Skip to main content
Version: 2.0.0



An application that allows users to visually create and manage multiple agents. Within each agent, users can also build and maintain different conversation flows.


A virtual staff that can be used online independently and is capable of understanding natural human language through analysis. It takes the form of daily conversations to assist enterprises with various real business problem resolutions in an efficient, rigorous and low-cost manner.


What the user wants to accomplish. For example, getting weather conditions, getting ways to handle matters, completing matters, etc.


An element of information in natural language, consisting of words or phrases. It includes the name of a person, organization, geographic location, time, date, etc.


The variable that stores the key information needed for intent, which can be inherited in a conversation, and the agent gives subsequent actions and feedback based on the value in the word slot.


A complete task that the agent accomplishes in conversations with the user. It can solve an independent need of the user through one or more rounds of conversation.


It is used in single-round questioning scenarios. A FAQ consists of a user question and an agent answer.

Dailogue Tree

A dialogue tree consists of hierarchical tree branches. A response is given based on the user's question and branch selection in every round of dialogue.

Dailogue Flow

It is used for more complex multi-round task scenarios, including but not limited to complex dialogue flows, data interaction with external interfaces, etc.


Channels are the medium through which agents interact with end users.

After creating a channel, you can use the send and receive message interface provided by the open platform to make calls and integrate the conversation agent capabilities in your own conversation scenarios for end-users to use.


It is a record of a conversation between a user and an agent with a complete contextual relationship - a session starts when the user initiates a message to the agent, and ends when the user does not continue sending messages within 15 minutes of the last message.


Logs show details of each message sent by the user and the corresponding agent's reply, sorted in reverse chronological order of when the message occurred. In another word, the later occurred message comes first.


Developers can train the agent with the real conversation records after it goes live, helping the agent to continuously optimize its knowledge system and make itself "smarter".