Skip to main content

Implement Chatbot Function Based on OpenIM

Brief Description​

Use the Webhook mechanism in OpenIM to implement chatbot functions. After sending a text message or image message to the chatbot, the bot will return the same message. Developers can replace this logic and call the LLM interface to achieve intelligent question answering features.

1. Modify Configuration File​

Set enable of the callback to true to indicate starting this callback. If you need to add a new callback, refer to the config/webhooks.yml configuration in open-im-server, and modify accordingly at the code level.

Tip:

  • url is the callback URL.
  • When afterSendSingleMsg.enable is set to true, this callback is enabled.

2. Create Chatbot Account​

  1. Log in to the management backend, refer to this document.
  2. Create a chatbot account in user management and record the userID of this account.
  3. For convenience of testing, you can set this userID as a default friend.

3. Write afterSendSingleMsg Interface​

Refer to the following example code.

Tip:

  1. Replace robotics in the example with the userID obtained in step 2.
func (m *ChatApi) CallbackExample(c *gin.Context) {
    // 1. Handling callbacks after sending a single chat message
    msgInfo, err := handlingCallbackAfterSendMsg(c)
    if err != nil {
        apiresp.GinError(c, err)
        return
    }

    // 2. If the user receiving the message is a customer service bot, return the message.
    // 2.1 UserID of the robot account
    robotics := "robotics"

    // 2.2 ChatRobot account validation and determining if messages are text and images
    if msgInfo.SendID == robotics || msgInfo.RecvID != robotics {
        return
    }
    if msgInfo.ContentType != constant.Picture && msgInfo.ContentType != constant.Text {
        return
    }

    // 2.3 Get administrator token
    adminToken, err := getAdminToken(c)
    if err != nil {
        apiresp.GinError(c, err)
        return
    }

    // 2.4 Get RobotAccount info
    robUser, err := getRobotAccountInfo(c, adminToken.AdminToken, robotics)
    if err != nil {
        apiresp.GinError(c, err)
        return
    }

    // 2.5 Constructing the contents of the message field or invoking an LLM to implement AI-driven question answering.
    mapStruct, err := contextToMap(c, msgInfo)
    if err != nil {
        apiresp.GinError(c, err)
        return
    }

    // 2.6 Send Message
    err = sendMessage(c, adminToken.ImToken, robotics, msgInfo, robUser, mapStruct)
    if err != nil {
        apiresp.GinError(c, err)
        return
    }
}

4. Effect Display​

PC Web Interface