,
+  z2.object({ error: zaiErrorDetailsSchema })
+]);
+var zaiErrorStructure = {
+  errorSchema: zaiErrorSchema,
+  errorToMessage: (data) => "error" in data ? data.error.message : data.message
+};
+
+// src/zai-chat-language-model.ts
+function transformZaiRequestBody(args) {
+  const {
+    doSample,
+    frequency_penalty: _frequencyPenalty,
+    presence_penalty: _presencePenalty,
+    requestId,
+    seed: _seed,
+    thinking,
+    toolStream,
+    user: _user,
+    userId,
+    verbosity: _verbosity,
+    ...restArgs
+  } = args;
+  return {
+    ...restArgs,
+    ...doSample !== void 0 && { do_sample: doSample },
+    ...thinking !== void 0 && {
+      thinking: {
+        ...thinking.type !== void 0 && { type: thinking.type },
+        ...thinking.clearThinking !== void 0 && {
+          clear_thinking: thinking.clearThinking
+        }
+      }
+    },
+    ...toolStream !== void 0 && { tool_stream: toolStream },
+    ...requestId !== void 0 && { request_id: requestId },
+    ...userId !== void 0 && { user_id: userId }
+  };
+}
+function mapZaiFinishReason(finishReason) {
+  switch (finishReason.raw) {
+    case "sensitive":
+      return { unified: "content-filter", raw: finishReason.raw };
+    case "model_context_window_exceeded":
+      return { unified: "length", raw: finishReason.raw };
+    case "network_error":
+      return { unified: "error", raw: finishReason.raw };
+    default:
+      return finishReason;
+  }
+}
+var ZaiChatLanguageModel = class _ZaiChatLanguageModel extends OpenAICompatibleChatLanguageModel {
+  static [WORKFLOW_SERIALIZE](model) {
+    return serializeModelOptions({
+      modelId: model.modelId,
+      config: model.zaiConfig
+    });
+  }
+  static [WORKFLOW_DESERIALIZE](options) {
+    return new _ZaiChatLanguageModel(
+      options.modelId,
+      options.config
+    );
+  }
+  constructor(modelId, config) {
+    const headers = config.headers;
+    super(modelId, {
+      provider: config.provider,
+      url: ({ path }) => `${config.baseURL}${path}`,
+      headers: headers == null ? void 0 : () => typeof headers === "function" ? headers() : headers,
+      fetch: config.fetch,
+      errorStructure: zaiErrorStructure,
+      transformRequestBody: transformZaiRequestBody,
+      supportedUrls: () => ({
+        "image/*": [/^https?:\/\//],
+        "video/*": [/^https?:\/\//]
+      })
+    });
+    this.zaiConfig = config;
+  }
+  async prepareCallOptions(options) {
+    const warnings = [];
+    const zaiOptions = await parseProviderOptions({
+      provider: "zai",
+      providerOptions: options.providerOptions,
+      schema: zaiLanguageModelChatOptions
+    });
+    if (options.frequencyPenalty != null) {
+      warnings.push({ type: "unsupported", feature: "frequencyPenalty" });
+    }
+    if (options.presencePenalty != null) {
+      warnings.push({ type: "unsupported", feature: "presencePenalty" });
+    }
+    if (options.seed != null) {
+      warnings.push({ type: "unsupported", feature: "seed" });
+    }
+    let tools = options.tools;
+    let toolChoice = options.toolChoice;
+    if ((toolChoice == null ? void 0 : toolChoice.type) === "none") {
+      tools = void 0;
+      toolChoice = void 0;
+    } else if (toolChoice != null && toolChoice.type !== "auto") {
+      warnings.push({
+        type: "unsupported",
+        feature: `toolChoice ${toolChoice.type}`,
+        details: "Z.AI currently supports only automatic tool selection."
+      });
+      toolChoice = void 0;
+    }
+    const normalizedOptions = {
+      ...options,
+      frequencyPenalty: void 0,
+      presencePenalty: void 0,
+      seed: void 0,
+      tools,
+      toolChoice,
+      providerOptions: zaiOptions == null ? options.providerOptions : {
+        ...options.providerOptions,
+        zai: zaiOptions
+      }
+    };
+    return { normalizedOptions, warnings };
+  }
+  async doGenerate(options) {
+    const { normalizedOptions, warnings } = await this.prepareCallOptions(options);
+    const result = await super.doGenerate(normalizedOptions);
+    return {
+      ...result,
+      finishReason: mapZaiFinishReason(result.finishReason),
+      warnings: [...result.warnings, ...warnings]
+    };
+  }
+  async doStream(options) {
+    const { normalizedOptions, warnings } = await this.prepareCallOptions(options);
+    const result = await super.doStream(normalizedOptions);
+    return {
+      ...result,
+      stream: result.stream.pipeThrough(
+        new TransformStream({
+          transform(part, controller) {
+            if (part.type === "stream-start") {
+              controller.enqueue({
+                ...part,
+                warnings: [...part.warnings, ...warnings]
+              });
+              return;
+            }
+            if (part.type === "finish") {
+              controller.enqueue({
+                ...part,
+                finishReason: mapZaiFinishReason(part.finishReason)
+              });
+              return;
+            }
+            controller.enqueue(part);
+          }
+        })
+      )
+    };
+  }
+};
+
+// src/zai-provider.ts
+function createZai(options = {}) {
+  var _a;
+  const baseURL = (_a = withoutTrailingSlash(options.baseURL)) != null ? _a : "https://api.z.ai/api/paas/v4";
+  const getHeaders = () => withUserAgentSuffix(
+    {
+      Authorization: `Bearer ${loadApiKey({
+        apiKey: options.apiKey,
+        environmentVariableName: "ZAI_API_KEY",
+        description: "Z.AI API key"
+      })}`,
+      ...options.headers
+    },
+    `ai-sdk/zai/${VERSION}`
+  );
+  const createLanguageModel = (modelId) => new ZaiChatLanguageModel(modelId, {
+    provider: "zai.chat",
+    baseURL,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
+  const provider = (modelId) => createLanguageModel(modelId);
+  provider.specificationVersion = "v4";
+  provider.languageModel = createLanguageModel;
+  provider.chat = createLanguageModel;
+  provider.embeddingModel = (modelId) => {
+    throw new NoSuchModelError({ modelId, modelType: "embeddingModel" });
+  };
+  provider.textEmbeddingModel = provider.embeddingModel;
+  provider.imageModel = (modelId) => {
+    throw new NoSuchModelError({ modelId, modelType: "imageModel" });
+  };
+  return provider;
+}
+var zai = createZai();
+export {
+  VERSION,
+  createZai,
+  zai
+};
+//# sourceMappingURL=index.js.map

package/dist/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/zai-provider.ts","../src/version.ts","../src/zai-chat-language-model.ts","../src/zai-chat-language-model-options.ts","../src/zai-error.ts"],"sourcesContent":["import {\n  NoSuchModelError,\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 { VERSION } from './version';\nimport { ZaiChatLanguageModel } from './zai-chat-language-model';\nimport type { ZaiChatModelId } from './zai-chat-options';\n\nexport interface ZaiProviderSettings {\n  /**\n   * Z.AI API key. Defaults to the `ZAI_API_KEY` environment variable.\n   */\n  apiKey?: string;\n\n  /**\n   * Base URL for API calls. Defaults to\n   * `https://api.z.ai/api/paas/v4`.\n   */\n  baseURL?: string;\n\n  /**\n   * Custom headers to include in requests.\n   */\n  headers?: Record<string, string>;\n\n  /**\n   * Custom fetch implementation.\n   */\n  fetch?: FetchFunction;\n}\n\nexport interface ZaiProvider extends ProviderV4 {\n  /**\n   * Creates a Z.AI chat model for text generation.\n   */\n  (modelId: ZaiChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a Z.AI language model.\n   */\n  languageModel(modelId: ZaiChatModelId): LanguageModelV4;\n\n  /**\n   * Creates a Z.AI chat model.\n   */\n  chat(modelId: ZaiChatModelId): LanguageModelV4;\n\n  /**\n   * @deprecated Use `embeddingModel` instead.\n   */\n  textEmbeddingModel(modelId: string): never;\n}\n\nexport function createZai(options: ZaiProviderSettings = {}): ZaiProvider {\n  const baseURL =\n    withoutTrailingSlash(options.baseURL) ?? 'https://api.z.ai/api/paas/v4';\n\n  const getHeaders = () =>\n    withUserAgentSuffix(\n      {\n        Authorization: `Bearer ${loadApiKey({\n          apiKey: options.apiKey,\n          environmentVariableName: 'ZAI_API_KEY',\n          description: 'Z.AI API key',\n        })}`,\n        ...options.headers,\n      },\n      `ai-sdk/zai/${VERSION}`,\n    );\n\n  const createLanguageModel = (modelId: ZaiChatModelId) =>\n    new ZaiChatLanguageModel(modelId, {\n      provider: 'zai.chat',\n      baseURL,\n      headers: getHeaders,\n      fetch: options.fetch,\n    });\n\n  const provider = (modelId: ZaiChatModelId) => createLanguageModel(modelId);\n\n  provider.specificationVersion = 'v4' as const;\n  provider.languageModel = createLanguageModel;\n  provider.chat = createLanguageModel;\n\n  provider.embeddingModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'embeddingModel' });\n  };\n  provider.textEmbeddingModel = provider.embeddingModel;\n  provider.imageModel = (modelId: string) => {\n    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });\n  };\n\n  return provider;\n}\n\nexport const zai = createZai();\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","import { OpenAICompatibleChatLanguageModel } from '@ai-sdk/openai-compatible';\nimport type {\n  LanguageModelV4,\n  LanguageModelV4CallOptions,\n  LanguageModelV4FinishReason,\n  LanguageModelV4GenerateResult,\n  LanguageModelV4StreamPart,\n  LanguageModelV4StreamResult,\n  SharedV4Warning,\n} from '@ai-sdk/provider';\nimport {\n  parseProviderOptions,\n  serializeModelOptions,\n  WORKFLOW_DESERIALIZE,\n  WORKFLOW_SERIALIZE,\n  type FetchFunction,\n} from '@ai-sdk/provider-utils';\nimport type { ZaiChatModelId } from './zai-chat-options';\nimport { zaiLanguageModelChatOptions } from './zai-chat-language-model-options';\nimport { zaiErrorStructure } from './zai-error';\n\ntype OpenAICompatibleChatConfig = ConstructorParameters<\n  typeof OpenAICompatibleChatLanguageModel\n>[1];\n\nexport type ZaiChatConfig = {\n  provider: string;\n  baseURL: string;\n  headers?:\n    | Record<string, string | undefined>\n    | (() => Record<string, string | undefined>);\n  fetch?: FetchFunction;\n};\n\nfunction transformZaiRequestBody(\n  args: Record<string, any>,\n): Record<string, any> {\n  const {\n    doSample,\n    frequency_penalty: _frequencyPenalty,\n    presence_penalty: _presencePenalty,\n    requestId,\n    seed: _seed,\n    thinking,\n    toolStream,\n    user: _user,\n    userId,\n    verbosity: _verbosity,\n    ...restArgs\n  } = args;\n\n  return {\n    ...restArgs,\n    ...(doSample !== undefined && { do_sample: doSample }),\n    ...(thinking !== undefined && {\n      thinking: {\n        ...(thinking.type !== undefined && { type: thinking.type }),\n        ...(thinking.clearThinking !== undefined && {\n          clear_thinking: thinking.clearThinking,\n        }),\n      },\n    }),\n    ...(toolStream !== undefined && { tool_stream: toolStream }),\n    ...(requestId !== undefined && { request_id: requestId }),\n    ...(userId !== undefined && { user_id: userId }),\n  };\n}\n\nfunction mapZaiFinishReason(\n  finishReason: LanguageModelV4FinishReason,\n): LanguageModelV4FinishReason {\n  switch (finishReason.raw) {\n    case 'sensitive':\n      return { unified: 'content-filter', raw: finishReason.raw };\n    case 'model_context_window_exceeded':\n      return { unified: 'length', raw: finishReason.raw };\n    case 'network_error':\n      return { unified: 'error', raw: finishReason.raw };\n    default:\n      return finishReason;\n  }\n}\n\nexport class ZaiChatLanguageModel\n  extends OpenAICompatibleChatLanguageModel\n  implements LanguageModelV4\n{\n  private readonly zaiConfig: ZaiChatConfig;\n\n  static [WORKFLOW_SERIALIZE](model: ZaiChatLanguageModel) {\n    return serializeModelOptions({\n      modelId: model.modelId,\n      config: model.zaiConfig,\n    });\n  }\n\n  static [WORKFLOW_DESERIALIZE](options: {\n    modelId: string;\n    config: OpenAICompatibleChatConfig;\n  }) {\n    return new ZaiChatLanguageModel(\n      options.modelId,\n      options.config as unknown as ZaiChatConfig,\n    );\n  }\n\n  constructor(modelId: ZaiChatModelId, config: ZaiChatConfig) {\n    const headers = config.headers;\n\n    super(modelId, {\n      provider: config.provider,\n      url: ({ path }) => `${config.baseURL}${path}`,\n      headers:\n        headers == null\n          ? undefined\n          : () => (typeof headers === 'function' ? headers() : headers),\n      fetch: config.fetch,\n      errorStructure: zaiErrorStructure,\n      transformRequestBody: transformZaiRequestBody,\n      supportedUrls: () => ({\n        'image/*': [/^https?:\\/\\//],\n        'video/*': [/^https?:\\/\\//],\n      }),\n    });\n\n    this.zaiConfig = config;\n  }\n\n  private async prepareCallOptions(options: LanguageModelV4CallOptions) {\n    const warnings: SharedV4Warning[] = [];\n\n    const zaiOptions = await parseProviderOptions({\n      provider: 'zai',\n      providerOptions: options.providerOptions,\n      schema: zaiLanguageModelChatOptions,\n    });\n\n    if (options.frequencyPenalty != null) {\n      warnings.push({ type: 'unsupported', feature: 'frequencyPenalty' });\n    }\n    if (options.presencePenalty != null) {\n      warnings.push({ type: 'unsupported', feature: 'presencePenalty' });\n    }\n    if (options.seed != null) {\n      warnings.push({ type: 'unsupported', feature: 'seed' });\n    }\n\n    let tools = options.tools;\n    let toolChoice = options.toolChoice;\n\n    if (toolChoice?.type === 'none') {\n      tools = undefined;\n      toolChoice = undefined;\n    } else if (toolChoice != null && toolChoice.type !== 'auto') {\n      warnings.push({\n        type: 'unsupported',\n        feature: `toolChoice ${toolChoice.type}`,\n        details: 'Z.AI currently supports only automatic tool selection.',\n      });\n      toolChoice = undefined;\n    }\n\n    const normalizedOptions: LanguageModelV4CallOptions = {\n      ...options,\n      frequencyPenalty: undefined,\n      presencePenalty: undefined,\n      seed: undefined,\n      tools,\n      toolChoice,\n      providerOptions:\n        zaiOptions == null\n          ? options.providerOptions\n          : {\n              ...options.providerOptions,\n              zai: zaiOptions,\n            },\n    };\n\n    return { normalizedOptions, warnings };\n  }\n\n  async doGenerate(\n    options: LanguageModelV4CallOptions,\n  ): Promise<LanguageModelV4GenerateResult> {\n    const { normalizedOptions, warnings } =\n      await this.prepareCallOptions(options);\n    const result = await super.doGenerate(normalizedOptions);\n\n    return {\n      ...result,\n      finishReason: mapZaiFinishReason(result.finishReason),\n      warnings: [...result.warnings, ...warnings],\n    };\n  }\n\n  async doStream(\n    options: LanguageModelV4CallOptions,\n  ): Promise<LanguageModelV4StreamResult> {\n    const { normalizedOptions, warnings } =\n      await this.prepareCallOptions(options);\n    const result = await super.doStream(normalizedOptions);\n\n    return {\n      ...result,\n      stream: result.stream.pipeThrough(\n        new TransformStream<\n          LanguageModelV4StreamPart,\n          LanguageModelV4StreamPart\n        >({\n          transform(part, controller) {\n            if (part.type === 'stream-start') {\n              controller.enqueue({\n                ...part,\n                warnings: [...part.warnings, ...warnings],\n              });\n              return;\n            }\n\n            if (part.type === 'finish') {\n              controller.enqueue({\n                ...part,\n                finishReason: mapZaiFinishReason(part.finishReason),\n              });\n              return;\n            }\n\n            controller.enqueue(part);\n          },\n        }),\n      ),\n    };\n  }\n}\n","import { z } from 'zod/v4';\n\nexport const zaiLanguageModelChatOptions = z.object({\n  /**\n   * Enables or disables sampling. When disabled, temperature and topP do not\n   * take effect.\n   */\n  doSample: z.boolean().optional(),\n\n  /**\n   * Controls model thinking and whether reasoning from earlier turns is kept.\n   */\n  thinking: z\n    .object({\n      type: z.enum(['enabled', 'disabled']).optional(),\n      clearThinking: z.boolean().optional(),\n    })\n    .optional(),\n\n  /**\n   * Controls reasoning effort for GLM-5.2 and later models.\n   */\n  reasoningEffort: z\n    .enum(['none', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max'])\n    .optional(),\n\n  /**\n   * Enables incremental function-call argument streaming on supported models.\n   */\n  toolStream: z.boolean().optional(),\n\n  /**\n   * A caller-provided request identifier between 6 and 64 characters.\n   */\n  requestId: z.string().min(6).max(64).optional(),\n\n  /**\n   * A non-sensitive end-user identifier between 6 and 128 characters.\n   */\n  userId: z.string().min(6).max(128).optional(),\n});\n\nexport type ZaiLanguageModelChatOptions = z.infer<\n  typeof zaiLanguageModelChatOptions\n>;\n","import type { ProviderErrorStructure } from '@ai-sdk/openai-compatible';\nimport { z } from 'zod/v4';\n\nconst zaiErrorDetailsSchema = z.object({\n  code: z.union([z.number(), z.string()]).nullish(),\n  message: z.string(),\n});\n\nconst zaiErrorSchema = z.union([\n  zaiErrorDetailsSchema,\n  z.object({ error: zaiErrorDetailsSchema }),\n]);\n\nexport type ZaiErrorData = z.infer<typeof zaiErrorSchema>;\n\nexport const zaiErrorStructure: ProviderErrorStructure<ZaiErrorData> = {\n  errorSchema: zaiErrorSchema,\n  errorToMessage: data => ('error' in data ? data.error.message : data.message),\n};\n"],"mappings":";AAAA;AAAA,EACE;AAAA,OAGK;AACP;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OAEK;;;ACRA,IAAM,UACX,OACI,UACA;;;ACLN,SAAS,yCAAyC;AAUlD;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OAEK;;;AChBP,SAAS,SAAS;AAEX,IAAM,8BAA8B,EAAE,OAAO;AAAA;AAAA;AAAA;AAAA;AAAA,EAKlD,UAAU,EAAE,QAAQ,EAAE,SAAS;AAAA;AAAA;AAAA;AAAA,EAK/B,UAAU,EACP,OAAO;AAAA,IACN,MAAM,EAAE,KAAK,CAAC,WAAW,UAAU,CAAC,EAAE,SAAS;AAAA,IAC/C,eAAe,EAAE,QAAQ,EAAE,SAAS;AAAA,EACtC,CAAC,EACA,SAAS;AAAA;AAAA;AAAA;AAAA,EAKZ,iBAAiB,EACd,KAAK,CAAC,QAAQ,WAAW,OAAO,UAAU,QAAQ,SAAS,KAAK,CAAC,EACjE,SAAS;AAAA;AAAA;AAAA;AAAA,EAKZ,YAAY,EAAE,QAAQ,EAAE,SAAS;AAAA;AAAA;AAAA;AAAA,EAKjC,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,EAAE,EAAE,SAAS;AAAA;AAAA;AAAA;AAAA,EAK9C,QAAQ,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,IAAI,GAAG,EAAE,SAAS;AAC9C,CAAC;;;ACvCD,SAAS,KAAAA,UAAS;AAElB,IAAM,wBAAwBA,GAAE,OAAO;AAAA,EACrC,MAAMA,GAAE,MAAM,CAACA,GAAE,OAAO,GAAGA,GAAE,OAAO,CAAC,CAAC,EAAE,QAAQ;AAAA,EAChD,SAASA,GAAE,OAAO;AACpB,CAAC;AAED,IAAM,iBAAiBA,GAAE,MAAM;AAAA,EAC7B;AAAA,EACAA,GAAE,OAAO,EAAE,OAAO,sBAAsB,CAAC;AAC3C,CAAC;AAIM,IAAM,oBAA0D;AAAA,EACrE,aAAa;AAAA,EACb,gBAAgB,UAAS,WAAW,OAAO,KAAK,MAAM,UAAU,KAAK;AACvE;;;AFgBA,SAAS,wBACP,MACqB;AACrB,QAAM;AAAA,IACJ;AAAA,IACA,mBAAmB;AAAA,IACnB,kBAAkB;AAAA,IAClB;AAAA,IACA,MAAM;AAAA,IACN;AAAA,IACA;AAAA,IACA,MAAM;AAAA,IACN;AAAA,IACA,WAAW;AAAA,IACX,GAAG;AAAA,EACL,IAAI;AAEJ,SAAO;AAAA,IACL,GAAG;AAAA,IACH,GAAI,aAAa,UAAa,EAAE,WAAW,SAAS;AAAA,IACpD,GAAI,aAAa,UAAa;AAAA,MAC5B,UAAU;AAAA,QACR,GAAI,SAAS,SAAS,UAAa,EAAE,MAAM,SAAS,KAAK;AAAA,QACzD,GAAI,SAAS,kBAAkB,UAAa;AAAA,UAC1C,gBAAgB,SAAS;AAAA,QAC3B;AAAA,MACF;AAAA,IACF;AAAA,IACA,GAAI,eAAe,UAAa,EAAE,aAAa,WAAW;AAAA,IAC1D,GAAI,cAAc,UAAa,EAAE,YAAY,UAAU;AAAA,IACvD,GAAI,WAAW,UAAa,EAAE,SAAS,OAAO;AAAA,EAChD;AACF;AAEA,SAAS,mBACP,cAC6B;AAC7B,UAAQ,aAAa,KAAK;AAAA,IACxB,KAAK;AACH,aAAO,EAAE,SAAS,kBAAkB,KAAK,aAAa,IAAI;AAAA,IAC5D,KAAK;AACH,aAAO,EAAE,SAAS,UAAU,KAAK,aAAa,IAAI;AAAA,IACpD,KAAK;AACH,aAAO,EAAE,SAAS,SAAS,KAAK,aAAa,IAAI;AAAA,IACnD;AACE,aAAO;AAAA,EACX;AACF;AAEO,IAAM,uBAAN,MAAM,8BACH,kCAEV;AAAA,EAGE,QAAQ,kBAAkB,EAAE,OAA6B;AACvD,WAAO,sBAAsB;AAAA,MAC3B,SAAS,MAAM;AAAA,MACf,QAAQ,MAAM;AAAA,IAChB,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ,oBAAoB,EAAE,SAG3B;AACD,WAAO,IAAI;AAAA,MACT,QAAQ;AAAA,MACR,QAAQ;AAAA,IACV;AAAA,EACF;AAAA,EAEA,YAAY,SAAyB,QAAuB;AAC1D,UAAM,UAAU,OAAO;AAEvB,UAAM,SAAS;AAAA,MACb,UAAU,OAAO;AAAA,MACjB,KAAK,CAAC,EAAE,KAAK,MAAM,GAAG,OAAO,OAAO,GAAG,IAAI;AAAA,MAC3C,SACE,WAAW,OACP,SACA,MAAO,OAAO,YAAY,aAAa,QAAQ,IAAI;AAAA,MACzD,OAAO,OAAO;AAAA,MACd,gBAAgB;AAAA,MAChB,sBAAsB;AAAA,MACtB,eAAe,OAAO;AAAA,QACpB,WAAW,CAAC,cAAc;AAAA,QAC1B,WAAW,CAAC,cAAc;AAAA,MAC5B;AAAA,IACF,CAAC;AAED,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAc,mBAAmB,SAAqC;AACpE,UAAM,WAA8B,CAAC;AAErC,UAAM,aAAa,MAAM,qBAAqB;AAAA,MAC5C,UAAU;AAAA,MACV,iBAAiB,QAAQ;AAAA,MACzB,QAAQ;AAAA,IACV,CAAC;AAED,QAAI,QAAQ,oBAAoB,MAAM;AACpC,eAAS,KAAK,EAAE,MAAM,eAAe,SAAS,mBAAmB,CAAC;AAAA,IACpE;AACA,QAAI,QAAQ,mBAAmB,MAAM;AACnC,eAAS,KAAK,EAAE,MAAM,eAAe,SAAS,kBAAkB,CAAC;AAAA,IACnE;AACA,QAAI,QAAQ,QAAQ,MAAM;AACxB,eAAS,KAAK,EAAE,MAAM,eAAe,SAAS,OAAO,CAAC;AAAA,IACxD;AAEA,QAAI,QAAQ,QAAQ;AACpB,QAAI,aAAa,QAAQ;AAEzB,SAAI,yCAAY,UAAS,QAAQ;AAC/B,cAAQ;AACR,mBAAa;AAAA,IACf,WAAW,cAAc,QAAQ,WAAW,SAAS,QAAQ;AAC3D,eAAS,KAAK;AAAA,QACZ,MAAM;AAAA,QACN,SAAS,cAAc,WAAW,IAAI;AAAA,QACtC,SAAS;AAAA,MACX,CAAC;AACD,mBAAa;AAAA,IACf;AAEA,UAAM,oBAAgD;AAAA,MACpD,GAAG;AAAA,MACH,kBAAkB;AAAA,MAClB,iBAAiB;AAAA,MACjB,MAAM;AAAA,MACN;AAAA,MACA;AAAA,MACA,iBACE,cAAc,OACV,QAAQ,kBACR;AAAA,QACE,GAAG,QAAQ;AAAA,QACX,KAAK;AAAA,MACP;AAAA,IACR;AAEA,WAAO,EAAE,mBAAmB,SAAS;AAAA,EACvC;AAAA,EAEA,MAAM,WACJ,SACwC;AACxC,UAAM,EAAE,mBAAmB,SAAS,IAClC,MAAM,KAAK,mBAAmB,OAAO;AACvC,UAAM,SAAS,MAAM,MAAM,WAAW,iBAAiB;AAEvD,WAAO;AAAA,MACL,GAAG;AAAA,MACH,cAAc,mBAAmB,OAAO,YAAY;AAAA,MACpD,UAAU,CAAC,GAAG,OAAO,UAAU,GAAG,QAAQ;AAAA,IAC5C;AAAA,EACF;AAAA,EAEA,MAAM,SACJ,SACsC;AACtC,UAAM,EAAE,mBAAmB,SAAS,IAClC,MAAM,KAAK,mBAAmB,OAAO;AACvC,UAAM,SAAS,MAAM,MAAM,SAAS,iBAAiB;AAErD,WAAO;AAAA,MACL,GAAG;AAAA,MACH,QAAQ,OAAO,OAAO;AAAA,QACpB,IAAI,gBAGF;AAAA,UACA,UAAU,MAAM,YAAY;AAC1B,gBAAI,KAAK,SAAS,gBAAgB;AAChC,yBAAW,QAAQ;AAAA,gBACjB,GAAG;AAAA,gBACH,UAAU,CAAC,GAAG,KAAK,UAAU,GAAG,QAAQ;AAAA,cAC1C,CAAC;AACD;AAAA,YACF;AAEA,gBAAI,KAAK,SAAS,UAAU;AAC1B,yBAAW,QAAQ;AAAA,gBACjB,GAAG;AAAA,gBACH,cAAc,mBAAmB,KAAK,YAAY;AAAA,cACpD,CAAC;AACD;AAAA,YACF;AAEA,uBAAW,QAAQ,IAAI;AAAA,UACzB;AAAA,QACF,CAAC;AAAA,MACH;AAAA,IACF;AAAA,EACF;AACF;;;AF5KO,SAAS,UAAU,UAA+B,CAAC,GAAgB;AA5D1E;AA6DE,QAAM,WACJ,0BAAqB,QAAQ,OAAO,MAApC,YAAyC;AAE3C,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,cAAc,OAAO;AAAA,EACvB;AAEF,QAAM,sBAAsB,CAAC,YAC3B,IAAI,qBAAqB,SAAS;AAAA,IAChC,UAAU;AAAA,IACV;AAAA,IACA,SAAS;AAAA,IACT,OAAO,QAAQ;AAAA,EACjB,CAAC;AAEH,QAAM,WAAW,CAAC,YAA4B,oBAAoB,OAAO;AAEzE,WAAS,uBAAuB;AAChC,WAAS,gBAAgB;AACzB,WAAS,OAAO;AAEhB,WAAS,iBAAiB,CAAC,YAAoB;AAC7C,UAAM,IAAI,iBAAiB,EAAE,SAAS,WAAW,iBAAiB,CAAC;AAAA,EACrE;AACA,WAAS,qBAAqB,SAAS;AACvC,WAAS,aAAa,CAAC,YAAoB;AACzC,UAAM,IAAI,iBAAiB,EAAE,SAAS,WAAW,aAAa,CAAC;AAAA,EACjE;AAEA,SAAO;AACT;AAEO,IAAM,MAAM,UAAU;","names":["z"]}

package/docs/200-zai.mdx

@@ -0,0 +1,159 @@
+---
+title: Z.AI
+description: Learn how to use Z.AI's GLM models with the AI SDK.
+---
+
+# Z.AI Provider
+
+The [Z.AI](https://z.ai/) provider gives you access to GLM language and vision models through the Z.AI API.
+
+API keys can be created in the [Z.AI API key console](https://z.ai/manage-apikey/apikey-list).
+
+## Setup
+
+The Z.AI provider is available through the `@ai-sdk/zai` package. Install it with:
+
+<InstallPackages packages="@ai-sdk/zai" />
+
+Set the `ZAI_API_KEY` environment variable:
+
+```bash
+ZAI_API_KEY=your-api-key
+```
+
+## Provider Instance
+
+Import the default provider instance `zai` from `@ai-sdk/zai`:
+
+```ts
+import { zai } from '@ai-sdk/zai';
+```
+
+For custom configuration, use `createZai`:
+
+```ts
+import { createZai } from '@ai-sdk/zai';
+
+const zai = createZai({
+  apiKey: process.env.ZAI_API_KEY,
+  baseURL: 'https://api.z.ai/api/paas/v4',
+});
+```
+
+The provider accepts these optional settings:
+
+- **apiKey** _string_
+
+  API key sent in the `Authorization` header. It defaults to the `ZAI_API_KEY` environment variable.
+
+- **baseURL** _string_
+
+  URL prefix for API requests. It defaults to `https://api.z.ai/api/paas/v4`.
+
+- **headers** _Record&lt;string, string&gt;_
+
+  Additional request headers.
+
+- **fetch** _FetchFunction_
+
+  Custom fetch implementation, for example to intercept requests in middleware.
+
+## Language Models
+
+Create a language model by passing its model id to the provider:
+
+```ts
+import { zai } from '@ai-sdk/zai';
+import { generateText } from 'ai';
+
+const { text } = await generateText({
+  model: zai('glm-5.3'),
+  prompt: 'Explain quantum entanglement in simple terms.',
+});
+```
+
+Current model families include GLM 5, GLM 4.7, GLM 4.6, GLM 4.5, and the GLM vision models. Model availability changes over time; see the [Z.AI model documentation](https://docs.z.ai/guides/llm) for the current catalog.
+
+The provider supports:
+
+- text generation and streaming
+- reasoning output and reasoning history
+- function calling, including incremental tool-call streaming
+- JSON object output
+- URL-based image and video inputs on compatible vision models
+
+## Provider Options
+
+Z.AI-specific options can be passed through `providerOptions.zai`:
+
+```ts
+import { zai } from '@ai-sdk/zai';
+import { generateText } from 'ai';
+
+const result = await generateText({
+  model: zai('glm-5.3'),
+  prompt: 'Compare two approaches to implementing a rate limiter.',
+  providerOptions: {
+    zai: {
+      thinking: { type: 'enabled', clearThinking: true },
+      reasoningEffort: 'high',
+      requestId: 'request-123456',
+      userId: 'user-123456',
+    },
+  },
+});
+```
+
+The following options are available:
+
+- **doSample** _boolean_
+
+  Enables sampling. When disabled, `temperature` and `topP` do not take effect.
+
+- **thinking** _object_
+
+  Controls thinking with `type: 'enabled' | 'disabled'`. Set `clearThinking` to `false` to retain reasoning from previous assistant messages.
+
+- **reasoningEffort** _'none' | 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' | 'max'_
+
+  Controls reasoning effort on supported models.
+
+- **toolStream** _boolean_
+
+  Enables incremental function-call argument streaming on supported models.
+
+- **requestId** _string_
+
+  A caller-provided request id between 6 and 64 characters.
+
+- **userId** _string_
+
+  A non-sensitive end-user id between 6 and 128 characters.
+
+## Streaming Tool Calls
+
+Enable `toolStream` when you want supported GLM models to stream function-call arguments incrementally:
+
+```ts
+import { zai } from '@ai-sdk/zai';
+import { streamText, tool } from 'ai';
+import { z } from 'zod';
+
+const result = streamText({
+  model: zai('glm-5.3'),
+  prompt: 'What is the weather in San Francisco?',
+  tools: {
+    weather: tool({
+      description: 'Get the weather for a city',
+      inputSchema: z.object({ city: z.string() }),
+    }),
+  },
+  providerOptions: {
+    zai: { toolStream: true },
+  },
+});
+
+for await (const part of result.fullStream) {
+  console.log(part);
+}
+```

package/package.json

@@ -1,23 +1,77 @@
 {
   "name": "@ai-sdk/zai",
-  "version": "0.0.0",
-  "description": "AI SDK provider for Z.ai",
-  "license": "Apache-2.0",
+  "version": "3.0.1",
   "type": "module",
-  "main": "./index.js",
-  "exports": {
-    ".": "./index.js"
-  },
+  "license": "Apache-2.0",
+  "sideEffects": false,
+  "main": "./dist/index.js",
+  "types": "./dist/index.d.ts",
   "files": [
-    "index.js",
+    "dist/**/*",
+    "docs/**/*",
+    "src",
+    "!src/**/*.test.ts",
+    "!src/**/*.test-d.ts",
+    "!src/**/__snapshots__",
+    "!src/**/__fixtures__",
+    "CHANGELOG.md",
     "README.md"
   ],
+  "directories": {
+    "doc": "./docs"
+  },
+  "exports": {
+    "./package.json": "./package.json",
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js",
+      "default": "./dist/index.js"
+    }
+  },
+  "dependencies": {
+    "@ai-sdk/openai-compatible": "3.0.39",
+    "@ai-sdk/provider": "4.0.8",
+    "@ai-sdk/provider-utils": "5.0.32"
+  },
+  "devDependencies": {
+    "@types/node": "22.19.19",
+    "@vercel/ai-tsconfig": "0.0.0",
+    "tsup": "^8.5.1",
+    "typescript": "5.8.3",
+    "zod": "3.25.76"
+  },
+  "peerDependencies": {
+    "zod": "^3.25.76 || ^4.1.8"
+  },
+  "engines": {
+    "node": ">=22"
+  },
   "publishConfig": {
-    "access": "public"
+    "access": "public",
+    "provenance": true
   },
+  "homepage": "https://ai-sdk.dev/docs",
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/vercel/ai.git",
+    "url": "https://github.com/vercel/ai",
     "directory": "packages/zai"
+  },
+  "bugs": {
+    "url": "https://github.com/vercel/ai/issues"
+  },
+  "keywords": [
+    "ai"
+  ],
+  "description": "The Z.AI provider for the AI SDK contains language model support for Z.AI's GLM models.",
+  "scripts": {
+    "build": "pnpm clean && tsup --tsconfig tsconfig.build.json",
+    "build:watch": "pnpm clean && tsup --watch",
+    "clean": "del-cli dist docs *.tsbuildinfo",
+    "type-check": "tsc --build",
+    "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"
   }
-}
+}

package/src/index.ts

@@ -0,0 +1,6 @@
+export type { ZaiLanguageModelChatOptions } from './zai-chat-language-model-options';
+export type { ZaiChatModelId } from './zai-chat-options';
+export type { ZaiErrorData } from './zai-error';
+export { createZai, zai } from './zai-provider';
+export type { ZaiProvider, ZaiProviderSettings } from './zai-provider';
+export { VERSION } from './version';

package/src/version.ts

@@ -0,0 +1,6 @@
+// Version string of this package injected at build time.
+declare const __PACKAGE_VERSION__: string | undefined;
+export const VERSION: string =
+  typeof __PACKAGE_VERSION__ !== 'undefined'
+    ? __PACKAGE_VERSION__
+    : '0.0.0-test';

package/src/zai-chat-language-model-options.ts

@@ -0,0 +1,45 @@
+import { z } from 'zod/v4';
+
+export const zaiLanguageModelChatOptions = z.object({
+  /**
+   * Enables or disables sampling. When disabled, temperature and topP do not
+   * take effect.
+   */
+  doSample: z.boolean().optional(),
+
+  /**
+   * Controls model thinking and whether reasoning from earlier turns is kept.
+   */
+  thinking: z
+    .object({
+      type: z.enum(['enabled', 'disabled']).optional(),
+      clearThinking: z.boolean().optional(),
+    })
+    .optional(),
+
+  /**
+   * Controls reasoning effort for GLM-5.2 and later models.
+   */
+  reasoningEffort: z
+    .enum(['none', 'minimal', 'low', 'medium', 'high', 'xhigh', 'max'])
+    .optional(),
+
+  /**
+   * Enables incremental function-call argument streaming on supported models.
+   */
+  toolStream: z.boolean().optional(),
+
+  /**
+   * A caller-provided request identifier between 6 and 64 characters.
+   */
+  requestId: z.string().min(6).max(64).optional(),
+
+  /**
+   * A non-sensitive end-user identifier between 6 and 128 characters.
+   */
+  userId: z.string().min(6).max(128).optional(),
+});
+
+export type ZaiLanguageModelChatOptions = z.infer<
+  typeof zaiLanguageModelChatOptions
+>;

package/src/zai-chat-language-model.ts

@@ -0,0 +1,233 @@
+import { OpenAICompatibleChatLanguageModel } from '@ai-sdk/openai-compatible';
+import type {
+  LanguageModelV4,
+  LanguageModelV4CallOptions,
+  LanguageModelV4FinishReason,
+  LanguageModelV4GenerateResult,
+  LanguageModelV4StreamPart,
+  LanguageModelV4StreamResult,
+  SharedV4Warning,
+} from '@ai-sdk/provider';
+import {
+  parseProviderOptions,
+  serializeModelOptions,
+  WORKFLOW_DESERIALIZE,
+  WORKFLOW_SERIALIZE,
+  type FetchFunction,
+} from '@ai-sdk/provider-utils';
+import type { ZaiChatModelId } from './zai-chat-options';
+import { zaiLanguageModelChatOptions } from './zai-chat-language-model-options';
+import { zaiErrorStructure } from './zai-error';
+
+type OpenAICompatibleChatConfig = ConstructorParameters<
+  typeof OpenAICompatibleChatLanguageModel
+>[1];
+
+export type ZaiChatConfig = {
+  provider: string;
+  baseURL: string;
+  headers?:
+    | Record<string, string | undefined>
+    | (() => Record<string, string | undefined>);
+  fetch?: FetchFunction;
+};
+
+function transformZaiRequestBody(
+  args: Record<string, any>,
+): Record<string, any> {
+  const {
+    doSample,
+    frequency_penalty: _frequencyPenalty,
+    presence_penalty: _presencePenalty,
+    requestId,
+    seed: _seed,
+    thinking,
+    toolStream,
+    user: _user,
+    userId,
+    verbosity: _verbosity,
+    ...restArgs
+  } = args;
+
+  return {
+    ...restArgs,
+    ...(doSample !== undefined && { do_sample: doSample }),
+    ...(thinking !== undefined && {
+      thinking: {
+        ...(thinking.type !== undefined && { type: thinking.type }),
+        ...(thinking.clearThinking !== undefined && {
+          clear_thinking: thinking.clearThinking,
+        }),
+      },
+    }),
+    ...(toolStream !== undefined && { tool_stream: toolStream }),
+    ...(requestId !== undefined && { request_id: requestId }),
+    ...(userId !== undefined && { user_id: userId }),
+  };
+}
+
+function mapZaiFinishReason(
+  finishReason: LanguageModelV4FinishReason,
+): LanguageModelV4FinishReason {
+  switch (finishReason.raw) {
+    case 'sensitive':
+      return { unified: 'content-filter', raw: finishReason.raw };
+    case 'model_context_window_exceeded':
+      return { unified: 'length', raw: finishReason.raw };
+    case 'network_error':
+      return { unified: 'error', raw: finishReason.raw };
+    default:
+      return finishReason;
+  }
+}
+
+export class ZaiChatLanguageModel
+  extends OpenAICompatibleChatLanguageModel
+  implements LanguageModelV4
+{
+  private readonly zaiConfig: ZaiChatConfig;
+
+  static [WORKFLOW_SERIALIZE](model: ZaiChatLanguageModel) {
+    return serializeModelOptions({
+      modelId: model.modelId,
+      config: model.zaiConfig,
+    });
+  }
+
+  static [WORKFLOW_DESERIALIZE](options: {
+    modelId: string;
+    config: OpenAICompatibleChatConfig;
+  }) {
+    return new ZaiChatLanguageModel(
+      options.modelId,
+      options.config as unknown as ZaiChatConfig,
+    );
+  }
+
+  constructor(modelId: ZaiChatModelId, config: ZaiChatConfig) {
+    const headers = config.headers;
+
+    super(modelId, {
+      provider: config.provider,
+      url: ({ path }) => `${config.baseURL}${path}`,
+      headers:
+        headers == null
+          ? undefined
+          : () => (typeof headers === 'function' ? headers() : headers),
+      fetch: config.fetch,
+      errorStructure: zaiErrorStructure,
+      transformRequestBody: transformZaiRequestBody,
+      supportedUrls: () => ({
+        'image/*': [/^https?:\/\//],
+        'video/*': [/^https?:\/\//],
+      }),
+    });
+
+    this.zaiConfig = config;
+  }
+
+  private async prepareCallOptions(options: LanguageModelV4CallOptions) {
+    const warnings: SharedV4Warning[] = [];
+
+    const zaiOptions = await parseProviderOptions({
+      provider: 'zai',
+      providerOptions: options.providerOptions,
+      schema: zaiLanguageModelChatOptions,
+    });
+
+    if (options.frequencyPenalty != null) {
+      warnings.push({ type: 'unsupported', feature: 'frequencyPenalty' });
+    }
+    if (options.presencePenalty != null) {
+      warnings.push({ type: 'unsupported', feature: 'presencePenalty' });
+    }
+    if (options.seed != null) {
+      warnings.push({ type: 'unsupported', feature: 'seed' });
+    }
+
+    let tools = options.tools;
+    let toolChoice = options.toolChoice;
+
+    if (toolChoice?.type === 'none') {
+      tools = undefined;
+      toolChoice = undefined;
+    } else if (toolChoice != null && toolChoice.type !== 'auto') {
+      warnings.push({
+        type: 'unsupported',
+        feature: `toolChoice ${toolChoice.type}`,
+        details: 'Z.AI currently supports only automatic tool selection.',
+      });
+      toolChoice = undefined;
+    }
+
+    const normalizedOptions: LanguageModelV4CallOptions = {
+      ...options,
+      frequencyPenalty: undefined,
+      presencePenalty: undefined,
+      seed: undefined,
+      tools,
+      toolChoice,
+      providerOptions:
+        zaiOptions == null
+          ? options.providerOptions
+          : {
+              ...options.providerOptions,
+              zai: zaiOptions,
+            },
+    };
+
+    return { normalizedOptions, warnings };
+  }
+
+  async doGenerate(
+    options: LanguageModelV4CallOptions,
+  ): Promise<LanguageModelV4GenerateResult> {
+    const { normalizedOptions, warnings } =
+      await this.prepareCallOptions(options);
+    const result = await super.doGenerate(normalizedOptions);
+
+    return {
+      ...result,
+      finishReason: mapZaiFinishReason(result.finishReason),
+      warnings: [...result.warnings, ...warnings],
+    };
+  }
+
+  async doStream(
+    options: LanguageModelV4CallOptions,
+  ): Promise<LanguageModelV4StreamResult> {
+    const { normalizedOptions, warnings } =
+      await this.prepareCallOptions(options);
+    const result = await super.doStream(normalizedOptions);
+
+    return {
+      ...result,
+      stream: result.stream.pipeThrough(
+        new TransformStream<
+          LanguageModelV4StreamPart,
+          LanguageModelV4StreamPart
+        >({
+          transform(part, controller) {
+            if (part.type === 'stream-start') {
+              controller.enqueue({
+                ...part,
+                warnings: [...part.warnings, ...warnings],
+              });
+              return;
+            }
+
+            if (part.type === 'finish') {
+              controller.enqueue({
+                ...part,
+                finishReason: mapZaiFinishReason(part.finishReason),
+              });
+              return;
+            }
+
+            controller.enqueue(part);
+          },
+        }),
+      ),
+    };
+  }
+}

package/src/zai-chat-options.ts

@@ -0,0 +1,28 @@
+/**
+ * Z.AI chat model ids from the official OpenAPI 1.0.0 specification,
+ * retrieved from https://docs.z.ai/openapi.json on 2026-08-26.
+ */
+export type ZaiChatModelId =
+  | 'glm-5.3'
+  | 'glm-5.2'
+  | 'glm-5.1'
+  | 'glm-5-turbo'
+  | 'glm-5'
+  | 'glm-4.7'
+  | 'glm-4.7-flash'
+  | 'glm-4.7-flashx'
+  | 'glm-4.6'
+  | 'glm-4.5'
+  | 'glm-4.5-air'
+  | 'glm-4.5-x'
+  | 'glm-4.5-airx'
+  | 'glm-4.5-flash'
+  | 'glm-4-32b-0414-128k'
+  | 'glm-5.3-flash'
+  | 'glm-5v-turbo'
+  | 'glm-4.6v'
+  | 'glm-4.6v-flash'
+  | 'glm-4.6v-flashx'
+  | 'glm-4.5v'
+  | 'autoglm-phone-multilingual'
+  | (string & {});

package/src/zai-error.ts

@@ -0,0 +1,19 @@
+import type { ProviderErrorStructure } from '@ai-sdk/openai-compatible';
+import { z } from 'zod/v4';
+
+const zaiErrorDetailsSchema = z.object({
+  code: z.union([z.number(), z.string()]).nullish(),
+  message: z.string(),
+});
+
+const zaiErrorSchema = z.union([
+  zaiErrorDetailsSchema,
+  z.object({ error: zaiErrorDetailsSchema }),
+]);
+
+export type ZaiErrorData = z.infer<typeof zaiErrorSchema>;
+
+export const zaiErrorStructure: ProviderErrorStructure<ZaiErrorData> = {
+  errorSchema: zaiErrorSchema,
+  errorToMessage: data => ('error' in data ? data.error.message : data.message),
+};

package/src/zai-provider.ts

@@ -0,0 +1,103 @@
+import {
+  NoSuchModelError,
+  type LanguageModelV4,
+  type ProviderV4,
+} from '@ai-sdk/provider';
+import {
+  loadApiKey,
+  withoutTrailingSlash,
+  withUserAgentSuffix,
+  type FetchFunction,
+} from '@ai-sdk/provider-utils';
+import { VERSION } from './version';
+import { ZaiChatLanguageModel } from './zai-chat-language-model';
+import type { ZaiChatModelId } from './zai-chat-options';
+
+export interface ZaiProviderSettings {
+  /**
+   * Z.AI API key. Defaults to the `ZAI_API_KEY` environment variable.
+   */
+  apiKey?: string;
+
+  /**
+   * Base URL for API calls. Defaults to
+   * `https://api.z.ai/api/paas/v4`.
+   */
+  baseURL?: string;
+
+  /**
+   * Custom headers to include in requests.
+   */
+  headers?: Record<string, string>;
+
+  /**
+   * Custom fetch implementation.
+   */
+  fetch?: FetchFunction;
+}
+
+export interface ZaiProvider extends ProviderV4 {
+  /**
+   * Creates a Z.AI chat model for text generation.
+   */
+  (modelId: ZaiChatModelId): LanguageModelV4;
+
+  /**
+   * Creates a Z.AI language model.
+   */
+  languageModel(modelId: ZaiChatModelId): LanguageModelV4;
+
+  /**
+   * Creates a Z.AI chat model.
+   */
+  chat(modelId: ZaiChatModelId): LanguageModelV4;
+
+  /**
+   * @deprecated Use `embeddingModel` instead.
+   */
+  textEmbeddingModel(modelId: string): never;
+}
+
+export function createZai(options: ZaiProviderSettings = {}): ZaiProvider {
+  const baseURL =
+    withoutTrailingSlash(options.baseURL) ?? 'https://api.z.ai/api/paas/v4';
+
+  const getHeaders = () =>
+    withUserAgentSuffix(
+      {
+        Authorization: `Bearer ${loadApiKey({
+          apiKey: options.apiKey,
+          environmentVariableName: 'ZAI_API_KEY',
+          description: 'Z.AI API key',
+        })}`,
+        ...options.headers,
+      },
+      `ai-sdk/zai/${VERSION}`,
+    );
+
+  const createLanguageModel = (modelId: ZaiChatModelId) =>
+    new ZaiChatLanguageModel(modelId, {
+      provider: 'zai.chat',
+      baseURL,
+      headers: getHeaders,
+      fetch: options.fetch,
+    });
+
+  const provider = (modelId: ZaiChatModelId) => createLanguageModel(modelId);
+
+  provider.specificationVersion = 'v4' as const;
+  provider.languageModel = createLanguageModel;
+  provider.chat = createLanguageModel;
+
+  provider.embeddingModel = (modelId: string) => {
+    throw new NoSuchModelError({ modelId, modelType: 'embeddingModel' });
+  };
+  provider.textEmbeddingModel = provider.embeddingModel;
+  provider.imageModel = (modelId: string) => {
+    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });
+  };
+
+  return provider;
+}
+
+export const zai = createZai();

package/index.js

@@ -1 +0,0 @@
-export {};