XpertAI Plugin: ZhipuAI Models

Zhipu AI model provider plugin for the XpertAI platform.

Features

• Support for Zhipu conversational and completion models
• OpenAI-compatible API surface for easy integration
• Streaming responses for low-latency UI updates
• Tool-calling and instruction-following support
• CogVideoX text-to-video and image-to-video toolset with Workspace artifacts

Installation

This plugin is included in the XpertAI plugins monorepo and will be loaded automatically by the platform when installed.

Configuration

Provide credentials and optional endpoints via your environment or platform configuration.

Required Credentials

• api_key: Your Zhipu AI API key
• base_url (optional): Custom API endpoint, defaults to the official Zhipu API host

Usage

The plugin exposes an OpenAI-compatible interface so existing callers can switch providers with minimal changes. It supports streaming responses and tool-calling where the model suggests structured tool invocations.

CogVideoX Toolset

The zhipu_cogvideo builtin toolset adds two tools:

• zhipu_cogvideo_submit: submits a text-to-video or image-to-video task and returns its task ID.
• zhipu_cogvideo_query: checks the task, waits for a bounded period, and uploads completed MP4 files and cover images to the Xpert Workspace.

Configure the toolset with a ZhipuAI API key. The endpoint defaults to https://open.bigmodel.cn/api/paas/v4. Generated files are stored under files/zhipuai/cogvideo/ in the current project or Xpert scope.

Basic example (pseudo):

1. Initialize the client with api_key and optional base_url.
2. Send chat messages to the zhipuai-chat model.
3. Receive streaming partial responses or final completions.

License

AGPL-3.0

For more details, see the XpertAI platform docs and the example usages in the models/zhipuai folder.