@wuzhiguocarter/zhipu-ai-provider weekly downloads statistics

Zhipu AI Provider - Vercel AI SDK Community Provider

This is a Zhipu prodiver for the AI SDK. It enables seamless integration with GLM and Embedding Models provided on bigmodel.cn.

Setup

# npm
npm i @wuzhiguocarter/zhipu-ai-provider

# pnpm
pnpm add @wuzhiguocarter/zhipu-ai-provider

# yarn
yarn add @wuzhiguocarter/zhipu-ai-provider

Set up your .env file / environment with your API key.

ZHIPU_API_KEY=<your-api-key>

Provider Instance

You can import the default provider instance zhipu from zhipu-ai-provider (This automatically reads the API key from the environment variable ZHIPU_API_KEY):

import { zhipu } from '@wuzhiguocarter/zhipu-ai-provider'

Alternatively, you can create a provider instance with custom configuration with createZhipu:

import { createZhipu } from '@wuzhiguocarter/zhipu-ai-provider';

const zhipu = createZhipu({
  baseURL: "https://open.bigmodel.cn/api/paas/v4",
  apiKey: "your-api-key"
});

You can use the following optional settings to customize the Zhipu provider instance:

• baseURL: string
  • Use a different URL prefix for API calls, e.g. to use proxy servers. The default prefix is https://open.bigmodel.cn/api/paas/v4.
• apiKey: string
  • Your API key for Zhipu BigModel Platform. If not provided, the provider will attempt to read the API key from the environment variable ZHIPU_API_KEY.
• headers: Record<string,string>
  • Custom headers to include in the requests.

Language Model Example

import { generateText } from 'ai';
import { zhipu } from '@wuzhiguocarter/zhipu-ai-provider';

const { text } = await generateText({
  model: zhipu('glm-4-plus'),
  prompt: 'Why is the sky blue?',
});

console.log(result)

Supported Language Models

• GLM-4.7 (Latest flagship, 200K context): glm-4.7
• GLM-4.6 (High performance, 200K context): glm-4.6
• GLM-4.5 (Excellent performance): glm-4.5, glm-4.5-x, glm-4.5-air, glm-4.5-airx, glm-4.5-flash (free)
• GLM-4.6V (Vision reasoning): glm-4.6v, glm-4.6v-flash (free)
• GLM-4.5V (Vision reasoning): glm-4.5v
• Legacy models: glm-4-plus, glm-4-air, glm-4-flash, glm-4-long, glm-4v, glm-z1-*

Thinking Mode (GLM-4.5/4.6/4.7)

GLM-4.5, GLM-4.6, and GLM-4.7 models support a "thinking mode" for complex reasoning:

import { generateText } from 'ai';
import { zhipu } from '@wuzhiguocarter/zhipu-ai-provider';

const { text } = await generateText({
  model: zhipu('glm-4.7', {
    thinking: {
      type: 'enabled',        // Enable deep thinking
      clear_thinking: true    // Include reasoning process in response
    }
  }),
  prompt: 'Explain quantum computing in detail',
});

// To disable thinking for faster responses:
const { text: quickText } = await generateText({
  model: zhipu('glm-4.7', {
    thinking: { type: 'disabled' }
  }),
  prompt: 'What is 2+2?',
});

• thinking: { type: 'enabled' } - Enable dynamic thinking based on task complexity (default for GLM-4.5+)
• thinking: { type: 'disabled' } - Disable thinking for faster, more direct responses
• thinking: { clear_thinking: true } - Include the reasoning process in the response (reasoning_content field)
• thinking: { clear_thinking: false } - Hide the reasoning process, only show final answer

Tool Streaming (GLM-4.6/4.7)

GLM-4.7 and GLM-4.6 models support controlling tool call streaming:

const result = await generateText({
  model: zhipu('glm-4.7', {
    toolStream: true  // Stream tool call parameters
  }),
  tools: {
    getWeather: {
      description: 'Get weather information',
      parameters: z.object({
        city: z.string(),
      }),
      execute: async ({ city }) => `${city} Sunny 25°C`
    }
  },
  prompt: 'What is the weather in Beijing today?'
});

• toolStream: true - Tool call parameters are streamed in chunks for faster feedback
• toolStream: false - Wait for complete tool call before returning

Response Format

Control the output format of the model:

// JSON mode
const { text } = await generateText({
  model: zhipu('glm-4-flash'),
  responseFormat: { type: 'json' },
  prompt: 'List three fruits in JSON array format'
});

// Text mode (explicit)
const { text } = await generateText({
  model: zhipu('glm-4-flash'),
  responseFormat: { type: 'text' },
  prompt: 'Write a poem'
});

• { type: 'text' } - Plain text output (default)
• { type: 'json' } - JSON format output (text models only)

Note: Vision and reasoning models do not support JSON format.

Embedding Example

const { embedding } = await embed({
  model: zhipu.textEmbeddingModel("embedding-3", {
    dimensions: 256, // Optional, defaults to 2048
  }),
  value: "Hello, world!",
});

console.log(embedding);

Image Generation Example

Zhipu supports image generation with the cogview models, but the api does not return images in base64 or buffer format, so the image urls are returned in the providerMetadata field.

import { experimental_generateImage as generateImage } from 'ai';
import { zhipu } from '@wuzhiguocarter/zhipu-ai-provider';

const { image, providerMetadata } = await generateImage({
  model: zhipu.ImageModel('cogview-4-250304'),
  prompt: 'A beautiful landscape with mountains and a river',
  size: '1024x1024',  // optional
  providerOptions: {  // optional
      zhipu: {
          quality: 'hd'
      }
  }
});

console.log(providerMetadata.zhipu.images[0].url)

Testing

Unit Tests

The project includes comprehensive unit tests that use mock data and do not require API calls:

# Run all unit tests
pnpm test src

# Run specific test file
pnpm test src/zhipu-chat-language-model.test.ts

Integration Tests

Integration tests use the real Zhipu AI API and require an API key:…