Product Introduction

1. Product positioning

Agent Studio is an AI agent platform based on large language models—users do not need coding skills, they can easily create intelligent agents by simply setting roles and entering knowledge. It also supports publishing to external channels, helping you provide intelligent conversation services to external customers.

2. How to achieve customization

Agent Studio can configure the role information of robots, build Knowledge Base, arrange dialogue flow, use plugins to query data, and search online. Therefore, robots with different personalities, experiences, and abilities can be created.

2.1 Flexible role setting

Robot role information can be set, defining robots as various character roles or job roles, such as assistants, customer service, consultants, etc. It can not only be described through natural language, but also supports adding variables to dynamically fill in the content of role settings.

2.2 Support for Robot Exclusive Knowledge Base and Public Knowledge Base

By importing knowledge into the robot's exclusive Knowledge Base, the robot can quickly learn and train your knowledge. If you want your knowledge to be reused by multiple robots, you can choose to bind a public Knowledge Base.

2.3 Provide the ability to configure plugins independently

Provide the ability to independently configure plugins, bind your system data interface with the robot in the form of plugins, and your robot will automatically determine whether to call the plugin in the conversation scenario.

2.4 Dialog flow choreography capability

Provide the ability to arrange dialogue processes, allowing your robot to accurately identify the business intentions proposed by customers, ask questions according to the dialogue steps you have arranged, proactively guide the conversation, and thus have the ability to execute tasks step by step, such as "ordering food", "invoicing", "querying order status", etc., with a wide range of applications.

2.5 Provide an explanation for the answer

If the questions asked have relevant reference knowledge under the dialog box, you can view the knowledge source for the answers replied by the robot, making the reply content traceable.

2.6 Multi-end point release channels

Help you publish robots to different channels, including web page sharing, front-end JS embedding, IM systems such as DingTalk, API interface docking, etc. More release channels will be launched in the future.

2.7 Viewing and inspecting conversation history

You can view the conversation history in two ways: single Q & A and complete conversation. You can view the user's likes/dislikes when using it, and also support the robot management personnel to re-evaluate the reply effect of the conversation history, which is convenient for business personnel to mark and inspect the reply effect.

2.8 View data

You can view "Access Trends" and "Response Time" to grasp information such as the level of access and request time of the robot to external services. Based on the data report, you can further analyze user usage, optimize robot information configuration and Knowledge Base composition, and improve the accuracy and efficiency of robot replies.