Multi-agent cooperation management method, device, equipment, medium and program product
By configuring communication rules and resource scheduling through the intelligent agent management platform, dynamically scheduling task processes, and finely managing tool calls, the problems of resource waste and security risks in multi-agent collaborative work are solved, and efficient and flexible multi-agent collaborative processing is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-03-13
AI Technical Summary
Existing intelligent agent management platforms employ a single mode of multi-agent collaboration, which is insufficient to meet the flexible collaboration needs in complex scenarios. The management of tool calls is also rudimentary, leading to resource waste and security risks.
The intelligent agent management platform responds to users' custom process requests, configures the communication rules and resource scheduling of intelligent agents, dynamically schedules task processes, finely manages tool call permissions, and analyzes business needs through natural language understanding of intelligent agents to optimize tool function parameters.
It improves the efficiency of multi-agent collaborative task processing, simplifies user operations, ensures the flexibility and security of task processing, and avoids resource waste and tool call errors.
Smart Images
Figure CN121664804A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a multi-agent collaborative management method, apparatus, device, medium, and program product. Background Technology
[0002] In existing agent management platforms, each agent typically operates independently, and the ways in which agents collaborate are relatively limited. For example, some platforms only support simple sequential agent startup, that is, starting different agents in a preset order, and some platforms only provide each agent with a basic tool library, allowing each agent to directly call the tools in the tool library to complete tasks after startup.
[0003] It is evident that the collaborative working method of multi-agents is relatively simple and cannot meet the needs of flexible collaboration among multi-agents in complex scenarios. Furthermore, the tool call management is crude, which can easily lead to problems such as chaotic tool calls, waste of resources, and even security risks. For example, a certain tool may be called frequently, affecting the overall performance of the system, or sensitive tools may be called at will, thereby affecting the efficiency of task processing. Summary of the Invention
[0004] The purpose of this application is to provide a multi-agent collaborative management method, apparatus, device, medium, and program product to achieve the technical effect of improving the efficiency of multi-agent collaborative task processing.
[0005] In a first aspect, embodiments of this application provide a multi-agent collaborative management method, applied to an agent management platform, wherein multiple agents are deployed on the agent management platform; the method includes: In response to a user-initiated process customization request, determine the task collaboration process of at least some of the multiple intelligent agents; Configure communication rules for each agent in at least some of the agents; wherein the configuration items of the communication rules include interface specifications and communication protocols; Based on the task collaboration process and the current resource status of the agent management platform, the execution tasks of each agent are scheduled.
[0006] In the above implementation process, by utilizing the agent management platform, in response to user-initiated process customization requests, the task collaboration process of at least some of the agents deployed on the platform is determined, the communication rules of each agent in at least some of the agents are configured, and the execution tasks of each agent are scheduled according to the task collaboration process and the current resource status of the agent management platform. This enables users to customize the task collaboration process between multiple agents for complex business needs, and dynamically schedule tasks based on the task collaboration process and the current resource status of the platform, thereby ensuring flexible and secure guidance for multi-agent collaborative work and effectively improving the efficiency of multi-agent collaborative task processing.
[0007] Furthermore, the process customization request includes process creation instructions input by the user on the visual interface of the agent management platform, and the process creation instructions include one or more of the following: agent position adjustment instructions, agent topology connection instructions, and agent task allocation instructions; The process of determining the task collaboration flow of at least some of the multiple intelligent agents in response to a user-initiated process customization request includes: Execute the process creation instruction to create the task collaboration process.
[0008] In the above implementation process, by allowing users to initiate process customization requests by inputting process creation instructions on the visual interface of the agent management platform, and by utilizing the agent management platform to respond to the process customization requests, execute the process creation instructions, and create task collaboration processes, it is possible to support users to customize and create task collaboration processes on the visual interface of the agent management platform, simplifying users' process customization operations, thereby further improving the efficiency of multi-agent collaborative task processing.
[0009] Furthermore, the process customization request includes business requirement statements entered by the user on the visual interface of the intelligent agent management platform, and the multiple intelligent agents include natural language understanding intelligent agents; The process of determining the task collaboration flow of at least some of the multiple intelligent agents in response to a user-initiated process customization request includes: The natural language understanding agent analyzes the business requirement statement to determine the business type. The task collaboration process is determined by selecting any one of the predefined default collaboration processes that matches the business type.
[0010] In the above implementation process, by allowing users to initiate process customization requests by inputting business requirement statements on the visual interface of the agent management platform, the agent management platform responds to the process customization requests and analyzes the business requirement statements through natural language understanding agents deployed on the platform to determine the business type. Any default collaboration process that matches the business type from among multiple predefined default collaboration processes is determined as the task collaboration process. This enables users to automatically query the business-suitable task collaboration process after inputting business requirement statements on the visual interface of the agent management platform, simplifying the user's process customization operation and further improving the efficiency of multi-agent collaborative task processing.
[0011] Furthermore, scheduling the execution tasks of each intelligent agent based on the task collaboration process and the current resource status of the intelligent agent management platform includes: Based on the task collaboration process, the task scheduling order among the execution tasks of each intelligent agent is determined; The tasks of each agent are traversed according to the task scheduling order, and resources are allocated to the target agent corresponding to the current task based on the current resource status. After resources are allocated to the target intelligent agent and the target intelligent agent completes processing the currently executing task, the resources are reclaimed from the target intelligent agent. After determining that no resources will be allocated to the target agent, the execution tasks of each agent will continue to be traversed until the execution tasks of each agent are completed.
[0012] In the above implementation process, the intelligent agent management platform determines the task scheduling order between the execution tasks of each intelligent agent according to the task collaboration process. It then traverses the execution tasks of each intelligent agent according to the task scheduling order. Based on the current resource status of the platform, it determines whether to allocate resources to the target intelligent agent corresponding to the current execution task. After allocating resources to the target intelligent agent and the target intelligent agent completes the processing of the current execution task, the resources are reclaimed from the target intelligent agent. If it is determined not to allocate resources to the target intelligent agent, the execution tasks of each intelligent agent continue to be traversed until the execution tasks of each intelligent agent are completed. This process takes into account the dynamically changing resource status of the platform and dynamically schedules tasks based on the task collaboration process and the current resource status of the platform. This ensures flexible and safe guidance for multi-agent collaborative work and effectively improves the efficiency of multi-agent collaborative task processing.
[0013] Furthermore, after allocating resources to the target agent, the process also includes: Through the target intelligent agent, target tools are invoked from a pre-established tool library according to the tool invocation permissions of the target intelligent agent; wherein, the tool invocation permissions of the target intelligent agent are set according to one or more of the target intelligent agent's role and the type of the currently executed task, and the tool library includes one or more of data processing tools, knowledge retrieval tools, and image recognition tools; The target intelligent agent uses the target tool to process the currently executing task.
[0014] In the above implementation process, by pre-establishing a tool library and pre-setting the tool calling permissions of the target agent, after resources are allocated to the target agent, the target agent can call the target tool from the tool library according to the target agent's tool calling permissions and use the target tool to process the currently executing task. This enables fine-grained management of the agent's tool calling permissions and avoids problems such as resource waste caused by the agent erroneously calling tools to process tasks, thereby further improving the efficiency of multi-agent collaborative task processing.
[0015] Furthermore, the method also includes: Detect whether the target agent abnormally calls the target tool; If the target agent abnormally invokes the target tool, an alarm message will be sent.
[0016] In the above implementation process, the agent management platform detects whether the target agent is abnormally calling the target tool when the target agent calls the target tool to handle the current task. If an abnormal call to the target tool is detected, an alarm message is sent. This can promptly notify the user when the agent abnormally calls the tool, effectively ensuring the stability and security of the overall platform operation.
[0017] Furthermore, detecting whether the target agent abnormally invokes the target tool includes: Evaluate the target agent's invocation metrics for the target tool; wherein the invocation metrics are used to characterize the target agent's invocation of the target tool; If the value of the calling indicator meets the abnormal value condition, it is determined that the target agent is abnormally calling the target tool; otherwise, it is determined that the target agent is normally calling the target tool.
[0018] In the above implementation process, the intelligent agent management platform evaluates the target intelligent agent's call index to the target tool. If the value of the call index meets the abnormal value condition, it is determined that the target intelligent agent is abnormally calling the target tool. Otherwise, it is determined that the target intelligent agent is normally calling the target tool. Based on the actual call situation of the intelligent agent to the tool, it can quickly and accurately detect whether the intelligent agent is abnormally calling the tool.
[0019] Furthermore, the method also includes: For each tool in the tool library, optimize the functional parameters of the current tool based on its historical call events and agent feedback data.
[0020] In the above implementation process, by utilizing the agent management platform to optimize the functional parameters of each tool in the tool library based on the historical call events and agent feedback data of the current tool, it is possible to support adaptive optimization of the functions of each tool in the tool library, thereby further improving the efficiency of multi-agent collaborative task processing.
[0021] Secondly, embodiments of this application provide a multi-agent collaborative management device, applied to an agent management platform, wherein multiple agents are deployed on the agent management platform; the device includes: The collaboration process determination module is used to determine the task collaboration process of at least some of the multiple intelligent agents in response to a user-initiated process customization request. A communication rule configuration module is used to configure the communication rules of each agent in the at least some of the agents; wherein, the configuration items of the communication rules include interface specifications and communication protocols; The task scheduling module is used to schedule the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform.
[0022] Thirdly, embodiments of this application provide an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; when the processor executes the computer program, it implements the method described above.
[0023] Fourthly, embodiments of this application provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the method described above.
[0024] Fifthly, embodiments of this application provide a computer program product, the computer program product including instructions, which, when executed by a computer, cause the computer to perform the method described above. Attached Figure Description
[0025] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 A flowchart illustrating a multi-agent collaborative management method provided in the first embodiment of this application; Figure 2 This is a schematic diagram illustrating a task collaboration process exemplified in the first embodiment of this application; Figure 3 This is a schematic diagram of the structure of a multi-agent cooperative management device provided in the second embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in the third embodiment of this application. Detailed Implementation
[0027] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.
[0028] It should be noted that in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. Furthermore, the step numbers in the text are only for the convenience of explaining the embodiments of this application and are not intended to limit the order in which the steps are performed.
[0029] In related technologies, many agent management platforms typically operate their agents independently, and the ways in which agents collaborate are relatively limited. For example, some platforms only support simple sequential agent startup, that is, starting different agents in a preset order, and some platforms only provide each agent with a basic tool library, allowing each agent to directly call the tools in the tool library to complete tasks after startup.
[0030] It is evident that the collaborative working method of multi-agents is relatively simple and cannot meet the needs of flexible collaboration among multi-agents in complex scenarios. Furthermore, the tool call management is crude, which can easily lead to problems such as chaotic tool calls, waste of resources, and even security risks. For example, a certain tool may be called frequently, affecting the overall performance of the system, or sensitive tools may be called at will, thereby affecting the efficiency of task processing.
[0031] To address this, this application proposes a multi-agent collaborative management method. By utilizing an agent management platform, in response to user-initiated process customization requests, the method determines the task collaboration process of at least some of the agents deployed on the platform, configures the communication rules for each agent, and schedules the execution tasks of each agent based on the task collaboration process and the current resource status of the agent management platform. This method supports users in customizing the task collaboration process between multiple agents for complex business needs, and dynamically schedules tasks based on the task collaboration process and the current resource status of the platform, thereby ensuring flexible and secure guidance for multi-agent collaborative work and effectively improving the efficiency of multi-agent collaborative task processing.
[0032] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.
[0033] The methods provided in this application can be executed by relevant terminal devices, and the following descriptions all use user terminals as the execution subject.
[0034] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a multi-agent collaborative management method provided in the first embodiment of this application. The multi-agent collaborative management method is applied to an agent management platform, on which multiple agents are deployed; the method includes steps S101-S102: S101. In response to a user-initiated process customization request, determine the task collaboration process of at least some of the multiple agents.
[0035] As an example, multiple agents are pre-deployed on the agent management platform, such as natural language understanding agents, technical answering agents, case retrieval agents, and business policy interpretation agents.
[0036] In practical applications, multiple pre-established and trained agents from different business domains can be directly selected. The process of establishing and training agents will not be elaborated here.
[0037] Users can submit custom process requests to the intelligent agent management platform based on their actual business needs.
[0038] Upon receiving a user-initiated process customization request, the agent management platform responds to the request by determining the task collaboration process for at least some of the multiple agents.
[0039] In practical applications, at least some intelligent agents are usually multiple, and the structure of the task collaboration process can include one or more of serial, parallel, and conditional branching.
[0040] S102. Configure communication rules for each agent in at least some of the agents; wherein the configuration items for the communication rules include interface specifications and communication protocols.
[0041] As an example, after determining the task collaboration process of at least some of the intelligent agents, the intelligent agent management platform configures the communication rules of each intelligent agent by configuring the interface specifications and communication protocols of each intelligent agent in these at least some intelligent agents.
[0042] In practical applications, the input and output interface specifications of each intelligent agent can be configured. For example, the format of the interaction data of each intelligent agent can be specified as JSON (JavaScript Object Notation, a lightweight data exchange format). The network communication protocol of each intelligent agent can also be configured, such as specifying the network communication protocol of each intelligent agent as HTTP (HyperText Transfer Protocol) / HTTPS (HyperText Transfer Protocol over Secure Socket Layer).
[0043] By configuring the communication rules for each agent, real-time interaction between agents can be supported, ensuring that interactive data is accurately transmitted and processed between agents.
[0044] S103. Based on the task collaboration process and the current resource status of the agent management platform, schedule the execution tasks of each agent.
[0045] As an example, the intelligent agent management platform obtains its own current resource status in real time, and schedules the execution tasks of each intelligent agent according to the task collaboration process and the current resource status, until the execution tasks of each intelligent agent are completed.
[0046] This application embodiment utilizes an intelligent agent management platform to respond to user-initiated process customization requests, determine the task collaboration process of at least some of the multiple intelligent agents deployed on the platform, configure the communication rules of each intelligent agent, and schedule the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform. This enables users to customize the task collaboration process between multiple intelligent agents for complex business needs, and dynamically schedule tasks based on the task collaboration process and the current resource status of the platform, thereby ensuring flexible and secure guidance for multi-agent collaborative work and effectively improving the efficiency of multi-agent collaborative task processing.
[0047] In an optional embodiment, the process customization request includes a process creation instruction entered by the user on the visual interface of the agent management platform. The process creation instruction includes one or more of the following: agent position adjustment instruction, agent topology connection instruction, and agent task allocation instruction. The step of responding to the process customization request initiated by the user and determining the task collaboration process of at least some of the multiple agents includes: executing the process creation instruction and creating the task collaboration process.
[0048] As an example, in order to simplify users' process customization operations, the agent management platform can render a visual interface on the front end, which displays multiple graphical elements, each of which represents an agent deployed on the agent management platform.
[0049] Users can input process creation commands on the visual interface of the agent management platform according to actual business needs. These commands include one or more of the following: agent position adjustment commands, agent topology connection commands, and agent task allocation commands.
[0050] For example, a user drags graphical element A to a specified position on the visual interface of the agent management platform and enters a position adjustment command for agent A; a user connects graphical element A and graphical element B on the visual interface of the agent management platform and enters a topology connection command for agent A and agent B; a user edits the task input box of graphical element A on the visual interface of the agent management platform and enters a task assignment command for agent A; a user clicks the submit control on the visual interface of the agent management platform to initiate a process customization request to the agent management platform.
[0051] After receiving a user's request to customize a workflow, the intelligent agent management platform responds by extracting workflow creation instructions from the request and executing them to create a task collaboration workflow.
[0052] This application embodiment allows users to initiate a process customization request by inputting process creation instructions on the visual interface of the agent management platform. The agent management platform responds to the process customization request, executes the process creation instructions, and creates a task collaboration process. This enables users to customize and create task collaboration processes on the visual interface of the agent management platform, simplifies the user's process customization operation, and further improves the efficiency of multi-agent collaborative task processing.
[0053] In an optional embodiment, the process customization request includes a business requirement statement entered by the user on the visual interface of the agent management platform, and the multiple agents include a natural language understanding agent; the step of responding to the process customization request initiated by the user and determining the task collaboration process of at least some of the multiple agents includes: analyzing the business requirement statement through the natural language understanding agent to determine the business type; and determining any one of the multiple predefined default collaboration processes that matches the business type as the task collaboration process.
[0054] As an example, in order to simplify users' process customization operations, natural language understanding agents can be pre-deployed on the agent management platform, and the agent management platform can render a visual interface on the front end, which displays a business requirement input box.
[0055] Users can input business requirement statements on the visual interface of the intelligent agent management platform according to their actual business needs, and then click the submit control on the visual interface of the intelligent agent management platform to initiate a process customization request to the intelligent agent management platform.
[0056] It should be noted that the business requirement statement is used to describe the current business requirements.
[0057] After receiving a user's request to customize a process, the intelligent agent management platform responds by extracting business requirement statements from the request and inputting them into the natural language understanding intelligent agent.
[0058] Natural language understanding agents analyze business requirement statements to determine the business type.
[0059] After obtaining the business type output by the natural language understanding agent, the intelligent agent management platform retrieves multiple predefined default collaboration processes and selects any default collaboration process that matches the business type as the task collaboration process.
[0060] For example, assuming the business requirement statement is "handle a complex customer inquiry question Q", the natural language understanding agent first analyzes the business requirement statement to determine the business type, and then determines the task collaboration process based on the business type. For instance, if the business type is a technical problem-solving business, the default collaboration process of parallel collaboration between the technical problem-solving agent and the case retrieval agent is selected from multiple default collaboration processes as the task collaboration process. If the business type is a business problem-solving business, the default collaboration process of the business policy interpretation agent is selected from multiple default collaboration processes as the task collaboration process.
[0061] This application embodiment allows users to initiate process customization requests by inputting business requirement statements on the visual interface of the agent management platform. The agent management platform responds to these requests by analyzing the business requirement statements using natural language understanding agents deployed on the platform, determining the business type, and identifying any one of the predefined default collaboration processes that matches the business type as the task collaboration process. This enables users to automatically query the appropriate task collaboration process after inputting business requirement statements on the visual interface of the agent management platform, simplifying the process customization operation and further improving the efficiency of multi-agent collaborative task processing.
[0062] In an optional embodiment, scheduling the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform includes: determining the task scheduling order among the execution tasks of each intelligent agent according to the task collaboration process; traversing the execution tasks of each intelligent agent according to the task scheduling order, and determining whether to allocate resources to the target intelligent agent corresponding to the current execution task according to the current resource status; after allocating resources to the target intelligent agent and the target intelligent agent completing the processing of the current execution task, reclaiming the resources from the target intelligent agent; and after determining not to allocate resources to the target intelligent agent, continuing to traverse the execution tasks of each intelligent agent until the processing of the execution tasks of each intelligent agent is completed.
[0063] As an example, after determining the task collaboration process of at least some of the intelligent agents, the intelligent agent management platform determines the task scheduling order among the execution tasks of each intelligent agent according to the task collaboration process.
[0064] In practical applications, the task scheduling order between the tasks of each intelligent agent can be determined based on the task collaboration process, taking into account factors such as the topological hierarchy of the process nodes where the tasks of each intelligent agent are executed, and the task priority of each intelligent agent's tasks. The task priority can be set based on business rules, such as user identity level, task urgency, or service commitment response time.
[0065] For example, suppose the task collaboration process is as follows: Figure 2 As shown, considering the topological hierarchy of the process node where each agent's execution task is located, and the task priority of each agent's execution task itself, the task scheduling order between the execution tasks of each agent is determined as follows: Agent A's execution task Task1 → Agent B's execution task Task3 → Agent A's execution task Task2.
[0066] After determining the task scheduling order, the intelligent agent management platform begins to dynamically schedule tasks, and obtains the current resource status of the intelligent agent management platform itself in real time, such as CPU utilization, memory usage, network bandwidth and concurrent call volume of the target tool, etc. It traverses the execution tasks of each intelligent agent according to the task scheduling order, and determines whether to allocate resources to the currently traversed execution task, that is, the target intelligent agent corresponding to the current execution task, based on the current resource status.
[0067] In practical applications, the agent management platform can determine whether to allocate appropriate resources to the target agent by judging whether the current resource status supports the target agent in handling the current task.
[0068] If resources are allocated to the target agent, the target agent can use these resources to process the currently executing task. If resources are not allocated to the target agent, the target agent will skip processing the currently executing task.
[0069] After resources are allocated to the target agent and the target agent completes the currently executed task, the resources are reclaimed from the target agent in a timely manner for subsequent use.
[0070] After determining not to allocate resources to the target agent, continue iterating through the execution tasks of each agent until the execution tasks of each agent are completed.
[0071] Understandably, after completing the execution tasks of each agent, if some execution tasks were previously skipped, then those tasks need to be traversed again until all execution tasks are completed.
[0072] By scheduling the execution tasks of each intelligent agent based on the task collaboration process and the current resource status of the intelligent agent management platform, the static task collaboration process can be transformed into a dynamic resource allocation and execution plan that comprehensively considers real-time resource bottlenecks, task criticality, and execution dependencies, thereby achieving overall optimization of platform throughput efficiency and response speed in complex business scenarios.
[0073] This application embodiment determines the task scheduling order between the execution tasks of each intelligent agent according to the task collaboration process by an intelligent agent management platform. It then traverses the execution tasks of each intelligent agent according to the task scheduling order, determines whether to allocate resources to the target intelligent agent corresponding to the current execution task based on the platform's current resource status, and reclaims the resources from the target intelligent agent after allocating resources and the target intelligent agent has completed processing the current execution task. If it is determined not to allocate resources to the target intelligent agent, the traversal of the execution tasks of each intelligent agent continues until the execution tasks of each intelligent agent are completed. This approach takes into account the dynamically changing resource status of the platform and dynamically schedules tasks based on the task collaboration process and the platform's current resource status, thereby ensuring flexible and secure guidance for multi-agent collaborative work and effectively improving the efficiency of multi-agent collaborative task processing.
[0074] In an optional embodiment, after allocating resources to the target agent, the method further includes: calling a target tool from a pre-established tool library based on the target agent's tool calling permissions; wherein the target agent's tool calling permissions are set according to one or more of the target agent's role and the type of the currently executed task, and the tool library includes one or more of data processing tools, knowledge retrieval tools, and image recognition tools; and using the target tool to process the currently executed task.
[0075] As an example, after the agent management platform allocates resources to the target agent, the target agent can use these resources to process the currently executing task.
[0076] To improve the efficiency of intelligent agents in processing tasks, a tool library is pre-established, which includes one or more of the following: data processing tools, knowledge retrieval tools, and image recognition tools.
[0077] In practical applications, the tool library can cover a variety of types and functions of tools, and provides detailed descriptions of each tool's function and category. It also supports tool updates and expansions, allowing users to add or update tools to the library themselves.
[0078] To prevent the target agent from accidentally calling tools during task processing, the target agent is pre-assigned corresponding tool calling permissions. The tool calling permissions of the target agent are set according to one or more of the target agent's role and the type of the currently executed task.
[0079] For example, if the target agent has high-level tool access permissions, it is allowed to access tools that involve sensitive data processing; if the target agent has normal tool access permissions, it is restricted to accessing only basic data analysis tools.
[0080] The target agent invokes target tools from a pre-established tool library based on its own tool invocation permissions, and uses the target tools to process the currently executing task.
[0081] This application embodiment, by pre-establishing a tool library and pre-setting the tool calling permissions of the target intelligent agent, enables the target intelligent agent to call the target tool from the tool library according to the tool calling permissions after resources are allocated to the target intelligent agent, and use the target tool to process the currently executed task. It can finely manage the tool calling permissions of the intelligent agent and avoid problems such as resource waste caused by the intelligent agent calling the wrong tool to process the task, thereby further improving the efficiency of multi-agent collaborative task processing.
[0082] In an optional embodiment, the method further includes steps S104-S105: S104. Detect whether the target agent abnormally calls the target tool; S105. If the target intelligent agent abnormally calls the target tool, send an alarm message.
[0083] As an example, the agent management platform can detect in real time whether the target agent is abnormally calling the target tool when the target agent calls the target tool to handle the currently executed task.
[0084] If an abnormal call to the target tool is detected by the target agent, an alarm message is sent.
[0085] In practical applications, when the intelligent agent management platform detects that the target intelligent agent is abnormally calling the target tool, it can send an alarm message about the abnormal call to the target tool and trigger a predefined exception handling strategy at the same time. This strategy may include modifying the target intelligent agent's tool calling permissions to prevent the target intelligent agent from calling the target tool, or controlling the target intelligent agent to stop running, in order to ensure the stability and security of the overall platform operation.
[0086] Maintain the current state if the target agent calls the target tool normally.
[0087] This application embodiment detects whether the target intelligent agent is abnormally calling the target tool during the process of the target intelligent agent calling the target tool to process the current task. If abnormal calling of the target tool is detected, an alarm message is sent. This can promptly notify the user when the intelligent agent abnormally calls the tool, effectively ensuring the stability and security of the overall platform operation.
[0088] In an optional embodiment, detecting whether the target agent abnormally calls the target tool includes: evaluating the call index of the target agent to the target tool; wherein the call index is used to characterize the call status of the target agent to the target tool; if the value of the call index meets the abnormal value condition, it is determined that the target agent abnormally calls the target tool, otherwise it is determined that the target agent normally calls the target tool.
[0089] As an example, an agent management platform can detect whether a target agent is abnormally calling a target tool in the following ways: First, evaluate the target agent's invocation metrics for the target tool, where the invocation metrics characterize the target agent's invocation of the target tool.
[0090] In practical applications, the call metrics can be selected from one or more of the following: call frequency, call count, call duration, and call result.
[0091] Next, it is determined whether the value of the called indicator meets the pre-set abnormal value conditions.
[0092] In practical applications, assuming that the call metrics include call frequency, call duration, and call result, the abnormal value conditions can be set as follows: the call frequency is less than a preset call frequency threshold, the call duration is less than a preset call duration threshold, and the call result is verified as an erroneous result.
[0093] Finally, if the value of the called indicator meets the abnormal value condition, it is determined that the target agent is abnormally calling the target tool; otherwise, it is determined that the target agent is normally calling the target tool.
[0094] This application embodiment evaluates the target agent's call index to the target tool by the agent management platform. If the value of the call index meets the abnormal value condition, it determines that the target agent is abnormally calling the target tool; otherwise, it determines that the target agent is normally calling the target tool. Based on the actual call situation of the agent to the tool, it can quickly and accurately detect whether the agent is abnormally calling the tool.
[0095] In an optional embodiment, the method further includes step S106: S106. For each tool in the tool library, optimize the functional parameters of the current tool based on the historical call events and agent feedback data.
[0096] As an example, the agent management platform obtains the historical call events and agent feedback data of the current tool for each tool in the tool library.
[0097] It should be noted that the historical invocation events fully record which agent invoked the current tool at what time, what functional parameters were input to the current tool, what task was executed, and what result was obtained. Agent feedback data refers to the data provided by agents that have historically invoked the current tool, including the tool's effectiveness, performance, and improvement suggestions.
[0098] Based on the historical invocation events and agent feedback data of the current tool, optimize the functional parameters of the current tool. For example, analyze the functional parameters input by relevant agents when calling the current tool, cache relevant resources in advance, and reduce invocation latency; or prioritize tools according to invocation frequency and importance, optimize scheduling strategies, and improve the overall efficiency of tool invocation.
[0099] This application embodiment utilizes an agent management platform to optimize the functional parameters of each tool in the tool library based on the tool's historical call events and agent feedback data. This enables adaptive optimization of the functions of each tool in the tool library, thereby further improving the efficiency of multi-agent collaborative task processing.
[0100] Please refer to Figure 3 , Figure 3 This is a schematic diagram of a multi-agent collaborative management device provided in the second embodiment of this application. The second embodiment of this application provides a multi-agent collaborative management device 20, applied to an agent management platform, on which multiple agents are deployed. The device 20 includes: a collaborative process determination module 201, used to determine the task collaborative process of at least some of the agents in response to a user-initiated process customization request; a communication rule configuration module 202, used to configure the communication rules of each agent in at least some of the agents; wherein the configuration items of the communication rules include interface specifications and communication protocols; and an execution task scheduling module 203, used to schedule the execution tasks of each agent according to the task collaborative process and the current resource status of the agent management platform.
[0101] In an optional embodiment, the process customization request includes a process creation instruction entered by the user on the visual interface of the agent management platform. The process creation instruction includes one or more of the following: agent position adjustment instruction, agent topology connection instruction, and agent task allocation instruction. The step of responding to the process customization request initiated by the user and determining the task collaboration process of at least some of the multiple agents includes: executing the process creation instruction and creating the task collaboration process.
[0102] In an optional embodiment, the process customization request includes a business requirement statement entered by the user on the visual interface of the agent management platform, and the multiple agents include a natural language understanding agent; the step of responding to the process customization request initiated by the user and determining the task collaboration process of at least some of the multiple agents includes: analyzing the business requirement statement through the natural language understanding agent to determine the business type; and determining any one of the multiple predefined default collaboration processes that matches the business type as the task collaboration process.
[0103] In an optional embodiment, scheduling the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform includes: determining the task scheduling order among the execution tasks of each intelligent agent according to the task collaboration process; traversing the execution tasks of each intelligent agent according to the task scheduling order, and determining whether to allocate resources to the target intelligent agent corresponding to the current execution task according to the current resource status; after allocating resources to the target intelligent agent and the target intelligent agent completing the processing of the current execution task, reclaiming the resources from the target intelligent agent; and after determining not to allocate resources to the target intelligent agent, continuing to traverse the execution tasks of each intelligent agent until the processing of the execution tasks of each intelligent agent is completed.
[0104] In an optional embodiment, the task scheduling module 203 is further configured to: after allocating resources to the target agent, invoke a target tool from a pre-established tool library according to the target agent's tool invocation permissions; wherein the target agent's tool invocation permissions are set according to one or more of the target agent's role and the type of the currently executed task, and the tool library includes one or more of data processing tools, knowledge retrieval tools, and image recognition tools; and use the target tool to process the currently executed task through the target agent.
[0105] In an optional embodiment, the device further includes a tool call monitoring module, used to: detect whether the target agent abnormally calls the target tool; and send an alarm message when the target agent abnormally calls the target tool.
[0106] In an optional embodiment, detecting whether the target agent abnormally calls the target tool includes: evaluating the call index of the target agent to the target tool; wherein the call index is used to characterize the call status of the target agent to the target tool; if the value of the call index meets the abnormal value condition, it is determined that the target agent abnormally calls the target tool, otherwise it is determined that the target agent normally calls the target tool.
[0107] In an optional embodiment, the device further includes a tool call optimization module, used to optimize the functional parameters of each tool in the tool library based on the historical call events of the current tool and the feedback data from the agent.
[0108] The specific implementation process of the functions and roles of each module in the above-mentioned device 20 can be found in the implementation process of the corresponding steps in the method described in the first embodiment of this application, and will not be repeated here.
[0109] Please refer to Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in the third embodiment of this application. The third embodiment of this application provides an electronic device 30, including a processor 301, a memory 302, and a computer program stored in the memory 302 and configured to be executed by the processor 301; when the processor 301 executes the computer program, it implements the method described in the first embodiment of this application and can achieve the same beneficial effects.
[0110] When the processor 301 reads a computer program from the memory 302 via the bus 303 and executes the computer program, it can implement any of the methods described in the first embodiment of this application.
[0111] Processor 301 can process digital signals and may include various computing architectures. For example, it may be a complex instruction set computer architecture, a reduced instruction set computer architecture, or an architecture that implements multiple instruction set combinations. In some examples, processor 301 may be a microprocessor.
[0112] The memory 302 can be used to store instructions executed by the processor 301 or data related to the execution of instructions. These instructions and / or data may include code for implementing some or all of the functions of one or more modules described in the embodiments of this application. The processor 301 of this disclosure embodiment can be used to execute instructions in the memory 302 to implement the method described in the first embodiment of this application. The memory 302 includes dynamic random access memory, static random access memory, flash memory, optical memory, or other memories well known to those skilled in the art.
[0113] The fourth embodiment of this application provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to perform the method described in the first embodiment of this application, and can achieve the same beneficial effects.
[0114] The method described in the first embodiment of this application can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the various embodiments of this application are executed, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, a core network device, an OAM (Open Application Model), or other programmable devices.
[0115] The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; or an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.
[0116] The fifth embodiment of this application provides a computer program product, which includes instructions that, when executed by a computer, cause the computer to perform the method described in the first embodiment of this application and achieve the same beneficial effects.
[0117] The methods described in the first embodiment of this application can be implemented, in whole or in part, by software, hardware, firmware, or any combination thereof. When implemented in software, they can be implemented, in whole or in part, in the form of a computer program product. A computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the various embodiments of this application are performed, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, a core network device, an OAM (Open Application Model), or other programmable devices.
[0118] Computer programs or instructions can be stored in or transferred from one computer-readable storage medium to another. For example, a computer program or instructions can be transferred from one website, computer, server, or data center to another via wired or wireless means. A computer-readable storage medium can be any usable medium that a computer can access, or a data storage device such as a server or data center that integrates one or more usable media. Usable media can be magnetic media, such as floppy disks, hard disks, and magnetic tapes; optical media, such as digital video discs; or semiconductor media, such as solid-state drives. The computer-readable storage medium can be volatile or non-volatile, or may include both types.
[0119] In summary, this application provides a multi-agent collaborative management method, apparatus, device, medium, and program product. The multi-agent collaborative management method is applied to an agent management platform, on which multiple agents are deployed. The method includes: responding to a user-initiated process customization request, determining a task collaboration process for at least some of the agents; configuring communication rules for each agent; wherein the configuration items of the communication rules include interface specifications and communication protocols; and scheduling the execution tasks of each agent based on the task collaboration process and the current resource status of the agent management platform. This application, by utilizing an agent management platform, responding to a user-initiated process customization request, determining a task collaboration process for at least some of the agents deployed on the platform, configuring communication rules for each agent, and scheduling the execution tasks of each agent based on the task collaboration process and the current resource status of the agent management platform, enables users to customize task collaboration processes between multiple agents for complex business needs. It also allows for dynamic task scheduling based on the task collaboration process and the current resource status of the platform, thereby ensuring flexible and secure guidance for multi-agent collaborative work and effectively improving the efficiency of multi-agent collaborative task processing.
[0120] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0121] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0122] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A multi-agent collaborative management method, characterized in that, The method is applied to an agent management platform, on which multiple agents are deployed; the method includes: In response to a user-initiated process customization request, determine the task collaboration process of at least some of the multiple intelligent agents; Configure communication rules for each agent in at least some of the agents; wherein the configuration items of the communication rules include interface specifications and communication protocols; Based on the task collaboration process and the current resource status of the agent management platform, the execution tasks of each agent are scheduled.
2. The method according to claim 1, characterized in that, The process customization request includes process creation instructions entered by the user on the visual interface of the agent management platform. The process creation instructions include one or more of the following: agent position adjustment instructions, agent topology connection instructions, and agent task allocation instructions. The process of determining the task collaboration flow of at least some of the multiple agents in response to a user-initiated flow customization request includes: Execute the process creation instruction to create the task collaboration process.
3. The method according to claim 1, characterized in that, The process customization request includes business requirement statements entered by the user on the visual interface of the intelligent agent management platform, and the multiple intelligent agents include natural language understanding intelligent agents; The process of determining the task collaboration flow of at least some of the multiple agents in response to a user-initiated flow customization request includes: The natural language understanding agent analyzes the business requirement statement to determine the business type. The task collaboration process is determined by selecting any one of the predefined default collaboration processes that matches the business type.
4. The method according to any one of claims 1 to 3, characterized in that, The step of scheduling the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform includes: Based on the task collaboration process, the task scheduling order among the execution tasks of each intelligent agent is determined; The tasks of each agent are traversed according to the task scheduling order, and resources are allocated to the target agent corresponding to the current task based on the current resource status. After resources are allocated to the target intelligent agent and the target intelligent agent completes processing the currently executing task, the resources are reclaimed from the target intelligent agent. After determining that no resources will be allocated to the target agent, the execution tasks of each agent will continue to be traversed until the execution tasks of each agent are completed.
5. The method according to claim 4, characterized in that, After allocating resources to the target agent, the process further includes: Through the target intelligent agent, target tools are invoked from a pre-established tool library according to the tool invocation permissions of the target intelligent agent; wherein, the tool invocation permissions of the target intelligent agent are set according to one or more of the target intelligent agent's role and the type of the currently executed task, and the tool library includes one or more of data processing tools, knowledge retrieval tools, and image recognition tools; The target intelligent agent uses the target tool to process the currently executing task.
6. The method according to claim 5, characterized in that, The method further includes: Detect whether the target agent abnormally calls the target tool; If the target agent abnormally invokes the target tool, an alarm message is sent.
7. The method according to claim 6, characterized in that, The step of detecting whether the target agent abnormally calls the target tool includes: Evaluate the target agent's invocation metrics for the target tool; wherein the invocation metrics are used to characterize the target agent's invocation of the target tool; If the value of the calling indicator meets the abnormal value condition, it is determined that the target agent is abnormally calling the target tool; otherwise, it is determined that the target agent is normally calling the target tool.
8. The method according to claim 5, characterized in that, The method further includes: For each tool in the tool library, optimize the functional parameters of the current tool based on its historical call events and agent feedback data.
9. A multi-agent collaborative management device, characterized in that, An application in an agent management platform, wherein multiple agents are deployed on the agent management platform; the device includes: The collaboration process determination module is used to determine the task collaboration process of at least some of the multiple intelligent agents in response to a user-initiated process customization request. A communication rule configuration module is used to configure the communication rules of each agent in the at least some agents; wherein the configuration items of the communication rules include interface specifications and communication protocols; The task scheduling module is used to schedule the execution tasks of each intelligent agent according to the task collaboration process and the current resource status of the intelligent agent management platform.
10. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; when the processor executes the computer program, it implements the method according to any one of claims 1 to 8.
11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1 to 8.
12. A computer program product, characterized in that, The computer program product includes instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 8.