A multi-agent cooperation method and device based on a unified task management platform

Abstract

This invention discloses a multi-agent collaboration method and apparatus based on a unified task management platform, relating to the field of artificial intelligence technology. The method includes: constructing and running a unified task management platform to centrally store and manage multi-agent teams, task attributes, and message information, providing agent operation interfaces and user interfaces; configuring long-term, recurring, timed tasks for each agent, with each cycle querying platform information through the interface, autonomously completing task planning, allocation, and processing, and writing back updates; responding to user creation commands and initializing team and task attribute information; enabling each agent to execute timed tasks in parallel, autonomously accepting and processing tasks, synchronizing status through the operation interface, and achieving human-computer interaction until the task is completed, thus achieving efficient multi-agent collaboration. This invention combines a passive platform with active agent polling, separating task execution and state storage, reducing coupling, and solving the problems of poor compatibility and scalability of heterogeneous agents.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a multi-agent collaboration method and apparatus based on a unified task management platform. Background Technology

[0002] With the continuous iteration and widespread application of large language model technology, intelligent agents with large models as their cognitive core are no longer limited to executing single, simple instructions. Instead, they are gradually evolving into complex intelligent execution units with autonomous planning, logical reasoning, experiential memory, and multi-turn interaction capabilities. When a single intelligent agent struggles to handle larger-scale and more logically complex comprehensive tasks, multi-agent collaborative systems composed of multiple heterogeneous and multifunctional intelligent agents have emerged and quickly become a research and application hotspot in the field of artificial intelligence. By integrating intelligent agent resources with different capabilities and architectures, multi-agent collaborative systems achieve collaborative task processing, information sharing, and complementary decision-making. They can be widely applied in software development, content generation, business process automation, knowledge engineering, and complex problem solving, significantly improving the processing efficiency and completion quality of complex tasks.

[0003] Current research on multi-agent collaboration technologies mainly revolves around four major directions: First, task scheduling and process orchestration, which achieves task decomposition, allocation, and execution control by constructing a unified scheduling logic; second, task planning based on large language models, which leverages the reasoning and generation capabilities of large models to complete task decomposition, strategy formulation, and execution plan generation; third, heterogeneous agent integration, which aims to solve the communication and interoperability problems of agents from different frameworks and platforms; and fourth, collaborative optimization control for physical scenarios, which uses algorithms such as reinforcement learning and cluster control to achieve collaborative movement and state consistency control of agents under resource constraints. These technologies provide a basic implementation path for multi-agent collaborative work from the perspectives of scheduling modes, decision-making algorithms, communication mechanisms, and control strategies, promoting the initial implementation of agent collaboration systems.

[0004] However, existing multi-agent collaboration technologies still have significant limitations in practical applications. In terms of collaborative architecture, most systems adopt a scheduling model where a central node actively pushes tasks, relying on a central module to complete task decomposition, instruction issuance, and status monitoring. This requires the central node to maintain real-time communication with each agent and fully perceive their operational status. This results in poor compatibility with agents across ecosystems and heterogeneous architectures, making it difficult to achieve flexible access and plug-and-play functionality for agents from different platforms, and also limiting the dynamic expansion of the collaboration team. At the information management level, the task status, execution progress, and interaction information of each agent are mostly stored in a decentralized manner, lacking a standardized and unified management carrier. Some solutions rely on complex consensus algorithms to achieve information synchronization, which not only increases the complexity of collaboration but also easily leads to inconsistencies in task status transitions and untraceable execution processes. At the human-machine collaboration level, existing human-machine interactions are mostly based on simple instruction correction and result feedback, lacking a structured intervention and review mechanism based on the entire task lifecycle. Users cannot efficiently control, provide targeted guidance, and adjust the status of task execution through a unified entry point, resulting in insufficient controllability and accuracy of human-machine collaboration.

[0005] Existing technological limitations make it difficult for heterogeneous intelligent agents across ecosystems to collaborate efficiently, lack unified control over task execution, and have low human-machine collaboration efficiency. These limitations fail to meet the needs of flexible networking, standardized collaboration, and refined control of multiple intelligent agents in complex business scenarios. Therefore, there is an urgent need for a loosely coupled, standardized multi-agent collaboration solution that supports structured user intervention to overcome the limitations of the existing architecture and improve the scalability, consistency, and controllability of multi-agent collaboration. Summary of the Invention

[0006] The main objective of this invention is to provide a multi-agent collaboration method based on a unified task management platform.

[0007] Another objective of this invention is to propose a multi-agent collaborative device based on a unified task management platform.

[0008] The third objective of this invention is to provide an electronic device.

[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0010] To achieve the above objectives, a first aspect of the present invention proposes a multi-agent collaboration method based on a unified task management platform, comprising:

[0011] Build and run a unified task management platform to centrally store and manage team attribute information, task attribute information and message information of multi-agent teams, and provide operation interfaces and user interfaces adapted for intelligent agent calls. Configure each intelligent agent to perform long-term loop timed operations. In each loop, query team attribute information, task attribute information and message message information through the operation interface. Based on the query results, perform task planning or task processing, and write back the updated information generated by the processing to the unified task management platform through the operation interface. Responding to and executing the creation command initiated by the user through the user interface, the unified task management platform establishes team attribute information including a member list and task attribute information including task status to initialize the collaboration environment. Each intelligent agent can perform timed operations in parallel, autonomously acquire and process the tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through the operation interface during task processing until the task is completed, thus realizing efficient collaborative work among multiple intelligent agents.

[0012] Optionally, a unified task management platform is built and run to centrally store and manage team attribute information, task attribute information, and message information of multi-agent teams, and to provide operation interfaces for agents to call and user interfaces for users to operate, including: Deploy a data storage model with parallel security mechanisms to centrally store team attribute information, task attribute information, and message information of multi-agent teams; Configure an operation interface for intelligent agents to call, allowing the creation, deletion, query and modification of team attribute information, task attribute information and message information through application programming interface, and ensuring that the interface call does not cause resource read and write conflicts or deadlocks; A user-friendly graphical user interface is constructed, allowing users to intuitively operate the team attribute information, task attribute information, and message message information through a browser, and to send task progress notifications to users according to user settings.

[0013] Optionally, each agent can be configured to perform a long-term, recurring, timed operation. Each cycle queries team attribute information, task attribute information, and message information via an operation interface, including: By utilizing the built-in timing mechanism of the intelligent agent or the timed task mechanism of the external operating system, a long-term loop execution process can be established. The built-in timing mechanism of the intelligent agent includes the loop task function built into the intelligent agent, and the timed task mechanism of the external operating system includes the task scheduler of the Windows operating system and the cron task of the Linux operating system. In each loop, the agent queries the unified task management platform for current team attribute information, task attribute information, and message information through the operation interface. Based on the query results, it analyzes the information to determine task planning, task allocation, or task processing strategies. The agent then returns the generated updated information to the unified task management platform through the operation interface and waits for the information update to be completed.

[0014] Optionally, task planning or task processing can be performed based on the query results, including: Using preset skills or prompts, combined with team attribute information, task attribute information and message information obtained from the unified task management platform, the system generates corresponding analysis results through a large model or intelligent agent, generates corresponding task instructions based on the analysis results, and executes corresponding operations according to the task instructions. When the task instruction executes the initialization operation, it configures its corresponding skills or prompt words based on the obtained team member names, descriptions, roles, and skill information; When a task instruction is to be executed as a subtask or needs to be assigned to a team member, the current task is broken down into subtasks, and the subtask instructions are assigned to other corresponding intelligent agents as needed. When a task instruction is given to perform subtask processing, the task instruction is followed to conduct analysis and to call upon the appropriate tools or skills to complete the assigned subtask.

[0015] Optionally, responding to user-initiated creation commands via the user interface, the unified task management platform is used to establish team attribute information including a member list and task attribute information including task status to initialize the collaboration environment, including: Create a new team on the unified task management platform through the user interface, enter the team name, description and member list information, wherein the member list information includes the name, description, role and skill information of each team member, and wait for the unified task management platform to complete the information update; Create a new task on the unified task management platform through the user interface, enter the task name, objective and description information, and choose to manually assign it to an agent or agent team or not manually assign it. If manually assigned, the task status will be set to assigned; if not manually assigned, the task status will be set to unassigned, and wait for the unified task management platform to complete the information update.

[0016] Optionally, each intelligent agent can execute timed operations in parallel, autonomously acquire and process tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through an operation interface during task processing until the task is completed, thereby realizing efficient collaborative work among multiple intelligent agents, including: Multiple intelligent agents can independently execute their own timed operations in parallel, autonomously identify and retrieve the corresponding tasks recorded in the task attribute information, and execute and process the retrieved tasks according to their own processing logic. Throughout the task processing process, each intelligent agent writes the task execution status back to the unified task management platform in real time according to the operation interface, and updates the task attribute information synchronously to achieve global synchronization of task status. The unified task management platform pushes task progress information and receives user intervention commands through the user interface according to user configuration, and transmits task-related interactive data through the operation interface to complete the human-computer interaction between the intelligent agent and the user. The intelligent agent performs information query, task planning, task processing and information writing in a timed cycle, continuously iterating and advancing task execution until the corresponding task in the task attribute information reaches the completion state, thus realizing efficient collaborative work among multiple intelligent agents.

[0017] To achieve the above objectives, a second aspect of the present invention provides a multi-agent collaborative device based on a unified task management platform, comprising: The platform building module is used to build and run a unified task management platform. Through the unified task management platform, team attribute information, task attribute information and message information of multi-agent teams are centrally stored and managed, and operation interfaces and user interfaces adapted for intelligent agent calls are provided. The timed configuration module is used to configure each intelligent agent to perform long-term loop timed operations. In each loop, the agent queries team attribute information, task attribute information and message message information through the operation interface, performs task planning or task processing based on the query results, and writes the updated information generated by the processing back to the unified task management platform through the operation interface. The instruction response module is used to respond to and execute the creation instruction initiated by the user through the user interface, and to establish team attribute information including a member list and task attribute information including task status in the unified task management platform to initialize the collaboration environment. The collaborative operation module enables each intelligent agent to perform timed operations in parallel, autonomously acquire and process the tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through the operation interface during task processing until the task is completed, thus realizing efficient collaborative operation of multiple intelligent agents.

[0018] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0019] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code stored in the memory, in order to implement a multi-agent cooperation method based on a unified task management platform as described in the first aspect embodiment.

[0020] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements a multi-agent cooperation method based on a unified task management platform as described in the first aspect embodiment.

[0021] The embodiments of the present invention have the following beneficial effects: 1. This invention effectively reduces the coupling degree of multi-agent collaborative systems. By combining a passive unified task management platform with an agent's active timed polling mechanism, it achieves the separation of task logic execution and state storage. The platform only undertakes the responsibilities of state storage and interaction mediation, without the need for complex active scheduling logic. It can seamlessly support heterogeneous agents with different ecosystems and architectures, greatly improving the scalability and flexibility of the collaborative system and expanding the scope of agent collaboration.

[0022] 2. This invention significantly improves the stability and maintainability of multi-agent collaboration. The "passive platform + active polling" architecture reduces the system's dependence on real-time communication and can still maintain stable operation in scenarios such as network fluctuations and dynamic online / offline of agents. Moreover, the core technology modules are built on mature existing technologies, resulting in low development risk and convenient deployment and maintenance. At the same time, the unified platform enables centralized management of tasks, teams, and message information, ensuring the consistency and traceability of task status flow.

[0023] 3. This invention optimizes the human-machine collaboration mode by constructing a structured human-machine intervention mechanism through a platform message list and task state machine. Users can achieve precise guidance, progress control, and result review for task execution through a unified interface, which improves the controllability and accuracy of human-machine collaboration. At the same time, the standardized interface design reduces the integration cost of existing intelligent agents. Combined with a wide range of application scenarios, it has outstanding commercial value and promotion prospects. Attached Figure Description

[0024] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a multi-agent collaboration method based on a unified task management platform, provided in an embodiment of the present invention; Figure 2 This is a structural diagram of a multi-agent collaborative device based on a unified task management platform, provided in an embodiment of the present invention. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] The following description, with reference to the accompanying drawings, illustrates a multi-agent collaboration method and apparatus based on a unified task management platform according to an embodiment of the present invention.

[0028] Example 1 This invention provides a multi-agent collaboration method based on a unified task management platform. Figure 1 This is a flowchart illustrating a multi-agent collaboration method based on a unified task management platform, provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S1: Build and run a unified task management platform. The unified task management platform centrally stores and manages the team attribute information, task attribute information and message message information of the multi-agent team, and provides operation interfaces and user interfaces adapted for intelligent agent calls.

[0029] In this embodiment, the unified task management platform serves as the information hub of the entire collaborative system. The information system it maintains is complete and structured, specifically covering three core categories: team attribute information, task attribute information, and message information. This ensures smooth information synchronization and interaction between multiple agents and users. Team attribute information is key data characterizing the overall features of the agent team and individual members. It includes basic information such as the team name, functional description, and member list. The member list, as a core component of the team attribute information, further details the specific information of each team member, including member name, functional description, role, and skill information. It should be noted that "skill" here is a professional term in the agent field, specifically used to characterize the professional capabilities, execution scope, and core advantages possessed by agent members. It is an important basis for agents to allocate and process tasks.

[0030] At the task management level, task attribute information is mainly used to describe the core characteristics and execution status of a task, specifically including key content such as the task name, execution objective, detailed description, and execution status. The classification and flow logic of task status is crucial to ensuring the orderly progress of tasks. This application's embodiment clearly classifies task status into five types: unassigned, assigned, processing, pending review, and completed, and clarifies the flow rules between each status: During the task initialization phase, if the user has not manually assigned the task, the initial status of the task is set to unassigned by default; if the user has completed the manual assignment operation, the initial status of the task is directly set to assigned; unassigned status... Once a task is claimed by any team member, it automatically switches to the assigned status. Once an assigned task is initiated by the corresponding member, its status changes to "Processing". After a "Processing" task is completed, it will move to a different status based on the task's pre-defined review requirements: if user review is required, it switches to "Pending Review"; otherwise, it switches directly to "Completed". A task in the "Pending Review" status will be updated to "Completed" after user approval; otherwise, the user can revert it to "Unassigned", "Assigned", or "Processing" status as needed, ensuring that tasks can be flexibly adjusted based on actual execution.

[0031] In terms of information interaction, message messages serve as a crucial medium for information exchange and status feedback between intelligent agents and between intelligent agents and users. They are centrally managed in the form of a message list, specifically including three types of messages: first, task messages bound to specific tasks, mainly used to provide feedback on task execution progress, encountered problems, and processing results—information directly related to the task; second, team messages not bound to any task, used for team-level communication between intelligent agents and between intelligent agents and users; and third, other forms of broadcast messages, used to publish team notices, system announcements, and other information that all members need to know. It is also clearly stipulated that both intelligent agents and users have the authority to initiate all types of message messages, ensuring the two-way and timely nature of information interaction.

[0032] To adapt to different interaction objects, the unified task management platform has set up differentiated interaction channels to ensure the convenience and security of various interactive operations. Among them, the operation interface for intelligent agents or fixed programs supports efficient interaction with the unified task management platform via API (Application Programming Interface) calls. Intelligent agents or fixed programs can use this interface to perform full-process operations such as creation, deletion, querying, and modification of various information such as team attributes, task attributes, and message messages. Meanwhile, to avoid anomalies when multiple intelligent agents or programs concurrently call the interface, this operation interface has strict parallel security features. Through a preset concurrency control mechanism, it effectively avoids resource read / write conflicts, deadlocks, and other problems that may occur during interface calls, ensuring the consistency of platform information and the stability of system operation.

[0033] Meanwhile, this application embodiment provides a graphical user interface for users. This interface adopts a simple and intuitive design style, and its functions can be implemented through the backend operation interface or other adapted technologies. Users do not need to master complex technical operations; they can intuitively and conveniently create, delete, query, and modify various information such as team attributes, task attributes, and message messages through terminal devices such as browsers. In addition, the user interface also supports automatically sending task progress notifications to users according to user-preset configuration rules, ensuring that users can keep track of task execution dynamics in real time. It is worth noting that various operations performed by users on the user interface can also be completed by sending relevant instructions to a designated intelligent agent, which will then call the operation interface to complete the operation, further improving the flexibility and convenience of human-computer interaction.

[0034] Step S2: Configure each agent to perform a long-term loop of timed operations. In each loop, query team attribute information, task attribute information and message message information through the operation interface. Based on the query results, perform task planning or task processing, and write back the updated information generated by the processing to the unified task management platform through the operation interface.

[0035] In the embodiments of this application, the scope of intelligent agents involved is broad, covering any single intelligent agent and various intelligent agent teams. It is not limited to intelligent agents of a specific type or specific ecosystem. For example, intelligent agents or intelligent agent teams commonly used in the industry such as OpenClaw, Claude Code, and Manus can all be included in the collaborative system of this application. These agents exhibit high heterogeneity, specifically manifested in the following ways: they can belong to different agent ecosystems, including but not limited to the OpenClaw ecosystem, Claude ecosystem, and Dify workflow ecosystem; they can adopt different team architecture models, such as master-slave architecture (i.e., there is a master agent and subordinate agents, with the master agent responsible for overall planning and subordinate agents responsible for executing specific tasks) and peer-to-peer architecture (i.e., all agents have equal status, autonomously collaborate, and cooperate with each other); they can be deployed in different operating environments, either as a standalone deployment or a distributed cluster deployment, adapting to different application scenario requirements; at the same time, each agent also has different roles, such as simulating a product manager role (responsible for task requirement analysis and goal planning), a full-stack engineer role (responsible for technical implementation and tool calls), and a digital media production role (responsible for content creation and material processing), etc. Agents with different roles leverage their own skill advantages to collaboratively complete complex tasks.

[0036] To ensure continuous and stable operation, the scheduled operations of the agent are not executed temporarily, but rather run in a long-term loop through a specific timing mechanism. This ensures that the agent can continuously interact with the unified task management platform. This timing mechanism is mainly divided into two categories: built-in timing mechanisms and external timing mechanisms, covering the adaptation needs of different agents and different operating environments. Built-in timing mechanisms refer to the timing functions inherent in the agent itself, such as the "Cron Tasks" or "Hearbeat" mechanism of the OpenClaw agent, the loop execution function of the Claude agent, and the "Scheduled Tasks" of the Manus agent. These can achieve scheduled loops without relying on external systems. External timing mechanisms, on the other hand, rely on the scheduling task functions provided by the operating system on which the agent runs, such as the "Task Scheduler" function of the Windows operating system and the "cron" scheduling task mechanism of the Linux operating system. These external systems trigger the agent's scheduled operations, ensuring the stability and reliability of the scheduled loop. Regardless of the timing mechanism used, the core purpose is to achieve the long-term operation of the "query-analysis-update-wait" loop, ensuring continuous interaction between the agent and the platform.

[0037] At the execution level, each timed loop follows a fixed logical sequence to ensure the agent's operation is standardized and efficient. The specific process is as follows: First, at the initial stage of each loop, the agent proactively initiates a query request to the platform through the operation interface provided by the unified task management platform in step S1. This retrieves all core data in the current system, such as team attribute information, task attribute information, and message information, ensuring that the agent can grasp the latest system status and task information, providing data support for subsequent analysis and operations. Subsequently, based on its preset skills or prompts, and combined with various information retrieved from the platform, the agent performs a comprehensive analysis using its integrated large model or built-in algorithms. The core content of the analysis includes task planning, task allocation, specific solutions for task processing, and whether various information in the platform needs to be updated. Ultimately, a clear analysis result is formed to guide the execution of subsequent operations.

[0038] In the specific execution of task planning, task allocation, or task processing, the embodiments of this application clarify the core operational logic of the intelligent agent to ensure orderly task progress and reasonable division of labor: The intelligent agent first autonomously receives and comprehensively analyzes all tasks assigned to it, and performs corresponding operations for different types of tasks based on its own role and skills. If the current task instructs the intelligent agent to perform an initialization operation, the intelligent agent will autonomously configure its own skills or prompts based on detailed information such as team member names, descriptions, roles, and skills obtained from the platform, ensuring that its own functions match the task requirements and team collaboration requirements; if the current task requires sub-task planning or sub-task allocation, the intelligent agent will break down the current overall task into multiple independent and cooperative sub-tasks, and allocate the sub-tasks to other corresponding intelligent agent members as needed, based on the roles and skill advantages of each team member, ensuring that the sub-tasks can be executed efficiently; if the current task is a sub-task processing task, the intelligent agent will strictly follow the specific instructions of the task, autonomously carry out relevant analysis work, and call upon its own corresponding tools and skills to complete the processing of the assigned sub-tasks, ensuring the accuracy and efficiency of sub-task processing.

[0039] After completing the above analysis and processing operations, the agent will autonomously determine whether it needs to supplement, delete, or modify the information in the unified task management platform to avoid the platform information from becoming disconnected from the actual execution status. The specific update logic is as follows: If the task processing involves adjustments to the division of labor among team members (e.g., role changes, responsibilities adjustments) or additions or subtractions of team members (e.g., adding an agent member, removing an existing member), the agent will autonomously update the team attribute information in the platform accordingly. If the task objective is adjusted, the task description is modified, or the task status changes (e.g., from assigned to being processed, from being processed to pending review), the agent will autonomously update the task attribute information in the platform accordingly. If the agent completes the sub-task processing and needs to provide feedback on the processing results to the user or other agents, or if new task information needs to be supplemented during the analysis process, the agent will autonomously organize the relevant details into a task message bound to the task and add it to the platform's message list to achieve real-time feedback and synchronization of information.

[0040] It is important to emphasize that, in this embodiment, all update operations of the intelligent agent on platform information must be completed through the operation interface provided by the platform in step S1 to ensure the standardization and security of the operation. After submitting update information to the platform, the intelligent agent will automatically enter a waiting state until it receives an information update completion receipt from the unified task management platform to avoid subsequent operation anomalies caused by incomplete information updates. After the waiting period, the intelligent agent will automatically enter a preset waiting phase. The core purpose of this waiting phase is to save system computing resources and avoid resource waste caused by frequent looping of the intelligent agent. After the waiting phase ends, the intelligent agent will automatically enter the next timed loop and continue to repeat the above "query-analysis-update-wait" process to ensure continuous interaction with the unified task management platform and provide continuous support for multi-agent collaborative task execution.

[0041] Step S3: Respond to and execute the creation command initiated by the user through the user interface, and establish team attribute information containing a member list and task attribute information containing task status in the unified task management platform to initialize the collaboration environment.

[0042] In this embodiment, the user, as the initiator and controller of the entire collaborative system, can initiate team and task creation commands simply by relying on the graphical user interface of the unified task management platform built in step S1, without needing complex technical operations. The operation is convenient and intuitive. After receiving the creation command initiated by the user, the unified task management platform will immediately activate the response mechanism to receive, verify, enter, and store various types of information input by the user. This ensures that the user's operations can be quickly transformed into structured data that the platform can recognize and manage, while ensuring the integrity and accuracy of the entered information, providing a reliable data foundation for subsequent intelligent agent queries and interactions.

[0043] In the team creation phase, this application embodiment clarifies the user's operation process and the platform's response logic to ensure the standardization and completeness of team creation. Specifically, through the platform's graphical user interface, users sequentially input the core information of the agent team according to the interface guidance, including the team name (used to uniquely identify the agent team, facilitating user and agent identification), team description (used to describe the team's core functions, collaborative goals, application scenarios, etc., helping users and agents clarify the team's positioning), and member list information. Among them, the member list is the core content of team creation. Users must clearly specify the specific information of each agent member in the member list, which is indispensable. Specifically, it includes the member's name (used to identify an individual agent), member description (used to describe the agent's functional characteristics and applicable scenarios), member role (used to clarify the agent's division of labor and responsibilities in the team, such as leader, executor, assistant, etc.), and member skill (Skill, as a professional term in the field of agents, used to characterize the agent's professional capabilities and execution scope, and is an important basis for subsequent task allocation). After receiving all the aforementioned information, the unified task management platform verifies its legality. Once confirmed to be complete and error-free, it stores the information in the platform's database and updates team attribute information in real time, ensuring that agents can access the latest team information through the operation interface. Furthermore, this embodiment fully considers the needs of practical application scenarios, supporting users in creating multiple different agent teams simultaneously. Each team is independent and does not interfere with others. Users can build agent teams with different architectures and functions according to different collaboration needs, improving the system's flexibility and applicability.

[0044] In the task creation phase, this embodiment also follows the design principles of convenience and flexibility, clearly defining the operation process and state initialization rules for task creation. Users input various core information about the task through the graphical user interface of the unified task management platform, including the task name (used to uniquely identify the task), task objective (used to clarify the task's execution direction and expected results, providing task guidance for the agent), and task description (used to detail the task's execution requirements, operation steps, constraints, etc., ensuring the agent can accurately understand the task requirements). After completing the core information input, users can choose the task allocation method according to actual collaboration needs: they can choose to manually allocate the task to a designated single agent or a team of agents, clearly defining the task's execution entity; or they can choose not to manually allocate the task, allowing subsequent agents to automatically receive the task through scheduled operations.

[0045] For different allocation methods, this application embodiment clarifies the rules for setting the initial state of tasks to ensure the standardization and traceability of task states: If the user chooses to manually allocate tasks, the initial state of the task will be automatically marked as "allocated" by the unified task management platform, indicating that the task has a determined execution subject, facilitating querying and execution by the intelligent agent; if the user chooses not to manually allocate tasks, the initial state of the task will be marked as "unallocated" by default, waiting for the intelligent agent to query and claim it through scheduled operations. After the user completes all information entry and allocation method selection, the unified task management platform will verify the entered task information, store it after confirming its accuracy, and update the task attribute information in the platform in real time to ensure that the task information can be queried by the intelligent agent in a timely manner.

[0046] In addition, similar to team creation, this application embodiment also supports users to create multiple tasks at the same time. Each task can exist independently and be executed in parallel. Users can initiate the creation of multiple tasks in batches according to actual needs, further improving collaboration efficiency and adapting to the multi-task collaboration needs in complex scenarios.

[0047] Step S4 enables each intelligent agent to perform timed operations in parallel, autonomously retrieve and process the tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through the operation interface during task processing until the task is completed, thereby realizing efficient collaborative work among multiple intelligent agents.

[0048] In this embodiment, each intelligent agent autonomously and continuously executes timed operations and automatically interacts with the unified task management platform, thereby fully realizing a collaborative task execution mechanism among multiple intelligent agents. This embodiment utilizes a decentralized autonomous polling mode, allowing multiple intelligent agents to run pre-configured timed loop processes in parallel and independently. Each intelligent agent actively accesses the unified task management platform according to its own cycle to obtain the latest team attribute information, task attribute information, and message information. Based on its own role, skills, and task requirements, it completes local task planning, task breakdown, task allocation, and specific processing operations, while simultaneously transmitting the execution process and results back to the platform in real time for information synchronization. At the overall operational level, although multiple intelligent agents operate independently, under the information coordination of the unified task management platform, they can form an orderly and interconnected collaborative relationship, jointly promoting the successful completion of one or more task objectives.

[0049] As the information hub of the entire system, the unified task management platform receives two types of information update requests in real time: those initiated by agents through operation interfaces and those submitted by users through a graphical user interface. Upon receiving these requests, the platform first verifies the legality and validity of the updated content, filtering out invalid or conflicting operations. Updates that comply with the rules are then formally applied, and the processing results are promptly fed back to the requesting agent or user in the form of an update receipt. Simultaneously, the platform also pushes task progress-related notifications to users in real time according to pre-set notification policies. Typical notification scenarios include adding task messages, task status changing to "pending review," and critical changes in task status, enabling users to stay informed about collaboration progress.

[0050] Throughout the task execution process, users can view the task status, processing progress, and message information in real time through the unified task management platform's user interface, and flexibly and controllably intervene in the task execution process as needed. When the task is executing normally and requires no adjustment, users can remain inactive, and the system will continue to operate autonomously. When users believe it is necessary to guide the task direction, supplement task requirements, or correct execution deviations, they can publish task messages bound to the corresponding task through the user interface, providing clear execution guidance to the agent. For tasks that have entered the pending review stage, users can combine task messages and execution results for review. If the review is passed, the task status will be updated to "completed"; if the review is failed, the task status can be rolled back to "unassigned," "assigned," or "processing" status, allowing the agent to re-execute. In addition, users can also publish team messages or broadcast messages to achieve team-level information synchronization, further enhancing the controllability of human-machine collaboration.

[0051] Platform information, updated by user intervention or autonomously by the agents, will be re-acquired and parsed by each agent in the next timed cycle. The agents will then adjust their task strategies and execution logic based on the updated task status, message content, and team configuration, continuously iterating and advancing the task process until the task status is finally changed to "completed." Through this mechanism, this application achieves efficient collaborative work between heterogeneous agents, while also ensuring user-monitored and flexible intervention throughout the task process, achieving the goal of stable, reliable, and efficient human-machine collaboration.

[0052] Example 2 This invention provides a multi-agent collaborative device based on a unified task management platform. Figure 2 This is a flowchart illustrating a multi-agent collaborative device based on a unified task management platform, provided as an embodiment of the present invention. Figure 2 As shown, the device includes: Platform building module 100 is used to build and run a unified task management platform. The unified task management platform centrally stores and manages team attribute information, task attribute information and message information of multi-agent teams, and provides operation interfaces and user interfaces that are adapted to intelligent agent calls. The timing configuration module 200 is used to configure each intelligent agent to perform long-term cyclical timing operations. It queries team attribute information, task attribute information and message message information through the operation interface, performs task planning or task processing based on the query results, and writes back the updated information generated by the processing to the unified task management platform through the operation interface. The instruction response module 300 is used to respond to the creation instruction initiated by the user through the user interface, and to establish team attribute information including a member list and task attribute information including task status in the unified task management platform to initialize the collaboration environment. The collaborative operation module 400 is used to control each intelligent agent to perform timed operations in parallel, autonomously pick up and process the tasks recorded in the task attribute information, and realize state synchronization and human-computer interaction through the operation interface during the task processing until the task is completed, so as to realize efficient collaborative operation of multiple intelligent agents.

[0053] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0054] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0055] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0057] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0058] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A multi-agent collaboration method based on a unified task management platform, characterized in that, include: Build and run a unified task management platform to centrally store and manage team attribute information, task attribute information and message information of multi-agent teams, and provide operation interfaces and user interfaces adapted for intelligent agent calls. Configure each intelligent agent to perform long-term loop timed operations. In each loop, query team attribute information, task attribute information and message message information through the operation interface. Based on the query results, perform task planning or task processing, and write back the updated information generated by the processing to the unified task management platform through the operation interface. Responding to and executing the creation command initiated by the user through the user interface, the unified task management platform establishes team attribute information including a member list and task attribute information including task status to initialize the collaboration environment. Each intelligent agent can perform timed operations in parallel, autonomously acquire and process the tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through the operation interface during task processing until the task is completed, thus realizing efficient collaborative work among multiple intelligent agents.

2. The method according to claim 1, characterized in that, A unified task management platform is constructed and operated. This platform centrally stores and manages team attribute information, task attribute information, and message information for multi-agent teams. It also provides operation interfaces adapted for agent invocation and user interfaces for user operation, including: Deploy a data storage model with parallel security mechanisms to centrally store team attribute information, task attribute information, and message information of multi-agent teams; Configure an operation interface for intelligent agents to call, allowing the creation, deletion, query and modification of team attribute information, task attribute information and message information through application programming interface, and ensuring that the interface call does not cause resource read and write conflicts or deadlocks; A user-friendly graphical user interface is constructed, allowing users to intuitively operate the team attribute information, task attribute information, and message message information through a browser, and to send task progress notifications to users according to user settings.

3. The method according to claim 2, characterized in that, Configure each agent to perform a long-term, looping, timed operation. Each loop queries team attribute information, task attribute information, and message information via the operation interface, including: By utilizing the built-in timing mechanism of the intelligent agent or the timed task mechanism of the external operating system, a long-term loop execution process can be established. The built-in timing mechanism of the intelligent agent includes the loop task function built into the intelligent agent, and the timed task mechanism of the external operating system includes the task scheduler of the Windows operating system and the cron task of the Linux operating system. In each loop, the agent queries the unified task management platform for current team attribute information, task attribute information, and message information through the operation interface. Based on the query results, it analyzes the information to determine task planning, task allocation, or task processing strategies. The agent then returns the generated updated information to the unified task management platform through the operation interface and waits for the information update to be completed.

4. The method according to claim 3, characterized in that, Task planning or processing based on query results includes: Using preset skills or prompts, combined with team attribute information, task attribute information and message information obtained from the unified task management platform, the system generates corresponding analysis results through a large model or intelligent agent, generates corresponding task instructions based on the analysis results, and executes corresponding operations according to the task instructions. When the task instruction executes the initialization operation, it configures its corresponding skills or prompt words based on the obtained team member names, descriptions, roles, and skill information; When a task instruction is to be executed as a subtask or needs to be assigned to a team member, the current task is broken down into subtasks, and the subtask instructions are assigned to other corresponding intelligent agents as needed. When a task instruction is given to perform subtask processing, the task instruction is followed to conduct analysis and to call upon the appropriate tools or skills to complete the assigned subtask.

5. The method according to claim 4, characterized in that, Responding to user-initiated creation commands via the user interface, the unified task management platform establishes team attribute information including a member list and task attribute information including task status to initialize the collaboration environment, including: Create a new team on the unified task management platform through the user interface, enter the team name, description and member list information, wherein the member list information includes the name, description, role and skill information of each team member, and wait for the unified task management platform to complete the information update; Create a new task on the unified task management platform through the user interface, enter the task name, objective and description information, and choose to manually assign it to an agent or agent team or not manually assign it. If manually assigned, the task status will be set to assigned; if not manually assigned, the task status will be set to unassigned, and wait for the unified task management platform to complete the information update.

6. The method according to claim 5, characterized in that, This enables each intelligent agent to execute timed operations in parallel, autonomously acquire and process tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through an operation interface during task processing until the task is completed, thus realizing efficient collaborative work among multiple intelligent agents, including: Multiple intelligent agents can independently execute their own timed operations in parallel, autonomously identify and retrieve the corresponding tasks recorded in the task attribute information, and execute and process the retrieved tasks according to their own processing logic. Throughout the task processing process, each intelligent agent writes the task execution status back to the unified task management platform in real time according to the operation interface, and updates the task attribute information synchronously to achieve global synchronization of task status. The unified task management platform pushes task progress information and receives user intervention commands through the user interface according to user configuration, and transmits task-related interactive data through the operation interface to complete the human-computer interaction between the intelligent agent and the user. The intelligent agent performs information query, task planning, task processing and information writing in a timed cycle, continuously iterating and advancing task execution until the corresponding task in the task attribute information reaches the completion state, thus realizing efficient collaborative work among multiple intelligent agents.

7. A multi-agent collaborative device based on a unified task management platform, characterized in that, include: The platform building module is used to build and run a unified task management platform. Through the unified task management platform, team attribute information, task attribute information and message information of multi-agent teams are centrally stored and managed, and operation interfaces and user interfaces adapted for intelligent agent calls are provided. The timed configuration module is used to configure each intelligent agent to perform long-term loop timed operations. In each loop, the agent queries team attribute information, task attribute information and message message information through the operation interface, performs task planning or task processing based on the query results, and writes the updated information generated by the processing back to the unified task management platform through the operation interface. The instruction response module is used to respond to and execute the creation instruction initiated by the user through the user interface, and to establish team attribute information including a member list and task attribute information including task status in the unified task management platform to initialize the collaboration environment. The collaborative operation module enables each intelligent agent to perform timed operations in parallel, autonomously acquire and process the tasks recorded in the task attribute information, and achieve state synchronization and human-computer interaction through the operation interface during task processing until the task is completed, thus realizing efficient collaborative operation of multiple intelligent agents.

8. The apparatus according to claim 7, characterized in that, The platform building module is also used for: Deploy a data storage model with parallel security mechanisms to centrally store team attribute information, task attribute information, and message information of multi-agent teams; Configure an operation interface for intelligent agents to call, allowing the creation, deletion, query and modification of team attribute information, task attribute information and message information through application programming interface, and ensuring that the interface call does not cause resource read and write conflicts or deadlocks; A user-friendly graphical user interface is constructed, allowing users to intuitively operate the team attribute information, task attribute information, and message message information through a browser, and to send task progress notifications to users according to user settings.

9. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images