Dialogue Boxes
1. Robot conversations
After configuring basic information, Knowledge Base, plugins, and dialogue flow, you can ask questions in the dialog box of the robot homepage.
1.1 Function of dialog box
Input and output methods: In addition to text input and text output, it also supports voice input and voice reading output.
Plugin call status display: After the plugin is called, the method and successful call status of the plugin will be displayed.
Reply content feedback: You can give feedback by liking or disliking the reply result below.
Other functions: support "copy", "view source", "voice playback". In addition, the time taken to reply and token consumption will be displayed.
Support "knowledge reference": Click the knowledge reference box below the reply to view the specific knowledge reference content.
Support “PDF file citation preview, highlighting”：When user ask questions concerning knowledge points in the PDF file,the robot will clearly show the source of the citation, click on PDF file, all citation can be previewed, quoted sections will be highlighted. Besides, the PDF file can also display the page number (e.g. “1 | n page”), to ensure the traceability of knowledge.
Support “Interrupt Generation”: Click the stop button below the reply to terminate the bot reply.
1.2 Chat Management
Click the "New Session" button at the top of the dialog box to start a new session.
You can view the conversation history in the upper left corner of the dialog box and continue the conversation based on the historical conversation.
2. Dialogue history
2.1 Display by item
In this mode, the conversation history can be displayed in a single bar.
In the lower right corner, you can perform operations such as "administrator evaluation", "conversation correction", "view correction history", and "remove top" for a single conversation history.
If the user clicks on it, it will be pinned to the top for the robot manager to follow up or correct the badcase.
In the lower left corner, click "View Context" to view the context of the conversation, and then assist in judgment by replying with the context.
2.2 Display by chat
In this mode, the conversation history can be displayed according to a complete conversation.
On the right side, you can view user feedback and user reviews made by users in this session dimension.
The robot manager can also evaluate and provide feedback to the entire conversation in this mode.
Click Chat History to view the conversation content of the conversation on the right. The administrator can perform feedback and evaluation by conversation here, which is suitable for the inspection of the conversation effect.
3. Data trends
3.1 Visit Trends
Homepage - "Visit Trends" is a data lake visualization display of the robot's Q & A status, which facilitates managers to follow up on daily data statistics and trend insights in real time.
Currently, you can switch to view 5 time intervals: today, yesterday, last 7 days, last 30 days, and a total of 5.
Support viewing of 5 indicator dimensions, see table below for details.
Indicator name
Meaning of indicators
Indicator calculation
Number of questions & number of replies
Number of questions: The number of times the user asked questions.Reply count: The number of normal replies by the robot, excluding system abnormalities.
Normal reply rate: reply times/question times * 100%
Number of replies & number of effective replies
Reply count: The number of normal replies by the robot, excluding system abnormalities.Effective reply count: The number of replies that users do not click on (including the number of times they do not comment and like).
Effective reply rate: number of effective replies/number of replies * 100%.
Number of visitors & number of questions
Number of visitors: The robot's unique viewerNumber of people asking questions: The number of users who asked questions to the bot.
Number of questions per person: Number of questions asked/Number of visitors * 100%
Number of Replies & Number of Error Replies
Reply count: The number of normal replies by the robot, excluding system abnormalities.Number of incorrect replies: The number of conversations rated as incorrect by the administrator.
Error reply rate: number of error replies/number of replies * 100%
Number of Replies & Number of Correct Replies
Reply count: The number of normal replies by the robot, excluding system abnormalities.Number of correct replies: The number of conversations rated as correct by the administrator.
Correct reply rate: Number of correct replies/Number of replies * 100%
3.2 Response time
Display the time consumption details and fluctuation trend of dialogue requests within a certain time range.
Last updated