Skip to main content

Implementing Chatbot Functionality with OpenIM

Brief Description​

Implement chatbot functionality using the Webhook mechanism in OpenIM. After sending a text message or an image message to the chatbot, the bot will return the same message. Developers can replace this logic by calling the LLM interface to enable intelligent question-answering.

1. Modify the Configuration File​

Refer to the template below to modify the config/webhooks.yml configuration in open-im-server.

PC Web Interface

Tips:

  • url is the callback URL.
  • Set afterSendSingleMsg.enable to true to enable this callback.

2. Create a Chatbot Account​

  1. Log in to the admin console, refer to this document.
  2. Create a chatbot account in user management and note the userID of the account.
  3. For convenience, set this userID as a default friend.

3. Write the afterSendSingleMsg Interface​

Refer to the example code below.

Tips:

  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. Demonstration of Results​

PC Web Interface