AI Chat

AI Chat means that in the menu, by clicking the dialogue button, you can use the edited content to have a dialogue with AI. It mainly provides 4 capabilities:

• Directly select the content of the editor as the context of the dialogue for continuous dialogue.
• The AI content of the dialogue can be inserted (or replaced) into the specified position of the editor at any time.
• Support historical dialogue
• Complete function configuration

Note: The current function is only available in the commercial Pro version, demonstration address: http://pro.aieditor.com.cn

Sample Code

new AiEditorPro({
    element: "#aiEditor",
    chat:{
        systemPrompt: "Your name is AIEditor Editing Assistant, you are good at...",
        helloMessage: "Welcome use AIEditor editing assistant, you can...",
        localStorageKey: "aie-chat-messages",
        placeholder: "what's your question?",
        maxHeight: 600,
        maxAttachedMessageCount: 10,
        historyMessageTruncateEnable: false,
        historyMessageTruncateLength: 1000,
        historyMessageTruncateProcessor?: (message: string) => string,
        appendEditorSelectedContentEnable: true,
        appendEditorSelectedContentProcessor?: (userContent: string, selectedContent: string) => string,
        messagesToPayloadProcessor?: (messages: any, aiModel: AiModel) => string,
    },
})

Configuration Instructions

systemPrompt

systemPrompt Messages with the system role act as top-level instructions for models, typically describing what the model should do and how it should generally behave and respond. When systemPrompt is configured, each message will carry systemPrompt, as shown in the following code:

{
  "model": "gpt-4o",
  "stream": true,
  "messages": [
    {
      "role": "system",
      "content": "Your name is AIEditor Editorial Assistant, and you are good at text creation and content optimization."
    },
    {
      "role": "user",
      "content": "hello"
    }
  ]
}

helloMessage

Used to configure the welcome message when the user first opens the AI dialog.

localStorageKey

The key value definition of historical messages saved in LocalStorage, the default value is: aie-chat-messages

placeholder

Set the placeholder attribute of the input box

maxHeight

Set the maximum height of the chat window

maxAttachedMessageCount

In the history conversation, the maximum number of history messages that can be carried is 10 by default. The more history messages there are, the more tokens are consumed.

historyMessageTruncateEnable

Whether to enable historical message truncation, the default value is: false.

In the messages responded by AI, there may be some messages with very long content. If we submit these long messages to AI, it may consume a lot of tokens. We can only intercept part of the content and submit it instead of the complete content.

historyMessageTruncateLength

The length of the intercepted content. The default value is: 1000. This configuration is only valid when historyMessageTruncateEnable: true is configured.

historyMessageTruncateProcessor

Customize the history message truncate processor. After this configuration, the configuration of historyMessageTruncateLength will become invalid.

In the AI history dialogue compression strategy, in addition to content truncate, there are also multiple strategies such as summary extraction and keyword extraction, which can be implemented by custom configuration historyMessageTruncateProcessor.

appendEditorSelectedContentEnable

In the message dialog, whether to automatically append the text content selected by the editor. The default value is: true.

appendEditorSelectedContentProcessor

Customize the handlers of user message and editor selection content. The default code is as follows:

appendEditorSelectedContentProcessor = (userContent, selectedContent)=>{
    return `${userContent}\n${selectedContent}`
}

messagesToPayloadProcessor

This configuration is used to convert historical message content into payload, which serves as the body content of AI request. It is only used when customizing large models.