Getting Started

The goal of this chapter is to take you from “knowing nothing” to “running the demo and seeing AI responses”, and to clarify what your backend needs to implement.

Required pieces (otherwise it will error or be unavailable):

1. At least one AI model configuration: ai.models.length > 0. Umo Editor Next uses ai.models[0] as the default model.
2. An accessible backend endpoint (SSE recommended): ai.models[].endpoint. The frontend will call this endpoint. You can test first by pointing to the default SSE demo endpoints provided by Umo Editor Server: /mock/sse/normal or /mock/sse/agui. See: API list.

Minimal configuration

The minimum working configuration:

const options = {
  ai: {
    models: [
      {
        value: 'default',
        label: { zh_CN: '默认模型', en_US: 'Default Model' },
        avatar: '/images/ai.svg',
        protocol: 'default', // or 'agui'
        endpoint: 'http://127.0.0.1:1235/mock/sse/normal', // SSE demo endpoint from Umo Editor Server
        stream: true,
        reasoning: true,
      },
    ],
    chat: { enabled: true }, // Sidebar chat
    assistant: { enabled: true }, // Document bubble assistant
  },
}

Choosing a protocol

For protocol details, see Supported protocols. Practical guidance:

• If your backend only needs to “return model output via SSE” and you want a custom data shape: use protocol: 'default'.
• If you already have (or plan to implement) AG-UI (with tool calls, status events, etc.): use protocol: 'agui'.

When protocol: 'agui', the frontend will not use custom parsing via ai.callbacks.onMessage. Instead it uses the built-in AG-UI parser and renderer. Your backend must output an AG-UI event stream (see Supported protocols).

Verify

You can simply:

1. Open the “Document Assistant” entry (bubble menu) or the chat entry at the bottom-right
2. Type anything and send
3. Check whether you can see an assistant reply

If there is no output, start with Troubleshooting.