GB/Z 185.6-2026Artificial intelligence - Agent interconnection - Part 6: Agent interaction (English PDF)

Foreword

GB/Z 185.6-2026: Artificial intelligence - Agent interconnection - Part 6: Agent interaction

GB /Z 185.6-2026.Artificial Intelligence - Interconnection of Intelligent Agents - Part

6.Interaction of Intelligent Agents GB /Z 185.6-2026 English version. Artificial intelligence - Agent interconnection - Part

6.Agent interaction National Standardization Guiding Technical Documents of the People's Republic of China ICS 35.100CCS L 79 AI-powered intelligent agent interconnection Part

6 Modes of Intelligent Agent Interaction

6.1 Overview Intelligent agent interaction modes mainly include three types. point-to-point mode, group mode, and hybrid mode. During a complete interaction, please... The agent selects a suitable mode to interact with the service agent based on the collaboration requirements. Typically, only one mode is used for the same interaction task. Interaction patterns. Furthermore, agents can assume different roles in different interaction tasks (such as requesting agents or serving agents).

6.2 Point-to-point mode The peer-to-peer mode refers to a scenario where, during a single interaction between intelligent agents, the requesting agent directly interacts with each service agent on a one-to-one basis. Mutual, as shown in Figure 1.

6.3 Group Mode Group mode refers to a scenario where, during a single agent interaction, the requesting agent interacts with multiple service agents via a message distribution module. The interaction is shown in Figure

2.In this mode, all agents within the same group (including requesting agents and service agents) can interact through... The message distribution module sends and receives messages. The creation, configuration, and management of groups can be initiated and maintained by the requesting agent, or pre-configured by a third-party organization. Requesting agent You can directly invite service agents to join the group, or review and process application requests submitted by service agents. Third-party organizations can also act as intermediaries. Invite or review applications to join the group.

6.4 Hybrid Mode Hybrid mode refers to a situation where, in a single agent interaction task, the interaction between the requesting agent and the service agent simultaneously exists in a point-to-point mode. In group mode, as shown in Figure 3, the interaction mode in which the service agent participates is determined by the requesting agent. The same service agent... An agent can participate in both point-to-point and group interactions simultaneously (as shown by agent n in Figure 3). Figure

7 Content elements and relationships of intelligent agent interaction

7.1 Content Elements of Intelligent Agent Interaction The content elements used in the interaction process of intelligent agents include, but are not limited to, the following.

a) Data, the smallest unit of content in intelligent agent interaction, can contain various data types, and its content varies depending on the type of information being transmitted. There are some differences; intelligent agents can determine the data type and specific content through various fields in the data.

b) A message is the basic unit of content in intelligent agent interaction, consisting of a sender (which can be a requesting intelligent agent or a service intelligent agent) and various identifiers. It consists of one or more data.

c) Task. The basic unit of work in agent interaction, with a well-defined lifecycle and state. (Requesting the agent to take responsibility) The creation and termination of tasks are handled by the service agent, which is responsible for the specific execution of tasks and the feedback of results. During task execution, the agent... Multiple rounds of message exchange are possible between them.

d) Sessions. Each session corresponds to a multi-agent interaction process, created by the requesting agent, and has a unique identifier for management. This process involves service agents, messages, and tasks.

7.2 Relationships between content elements The relationship between the participating roles and content elements in intelligent agent interaction is shown in Figure 4.

7.3 Data The data used during the interaction of intelligent agents should include, but is not limited to, the fields in Table 2.

7.4 Message The messages used during intelligent agent interaction should include, but are not limited to, the fields in Table 3.

7.6 Session The agent is requested to create and manage sessions based on task requirements. The sessions should include, but are not limited to, the fields in Table 5.

8 Intelligent Agent Interaction Process

8.1 Overview The requesting agent and the service agent can only interact after completing two-way agent identity authentication.

8.2 Point-to-point interaction process The reference interaction flow for the peer-to-peer mode is shown in Figure

5.Three typical implementation methods for this mode are listed in Appendix A; users can choose the method best suited to their specific needs. Choose the appropriate method based on the application scenario.

8.3 Group Mode Interaction Flow The request agent and multiple service agents can interact in a group mode, which enables group interaction through the message distribution module. The examination process is shown in Figure 6. Figure

6.Group Mode Reference Interaction Flow After receiving a message from an agent within a group, the message distribution module is responsible for distributing that message to all agents within the group except the one that sent it. All other participating agents besides the sender. The typical distribution process is shown in Figure 6, with steps such as 8, 11, and 13. Steps 17 and 22).