Multi-model agent collaboration method and system

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a multi-model agent collaboration method and system, and the method comprises the following steps: obtaining a task execution plan and an actual state, recognizing a deviation node, generating a synchronous deviation list, analyzing a task trend, labeling a progress label, recognizing a tool influence section, screening uninfluenced nodes, and judging the distribution efficiency. And extracting abnormal fluctuation, analyzing resource and task cycle difference, and generating a monitoring structure index. According to the method, task deviation identification is realized by extracting the task plan number and the time interval and comparing the real-time state of the intelligent agent, the plan execution monitoring precision is improved, the perception of progress change is enhanced, and the task distribution efficiency is evaluated by performing cross comparison on fluctuation tasks and tool calling data, identifying interference sections, mapping running logs and dispatching data. And in combination with resource release and task fluctuation differences, progress monitoring indexes are extracted, and the cooperation efficiency and the system regulation and control capability are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly to a multi-model agent collaboration method and system. Background Art

[0002] The field of artificial intelligence technology includes multiple branches such as agent technology, machine learning, deep learning, natural language processing, and computer vision. The core content of this technology field is to design automated systems that can autonomously learn, judge, and make decisions by simulating human intelligence. An agent refers to an automated system that can perceive the environment, make decisions, execute tasks, and provide feedback. With the improvement of computing power and the enrichment of data resources, artificial intelligence technology has been widely applied, covering multiple application scenarios such as automated control, image recognition, speech processing, and natural language understanding. The development of this field not only promotes the progress of science and technology but also brings profound changes in various industries, including multiple fields such as education and transportation.

[0003] Among them, the multi-model agent collaboration method refers to achieving more efficient task completion through the collaboration of multiple agents. The subject proposes an optimized collaboration method for problems such as task allocation, resource coordination, and information sharing existing in the collaborative work of multiple agents. This method enables each agent to flexibly collaborate and cooperate according to the requirements of different tasks by designing a collaboration mechanism between multi-model agents. In this method, agents share knowledge, data, and resources to achieve information transfer and joint task completion. Each agent independently or collaboratively executes tasks according to its own characteristics and capabilities, thereby improving the overall system efficiency and intelligence level. This method also optimizes the task execution process and efficiency by adjusting the collaboration strategy between agents.

[0004] Most of the existing technologies rely on static execution processes during task status recognition, lacking the ability to promptly capture actual operation deviations and being difficult to detect progress deviations existing in task execution. For example, in the case of parallel execution of multiple tasks, a delay in a certain task node causes the subsequent task chain to lag as a whole. However, due to the lack of a real-time monitoring mechanism, the delay risk cannot be warned. Task trend judgment is mostly based on fixed templates rather than the evolution law of actual operation data, making it difficult to accurately reflect the dynamic progress of tasks, thus forming resistance to resource allocation and intervention strategies. The analysis of the impact of tool calls is absent, resulting in the inability to evaluate the conflict between the frequent call of key tools and task node execution, causing resource congestion in high-concurrency scenarios. There is no comparison mechanism between operation logs and task assignment data, resulting in the difficulty of discovering inefficient task distributions and causing resource waste. In terms of progress monitoring, existing technologies mainly focus on coarse-grained statistics, ignoring the resource matching and task completion efficiency analysis in fine-grained sections and being difficult to meet the regulation requirements under dynamic changes. The problems are particularly prominent in the collaboration of multi-model agents, directly affecting the overall coordination and task completion quality. Summary of the Invention

[0005] The object of the present invention is to solve the deficiencies existing in the prior art, and to propose a multi-model agent collaboration method and system.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A multi-model agent collaboration method includes the following steps: S1: Obtain the knowledge base document, tool interface configuration information, and external API call records in the task scenario, extract the execution plan number and time interval of the agent task node, compare the real-time execution status of the agent with the planned node identifier, identify the task node number deviating from the plan, and generate a task node synchronization offset list; S2: Based on the task node synchronization offset list, screen the Agent-type intelligent assistant, extract the start and end times of task decomposition in combination with the task decomposition table, perform difference calculation and trend classification with the planned time point, and generate a task progress trend label group; S3: Invoke the task progress trend label group, extract the fluctuating task partition number, identify the regional tool call frequency data, compare the tool call period with the fluctuating task time period, record the number of overlapping time periods, and generate a list of affected sections of the task node by tool calls; S4: Based on the list of affected sections of the task node by tool calls, screen the partition node tasks not affected, extract the agent segmented task records and operation log data, map the daily task allocation volume to the agent operation period, and determine whether there is an inefficient distribution of tasks to obtain the agent node task completion fluctuation group.

[0007] As a further solution of the present invention, the task node synchronization offset list includes a task offset number, an execution status label, a time offset, and a task node classification. The task progress trend label group includes a progress offset level, a trend change type, a planned comparison result, and a task association number. The list of affected sections of the task node by tool calls includes a tool call type, an affected time section, the number of overlapping periods, and an affected task number. The agent node task completion fluctuation group includes a task distribution uneven number, a running duration record, a task completion deviation, and an agent operation matching degree.

[0008] As a further solution of the present invention, the specific steps for obtaining the task node synchronization offset list are as follows: S111: Obtain the knowledge base document, tool interface configuration information, and external API call records in the task scenario, extract the planned time and number of the agent task node, match the update time and coordinates of the knowledge base document, compare the matching document time range with the node planned time, and generate a partition node document record period; S112: Based on the document recording period of the partition node, perform a coincidence judgment with the planned time interval of the task node, extract the ratio of the coincidence time to the total duration, screen the node numbers with a ratio lower than the benchmark value, and update the status annotation of the node document to obtain the progress coverage deviation rate of the partition node. S113: According to the progress coverage deviation rate of the partition node, determine the deviation status of the task node number, identify the task node numbers with a deviation rate higher than the node synchronization threshold, integrate the node number, document coverage information, and deviation rate value, and generate a task node synchronization offset list.

[0009] As a further solution of the present invention, the obtaining steps of the task progress trend tag group are specifically as follows: S211: Based on the task node synchronization offset list, identify the Agent-type intelligent assistant with reasoning-planning-execution capabilities and the task decomposition table, extract the start and completion times of the real-time task decomposition, calculate the time difference between the start and end times of the task section, and compare it with the planned task node time difference to obtain the task time offset value. S212: Invoke the task time offset value, combine the section distribution, offset trend, and adjustment frequency, collect the task node offset data, identify and calculate the trend offset degree according to the section number, and judge the trend direction based on the adjustment frequency to obtain the task progress trend tag group.

[0010] As a further solution of the present invention, the obtaining steps of the list of sections of the task node affected by tool calls are specifically as follows: S311: Invoke the task progress trend tag group, screen the partition node numbers with fluctuating offsets, extract the task time period according to the node association table, process the section task period in the time dimension, identify the task time period index table, and obtain the set of fluctuating task activity periods. S312: According to the set of fluctuating task activity periods, collect the tool call frequencies in the same period section, identify the influence table, judge the interference based on the interference threshold, match it with the construction period of the fluctuating task, analyze the tool intervention intensity, judge the abnormal association of progress monitoring, and generate a list of sections of the task node affected by tool calls.

[0011] As a further solution of the present invention, the obtaining steps of the task completion fluctuation group of the intelligent agent node are specifically as follows: S411: Based on the list of sections of the task node affected by tool calls, screen the unmarked partition nodes, extract and record the job task list of the nodes, including the task number, planned process, start time, and end time, to obtain the set of task nodes not affected by tool call interference. S412: Invoke the set of node tasks not interfered by tool calls, match the daily segmented task records of the agent with the operation log data, extract the planned task volume according to the task number, count the running duration and the real-time running period, and generate an agent task execution matching data set; S413: According to the agent task execution matching data set, judge the matching degree between the task execution efficiency and the operation distribution, identify the efficiency fluctuation nodes and evaluate the task deviation degree, analyze the task efficiency fluctuation degree of the task nodes, and mark the tasks with a fluctuation degree exceeding the set benchmark value as abnormal task nodes to obtain an agent node task completion fluctuation group.

[0012] As a further solution of the present invention, the method further includes step S5: S5: Invoke the agent node task completion fluctuation group, extract the abnormal fluctuation task group, calculate the ratio of the resource input volume to the task volume in the knowledge base, record the difference distribution between the resource delivery cycle and the task execution cycle, and generate a task progress monitoring structure index; The task progress monitoring structure index includes a resource allocation ratio, a task intensity level, an execution cycle deviation, and a task efficiency index.

[0013] As a further solution of the present invention, the obtaining steps of the task progress monitoring structure index are specifically as follows: S511: Invoke the node numbers and the corresponding operation time intervals in the agent node task completion fluctuation group, identify the difference in task volume between adjacent task days, screen the nodes exceeding the threshold, record the time interval and the change amplitude of the task volume, and obtain a progress fluctuation abnormal identification set; S512: Based on the knowledge base resources corresponding to the nodes in the progress fluctuation abnormal identification set, identify the node resource input ratio sequence, extract the abnormal distribution interval and compare the critical coefficient, record the ratio deviation direction and the node number, and form a task resource matching deviation index group; S513: According to the task resource matching deviation index group, extract the knowledge base resource delivery and task execution time periods, analyze the difference between the resource delivery cycle and the operation cycle, identify the task time synchronization deviation distribution table, and generate a task progress monitoring structure index according to the progress benchmark sorting annotation difference.

[0014] The multi-model agent collaboration system is used to execute the above multi-model agent collaboration method, and the system includes: The task status extraction module obtains the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extracts the agent task node numbers and time intervals, compares the real-time execution status with the task node identifiers, marks the node tasks with inconsistent times, and generates a node status deviation label group; Based on the node status offset tag group, the node trend classification module locates the Agent-type intelligent assistant with reasoning - planning - execution capabilities, extracts the task decomposition table and planned time, identifies the difference between the real-time task time and the planned time, classifies and labels the difference types, and generates a task node progress trend identification group; Based on the task node progress trend identification group, the tool call recognition module filters out the fluctuating and lagging task nodes, locates the corresponding partitions, extracts the time periods with frequent tool call frequency interference, determines the coincidence with the task time period, filters out the frequently interfered sections, and generates a task progress tool call interference mapping set; Based on the task progress tool call interference mapping set, the task efficiency diagnosis module eliminates the tasks in the interfered sections, extracts the intelligent agent task records and operation logs, matches the task assignment and operation time periods, calculates the ratio of the operation volume to the task, identifies the intensive but inefficient node tasks, and obtains a node task execution deviation set; Based on the node task execution deviation set, the resource allocation analysis module locates the intelligent agent resource input records, extracts the ratio of the task volume to the resource configuration, compares the difference between the execution cycle and the input cycle, maps the resource usage and task progress status, and forms a task progress monitoring structure index.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by dynamically extracting the execution plan number and time interval of the task node and comparing with the real-time status of the intelligent agent, task deviation recognition is realized, and the dynamic monitoring accuracy of plan execution is improved. In the task trend analysis stage, an intelligent assistant with reasoning and planning capabilities is integrated. Based on the task decomposition time and the evaluation of the difference changes of the planned nodes, trend labels are effectively generated, enhancing the insight into the task progress changes. Further, through the cross-analysis of the fluctuating task time period and the tool call data, the time coincidence degree is recorded, so that the sections of the task affected by external factors can be accurately identified. The undisturbed task partitions map the assignment volume and operation time period through the intelligent agent operation log, realizing the quantitative evaluation of the intelligent agent resource and task distribution efficiency. Combining the task completion fluctuation and the difference in the resource delivery cycle, the matching degree between the task completion efficiency and the resource configuration is refined, forming a progress monitoring index system with timeliness and accuracy. The entire processing logic reflects a full-process closed loop from task execution status recognition, trend evolution analysis, tool call impact analysis, resource distribution rationality evaluation to index structure generation, ensuring the synchronous optimization of task execution and resource allocation in multiple dimensions, thereby effectively improving the overall operation efficiency and response ability, and enhancing the flexibility and adaptability of task regulation in complex multi-intelligent agent collaboration scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic diagram of the work flow of the present invention; Figure 2It is the flowchart for obtaining the task node synchronization offset list in the present invention; Figure 3 It is the flowchart for obtaining the task progress trend tag group in the present invention; Figure 4 It is the flowchart for obtaining the list of sections of task nodes affected by tool calls in the present invention; Figure 5 It is the flowchart for obtaining the task completion fluctuation group of the agent node in the present invention; Figure 6 It is the flowchart for obtaining the task progress monitoring structure index in the present invention. Detailed implementation manners

[0017] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0018] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0019] Embodiment 1: Please refer to Figure 1 , the present invention provides a technical solution: a multi-model agent collaboration method, including the following steps: S1: Obtain the knowledge base documents, tool interface configuration information and external API call records in the task scenario, extract the execution plan numbers and time intervals of the agent task nodes, compare the real-time execution status of the agent with the planned node identifiers, identify the task node numbers deviating from the plan, and generate a task node synchronization offset list; S2: Based on the task node synchronization offset list, screen the Agent-type intelligent assistants with reasoning-planning-execution capabilities, extract the start and end times of task decomposition in combination with the task decomposition table, calculate the differences from the planned time points and classify the trends, and generate a task progress trend tag group; S3: Call the task progress trend tag group, extract the numbers of the fluctuating task partitions, identify the tool call frequency data in the regions, compare the tool call time periods with the fluctuating task time periods, record the number of overlapping time periods, and generate a list of sections of task nodes affected by tool calls; S4: Based on the list of sections of task nodes affected by tool calls, filter the partition node tasks that are not affected, extract the intelligent agent segmented task records and operation log data, map the daily task distribution volume to the operation period of the intelligent agent, determine whether there is an inefficient distribution of tasks, and obtain the intelligent agent node task completion fluctuation group; S5: Invoke the intelligent agent node task completion fluctuation group, extract the abnormal fluctuation task group, calculate the ratio of the resource input amount to the task amount in the knowledge base, record the difference distribution between the resource delivery cycle and the task execution cycle, and generate the task progress monitoring structure indicators.

[0020] The task node synchronization offset list includes the task offset number, execution status label, time offset, task node classification. The task progress trend label group includes the progress offset level, trend change type, plan comparison result, task association number. The list of sections of task nodes affected by tool calls includes the tool call type, affected time section, number of overlapping periods, affected task number. The intelligent agent node task completion fluctuation group includes the task distribution unevenness number, operation duration record, task completion deviation, intelligent agent operation matching degree. The task progress monitoring structure indicators include the resource allocation ratio, task intensity level, execution cycle deviation, task efficiency indicator.

[0021] Please refer to Figure 2 , and the specific steps for obtaining the task node synchronization offset list are as follows: S111: Obtain the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extract the planned time and number of intelligent agent task nodes, match the update time and coordinates of the knowledge base documents, compare the matched document time range with the node planned time, and generate the document record period of the partition node; Obtain the knowledge base document data set in the task scenario from the background of the agent working platform, such as obtaining project requirement documents, interface specification manuals, original task reports, etc. Read the tool interface configuration information list from the Agent system configuration, including descriptions of various tool APIs, parameter definitions, call methods, expected response formats, etc. Query the details of external API call records within a specific time range from the log storage, such as capturing requests, response logs, call timestamps, return status codes, etc. when the Agent executes tasks and interacts with third-party services. Extract the planned start and end times of the agent task nodes from the task decomposition schedule, as well as the corresponding task node numbers. For example, the planned execution time of node Node_Alpha is from 10-27-10:00 to 10-27-10:15, and the node number is N001. The planned execution time of node Node Beta is from 10-27-10:20 to 10-27-10:35, and the node number is N002. Match the last update timestamp of the knowledge base documents. For example, the update time of document DocA (project requirement document) is 10-27-10:05, the update time of document DocB (interface specification manual) is 10-27-10:22, and the update time of document Doc_C (another version of the interface specification manual) is 10-27-10:30. Match the associated coordinates of the knowledge base documents in the task scenario. The coordinates can be document version numbers, specific section identifiers, or association tags with the task content. For example, the associated coordinate of Doc_A is "Requirementv1.1, Section3", the associated coordinate of Doc_B is "APIManualv2.0, EndpointA", and the associated coordinate of Doc_C is "APIManualv2.1, Endpoint". Compare the matched document time ranges. Here, the effective influence time of the document is defined as its update time point. If the update time point is earlier than the planned start time of the node, it is not used as a recording period. If the update time point is within the planned interval, record this update time point. For nodes with multiple matched documents or multiple update time points, record all update time points within the planned interval. For example, for node N001 with a planned execution time of [10:00, 10:15] and an associated document Doc_A with an update time of 10:05, the recording period is 10:05. For node N002 with a planned execution time of [10:20, 10:35] and associated documents Doc_B with an update time of 10:22 and Doc_C with an update time of 10:30, the recording periods are 10:22 and 10:30. Generate a partitioned node document recording period, which contains each task node number and all relevant knowledge base document update time points that fall within its planned execution time range. For example, the set is {N001: [10:05], N002: [10:22, 10:30]}.

[0022] S112: Based on the time period recorded in the partition node document, perform a coincidence judgment with the planned time interval of the task node, extract the ratio of the coincidence time to the total duration, screen the node numbers with a ratio lower than the benchmark value, and update the status annotation of the node document to obtain the deviation rate of the partition node progress coverage; Based on the set of document recording time periods for partitioned nodes, for example, the set is {N001: [10:05], N002: [10:22, 10:30]}, perform a coincidence judgment with the planned time intervals of each task node. For node N001, its planned time interval is [10:00, 10:15], and the recording time period is [10:05]. Determine whether the recording time period 10:05 falls within the planned interval [10:00, 10:15], indicating that there is a coincidence between document update and node execution time. Extract the ratio of the coincidence time to the total duration. Here, set the impact of the document update time point as an impact window with a fixed duration, for example, set it to 5 minutes, that is, within 5 minutes after the document update, it may have an impact on the task. For the recording time period 10:05, its impact window is [10:05, 10:10]. The total planned duration of node N001 is 15 minutes. Calculate the coincidence duration between the impact window [10:00, 10:10] and the planned interval [10:00, 10:15]. The coincidence interval is [10:05, 10:10], and the coincidence duration is 10:10 - 10:05 = 5 minutes. Calculate the ratio of the coincidence time to the total duration, ratio = coincidence duration / planned total duration = 5 / 15 ≈ 0.333. For node N002, the planned interval is [10:20, 10:35], the total duration is 15 minutes, and the recording time periods are [10:22, 10:30]. Consider two recording time periods, each with an impact window of 5 minutes: [10:22, 10:27] and [10:30, 10:35]. The coincidence durations with the planned interval [10:20, 10:35] are 5 minutes and 5 minutes respectively. The total coincidence duration is 5 + 5 = 10 minutes (assuming that the two impact windows do not overlap or the overlapping part is only calculated once for the actual coincidence duration. Here, for simplicity, calculate the respective coincidence durations and then sum them). Calculate the ratio of the coincidence time to the total duration, ratio = 10 / 15 ≈ 0.667. Screen out the node numbers with ratios lower than the benchmark value. The benchmark value is used to define the situation where the document update has a relatively small impact on the task. Its setting refers to the dependence of the task on the knowledge base document and the document update frequency. For example, for a data processing task that depends on a data specification document, its document update directly affects the processing logic and has a high dependence. For a task that only refers to interface parameters, the dependence is relatively low. The benchmark value can be obtained through raw data analysis. For example, statistically analyze the correlation between document updates and task failure rates and delay rates over a period of time to determine an empirical value, or manually set it according to the criticality level of the task. For example, set the benchmark value to 0.5, indicating that task nodes where the impact window of document update covers less than half of the planned duration need to be further marked. Compare the ratio 0.333 of node N001 with the benchmark value 0.5. Since 0.333 < 0.5, node N001 is screened out. Compare the ratio 0.667 of node N002 with the benchmark value 0.5. Since 0.667 > 0.5. Node N002 was not screened. Update the status annotation of the node document. Mark the screened node N001 as "low document coverage rate", and mark the unscreened node N002 as "document coverage rate meets the standard". Obtain the progress coverage deviation rate of the partition nodes. For the screened nodes, its deviation rate is defined as 1 - ratio, that is, the coverage deviation rate of node N001 is 1 - 0.333 ≈ 0.667. For the unscreened nodes, its deviation rate can be regarded as 0 or not high.

[0023] S113: According to the progress coverage deviation rate of the partition nodes, determine the deviation status of the task node numbers, identify the task node numbers with deviation rates higher than the node synchronization threshold, integrate the node numbers, document coverage information and deviation rate values, and generate a task node synchronization offset list; According to the progress coverage deviation rate of the partition nodes, for example, the coverage deviation rate of node N001 is about 0.667, and the coverage deviation rate of node N002 is regarded as not high (for example, 0). Determine the deviation status of the task node numbers. The judgment basis is whether the coverage deviation rate is higher than the set node synchronization threshold. The node synchronization threshold is used to define the minimum requirement for the synchronization of document updates and task execution. Its setting refers to the real-time requirements of the task and the sensitivity to the timeliness of the document. For example, for tasks with high real-time requirements where document updates must be reflected in the execution in a timely manner (such as financial trading strategy execution), the threshold can be set lower (such as 0.1), and any deviation will be marked. On the contrary, for tasks that can tolerate a certain delay (such as offline data analysis), a higher threshold can be set (such as 0.5). The setting of the threshold can also be determined by analyzing the correlation between synchronization deviation and task anomalies in the original data. For example, set the node synchronization threshold to 0.4, indicating that task nodes with a coverage deviation rate higher than 0.4 are determined to be synchronously deviated. Identify the task node numbers with deviation rates higher than the node synchronization threshold. Compare the deviation rate 0.667 of node N001 with the threshold 0.4. Since 0.667 > 0.4, node N001 is identified as a synchronously deviated node. Compare the deviation rate 0 of node N002 with the threshold 0.4. Since 0 < 0.4, node N002 is not identified as a synchronously deviated node. Integrate the identified synchronously deviated node numbers, relevant document coverage information, and the calculated deviation rate values, organize the information into structured data, such as in the form of a list or a table, and generate a task node synchronization offset list; Table 1: Task node synchronization offset list; 。

[0024] As shown in Table 1, the task node synchronization offset list lists the task nodes determined to have synchronization problems due to the low coincidence degree between the knowledge base document update and the task planned time, including node numbers, document coverage rates, coverage deviation rates, and their judgment statuses.

[0025] Please refer toFigure 3 , the steps for obtaining the task progress trend label group are specifically as follows: S211: Based on the task node synchronization offset list, identify the Agent-type intelligent assistant with reasoning-planning-execution capabilities and the task decomposition table, extract the start and completion times of real-time task decomposition, calculate the time difference between the start and end times of the task section, and compare it with the planned task node time difference to obtain the task time offset value; Based on the task node synchronization offset list, for example, the list contains nodes N001, N004, N005 to identify Agent-type intelligent assistants with reasoning-planning-execution capabilities. The Agent has specific role or ability tags in the configuration, such as marked as "Agent:RPE", and has the ability to adjust the plan according to environmental changes and call tools to execute tasks. Identify the task breakdown table related to the Agent's execution, and obtain the breakdown details of the Agent tasks (corresponding to the task nodes) that are currently being executed or recently executed from the Agent's task scheduling or task executor logs. The breakdown includes a series of finer-grained task segments or steps. For example, node N001 is decomposed into steps A.1 (data loading) and A.2 (data preprocessing). Extract the start and completion times of the real-time task breakdown. Obtain the actual start and end times of each sub-step or segment after the task breakdown of the Agent from the Agent's running log or task execution record. For example, Agent_A executes the task corresponding to node N001, where the real-time execution time of step A.1 starts at 10-27-10:02 and ends at 10-27-10:08, and the real-time execution time of step A.2 starts at 10-27-10:09 and ends at 10-27-10:17. Calculate the time difference between the start and end times of the task segment, that is, the real-time completion time minus the real-time start time. For example, the real-time time difference of step A.1 is 10:08 - 10:02 = 6 minutes, and the real-time time difference of step A.2 is 10:17 - 10:09 = 8 minutes. Compare it with the time difference of the planned task node. Here, it refers to comparing the duration of the real-time execution segment with the duration of this segment in the original plan. The original plan decomposes the total duration of the node into each segment. For example, the planned total duration of node N001 is 15 minutes, and the decomposed plan is that the planned duration of step A.1 is 5 minutes (for example, from 10:00 to 10:05), and the planned duration of step A.2 is 10 minutes (for example, from 10:05 to 10:15). Compare the real-time duration of step A.1, which is 6 minutes, with the planned duration of 5 minutes, and compare the real-time duration of step A.2, which is 8 minutes, with the planned duration of 10 minutes, to obtain the task time offset value. The offset value is defined as the real-time duration minus the planned duration. For example, the time offset value of step A.1 is 6 - 5 = +1 minute, where a positive value indicates delay and a negative value indicates early. The time offset value of step A.2 is 8 - 10 = -2 minutes. Repeat this process to obtain the time offset values of each task segment executed by the Agent. For node N001, its task segment time offset value is [+1, -2] minutes.

[0026] S212: Call the task time offset value, combine the segment distribution, offset trend, and adjustment frequency to collect the task node offset data, identify and calculate the trend offset degree according to the segment number, and judge the trend direction based on the adjustment frequency to obtain the task progress trend label group; Call the task time offset value. For example, the task section time offset value of node N001 is [+1, -2] minutes. Combine the section distribution, offset trend, and adjustment frequency. The section distribution refers to the relevance information such as the position of the task section in the entire task process, the pre - and post - relationship, and the task node it belongs to. The offset trend refers to whether the offset values of the same type of sections (such as all "data pre - processing" steps) or the sections executed by the same Agent are continuously positive (continuous delay), continuously negative (continuous advance), or fluctuating (alternating positive and negative) over a period of time. The adjustment frequency refers to how many retries, parameter adjustments, resource allocation changes, or plan modifications the user makes for this section, related tasks, or executing Agent. Aggregate the offset data of task nodes, associate the offset values (such as +1, -2), section numbers (such as A.1, A.2), executing Agent (such as Agent_A), and related adjustment frequencies of all task sections under the same task node (such as N001). Identify and calculate the trend offset degree according to the section number. For the same section type or number, calculate the average offset, cumulative offset, or maximum offset when it is executed multiple times (in different tasks or repeated execution of the same task) or on different Agents as the trend offset degree. For example, if step A.1 is executed multiple times in different tasks and its offset values are +1, +0.5, +2, +1.5 respectively, the average trend offset degree is (1 + 0.5 + 2 + 1.5) / 4 = +1.25 minutes. Judge the trend direction based on the adjustment frequency. If the offset degree of a certain section is high (for example, the average delay is more), and its adjustment frequency is also high, it indicates that there is a persistent problem resulting in repeated adjustments but still delays, and the trend direction is judged as "continuous delay and fluctuation". If the offset degree is positive but the adjustment frequency is low, it is a one - time problem or an unreasonable plan, and the trend direction is judged as "occasional delay". If the offset degree is close to zero and the adjustment frequency is low, it is judged as "stable progress". Here, the threshold for a "high" offset degree is set as the absolute value of the average offset being greater than 1.0 minute, and the threshold for a "high" adjustment frequency is set as 3 times per day. For example, if the average offset degree of step A.1 is +1.25 minutes (absolute value higher than 1.0 minute) and its adjustment frequency is 4 times per day (higher than 3 times per day), then its trend direction is judged as "continuous delay and fluctuation". If the average offset degree of step A.2 is -2 minutes (absolute value higher than 1.0 minute) and the adjustment frequency is 1 time per day (lower than 3 times per day), it is judged as "occasional advance". Associate the judgment result with the task node to obtain the task progress trend label group. For example, {N001: contains sections with continuous delay and fluctuation, N002: stable progress}. Specifically at the section level, the label group information is more detailed, such as (N001, A.1): continuous delay and fluctuation, (N001, A.2): occasional advance, …}

[0027] Please refer to Figure 4 , the specific steps for obtaining the list of sections affected by tool calls for task nodes are as follows: S311: Invoke the task progress trend tag group, filter the task partition numbers with fluctuations and offsets, extract the task time periods based on the node association table, process the section task time periods in the time dimension, identify the task time period index table, and obtain the set of active time periods for fluctuating tasks; Invoke the task progress trend tag group. For example, the tag group shows that node N001 contains "continuous delay fluctuation section". Filter the task partition numbers with fluctuations and offsets, and identify the task node numbers or section numbers in the tags that contain words such as "fluctuation" or "offset". For example, node N001 and its subordinate fluctuation sections (such as A.1). Extract the task time periods based on the node association table, and find the planned or actual execution time periods corresponding to the fluctuation nodes or sections from the task plan or execution records. For example, the planned time period of node N001 is [10:00, 10:15], the planned time period of the fluctuation section A.1 is [10:00, 10:05], and the actual execution period is [10:02, 10:08]. Process the section task time periods in the time dimension, sort, merge, or split the extracted time periods for subsequent time window analysis. For example, gather and sort the actual execution periods of all fluctuation sections to obtain a list of time periods. Identify the task time period index table, create an index, and associate the processed time periods with the corresponding fluctuation task nodes or sections. For example, the index entries are {[10:02, 10:08]:(N001,A1), [10:09, 10:17]:(N001,A.2)} (assuming A.2 is included in the fluctuation analysis although it is an occasional advance). Obtain the set of active time periods for fluctuating tasks, that is, the set containing the actual execution times of all fluctuation task nodes or sections. For example, the set is {[10:02,10:08], [10:09, 10:17]}.

[0028] S312: According to the set of active time periods for fluctuating tasks, collect the tool call frequencies in the same period sections, identify the impact table, judge the interference based on the interference threshold, match it with the construction time periods of the fluctuating tasks, analyze the tool intervention intensity, judge the abnormal association of progress monitoring, and generate a list of sections affected by tool calls for task nodes; According to the set of fluctuation task activity periods, for example, the set is {[10:02, 10:08], [10:09, 10:17]}, collect the tool call frequencies in the same period. During the activity period, count the number of times the Agent calls the tool interface or the number of calls per minute from the external API call records of the Agent system. For example, in the time period [10:02, 10:08] (a total of 6 minutes), the Agent called a certain tool APIX 15 times and APIY 5 times. In the time period [10:09, 10:17] (a total of 8 minutes), APIX was called 8 times and APIY 20 times. Calculate the average call frequency of APIX in [10:02, 10:08] as 15 / 6 = 2.5 times / minute. For APIY, according to the set of fluctuation task activity periods, for example, the set is {[10:02, 10:08], [10:09, 10:17]}, collect the tool call frequencies in the same period. During the activity period, count the number of times the Agent calls the tool interface or the number of calls per minute from the external API call records of the Agent system. For example, in the time period 10:02, 10:08 (a total of 6 minutes), the Agent called a certain tool APIX 15 times and APIY 5 times. In the time period 10:09, 10:17 (a total of 8 minutes), APIX was called 8 times and APIY 20 times. Calculate the average call frequency of APIX in 10:02, 10:08 as 15 / 6 = 2.5 times / minute, and the average call frequency of APIY is 5 / 6 ≈ 0.83 times / minute. Identify the impact table. The impact table is a predefined rule or data table that describes the types and intensities of the impacts of the call behaviors of specific tools (such as high-frequency calls, call failures, abnormal parameters) on the task progress. For example, the impact table entry {tool APIX call frequency > a certain threshold, potential impact: progress delay, impact intensity coefficient: 1.2}. Judge the interference based on the interference value. The interference value is used to define the degree to which the tool call affects the task progress, and its setting refers to the characteristics of the tool, the dependence of the Agent on the tool, and the normal tool call pattern. For example, for tool APIX, its normal average call frequency is 1.0 times / minute, and the interference threshold can be set to 1.5 times its normal mean, that is, 1.0 x 1.5 = 1.5 times / minute, indicating that an average call higher than 1.5 times per minute is determined to be interference. The setting of the threshold can be determined by analyzing the correlation between the tool call frequency and task anomalies or delays in the original operation logs. Here, the calculated frequency of APIX is 2.5 times / minute, and the threshold value is 1.5 times / minute. Since 2.5 > 1.5, it is determined that APIX has interference. The frequency of APIY is 0.83 times / minute. Assuming its threshold is 1.0 times / minute, then 0.83 < 1.0. Determine that APIY has no interference and matches the construction period of the fluctuation task. Check whether the tool call period determined to be interfering (i.e., the period for calculating the frequency [10:02, 10:08]) overlaps with the periods in the set of fluctuation task activity periods {[10:02, 10:08], [10:09, 10:17]} identified previously. [10:02, 10:08] overlaps with itself. Analyze the tool intervention intensity, which can be measured according to the degree or duration of the exceeded value and calculated in combination with the intensity coefficient in the impact table. For example, for the interference of APIX in [10:02, 10:08], the exceeded frequency is 2.5 - 1.5 = 1.0 times per minute, and the duration is 6 minutes. Assuming the impact intensity coefficient is 1.2, the intervention intensity is quantified as exceeded frequency × duration × impact coefficient = 1.0 × 6 × 1.2 = 7.2 unit intensities. Determine the abnormal association of progress monitoring. If there is tool interference and the interference period matches the fluctuation task period, then determine that the progress fluctuation anomaly (the fluctuation identified in S212, such as the continuous delay fluctuation of N001) is associated with the tool call interference, generate a list of task node sections affected by tool calls, and record information such as the task node numbers or section numbers determined to be affected by tool calls, their affected periods, tools, intervention intensities, etc.;. Table 2: List of task node sections affected by tool calls; 。

[0029] Referring to Table 2, the list of task node sections affected by tool calls shows the task progress fluctuations related to tool call interference, including the affected nodes or sections, the periods when interference occurs, the specific interfering tools, the quantified intervention intensities, and the associated task fluctuation labels.

[0030] Please refer to Figure 5 , the steps for obtaining the fluctuation group of intelligent agent node task completion are specifically as follows: S411: Based on the list of task node sections affected by tool calls, filter the unmarked partition nodes, extract and record the job task list of the nodes, including task numbers, planned processes, start times, and end times, to obtain the task set of nodes not interfered by tool calls; Based on the list of sections affected by tool calls for task nodes, for example, the list shows that section A.1 of node N001 is affected by tool calls. Filter the partition nodes that have not been marked, and select from all task nodes those that are not in the list of sections affected by tool calls generated in S312. The nodes are considered to be those not directly interfered with by tool calls. For example, node N003 is not in the list of sections affected by tools, nor is node N002. Extract and record the job task list of the nodes, and obtain the specific task list under the non-interfered nodes from the task decomposition plan table, including the number of each task, the description of the planned process, the planned start time, and the planned end time. For example, the job list of node N003: task number T003a, planned process "data preparation", planned start time 10:30, planned end time 10:40; task number T003b, planned process "model call", planned start time 10:40, planned end time 10:55. The job list of node N002: task number T002a, planned process "interface parameter configuration", planned start time 10:20, planned end time 10:25; task number T002b, planned process "interface call", planned start time 10:25, planned end time 10:35. Obtain the task set of nodes not interfered with by tool calls. This set contains the detailed planned job information of all task nodes determined not to be directly interfered with by tool calls. For example, the set contains {node: N003, task list: {number: T003a, process: data preparation, start: 10:30, end: 10:40}, {number: T003b, process: model call, start: 10:40, end: 10:55}, node: N002, task list: {number: T002a, process: interface parameter configuration, start: 10:20, end: 10:25}, {number: T002b, process: interface call, start: 10:25, end: 10:35}}.

[0031] S412: Invoke the task set of nodes not interfered with by tool calls, match the daily segmented task records of the agent with the operation log data, extract the planned task volume according to the task number, count the running duration and the real-time running period, and generate the agent task execution matching data set; Invoke the task set of nodes not interfered by tool calls. For example, the set contains tasks T003a and T003b (belonging to node N003), and tasks T002a and T002b (belonging to node N002). Match the agent's daily segmented task records with the operation log data. From the task execution log and the finer-grained operation log of the Agent system, find the actual execution records corresponding to the task numbers in the set. The records include information such as the actual start and end times of the task and the amount of tasks processed. Extract the planned task volume according to the task number, and obtain the expected workload of each task from the task plan or task decomposition table. For example, task T003a (data preparation) is planned to process 100 data records, task T003b (model call) is planned to process 50 requests, task T002a (interface parameter configuration) is planned to configure 10 parameters, and task T002b (interface call) is planned to initiate 20 calls. Statistically calculate the running duration and real-time running period. Record the actual start time, end time, and total duration of the task from the Agent operation log. For example, the actual execution time of task T003a starts at 10:32 and ends at 10:45, with a running duration of 13 minutes and a real-time running period of [10:32, 10:45]. The actual execution time of task T003b starts at 10:46 and ends at 10:50, with a running duration of 4 minutes and a real-time running period of [10:46, 10:50]. The actual execution time of task T002a starts at 10:20 and ends at 10:24, with a running duration of 4 minutes and a real-time running period of [10:20, 10:24]. The actual execution time of task T002b starts at 10:25 and ends at 10:32, with a running duration of 7 minutes and a real-time running period of [10:25, 10:32]. Generate an agent task execution matching data set, which associates the planned information with the actual execution information. For example, the data set contains the following entries: {task number: T003a, planned task volume: 100, planned duration: 10 minutes, real-time running duration: 13 minutes, real-time running period: [10:32, 10:45]}, {task number: T003b, planned task volume: 50, planned duration: 15 minutes, real-time running duration: 4 minutes, real-time running period: [10:46, 10:50]}, {task number: T002a, planned task volume: 10, planned duration: 5 minutes, real-time running duration: 4 minutes, real-time running period: [10:20, 10:24]}, {task number: T002b, planned task volume: 20, planned duration: 10 minutes, real-time running duration: 7 minutes, real-time running period: [10:25, 10:32]}.

[0032] S413: Match the dataset according to the agent task execution, judge the matching degree between the task execution efficiency and the running distribution, identify the efficiency fluctuation nodes and evaluate the task deviation degree, analyze the task efficiency fluctuation degree of the task nodes, mark the tasks with the fluctuation degree exceeding the set benchmark value as abnormal task nodes, and obtain the task completion fluctuation group of the agent nodes; Match the dataset according to the agent task execution. For example, if the dataset contains information about task T003a, judge the matching degree between the task execution efficiency and the running distribution. The task execution efficiency is measured by calculating the ratio of the planned task volume to the planned duration (planned efficiency) and the ratio of the actual completed task volume (assuming it is the same as the planned task volume) to the real-time running duration (actual efficiency). For task T003a, the planned efficiency is 100 / 10 = 10 tasks / minute, and the actual efficiency is 100 / 13 ≈ 7.69 tasks / minute. For task T003b, the planned efficiency is 50 / 15 ≈ 3.33 requests / minute, and the actual efficiency is 50 / 4 = 12.5 requests / minute. The running distribution refers to the characteristics such as the position of the actual execution period of the task relative to the planned period, whether it is continuous, and whether it exceeds the range. For example, for task T003a, the planned period is [10:30, 10:40], and the actual period is [10:32, 10:45]. The actual start is delayed by 2 minutes, and the end is delayed by 5 minutes. It is shifted backward and extended as a whole. For task T003b, the planned period is [10:40, 10:55], and the actual period is [10:46, 10:50]. The actual start is delayed by 6 minutes, and the end is advanced by 5 minutes. It is shifted backward and shortened as a whole. Identify the efficiency fluctuation nodes and evaluate the task deviation degree. If the actual efficiency is significantly lower or higher than the planned efficiency, or the running distribution is significantly different from the plan, it is considered that there is an efficiency fluctuation in this task or its affiliated node. The deviation degree can be evaluated by the ratio of the actual efficiency to the planned efficiency (efficiency ratio) or the absolute value of their difference. For example, for task T003a, the efficiency ratio = 7.69 / 10 = 0.769, and the deviation degree can be defined as |1 - efficiency ratio| = |1 - 0.769| = 0.231 (i.e., 23.1% deviation). For task T003b, the efficiency ratio = 12.5 / 3.33 ≈ 3.75, and the deviation degree is |1 - 3.75| = 2.75 (i.e., 275% deviation). Analyze the task efficiency fluctuation degree of the task node. The fluctuation degree is the quantified value of the deviation degree calculated above. Mark the tasks with a fluctuation degree exceeding the set benchmark value as abnormal task nodes. The benchmark value is used to define the tolerance of efficiency fluctuation, and its setting refers to the criticality of the task, the performance expectation of the Agent, the average fluctuation level of the original data, and the business requirements for efficiency stability. For example, for non-critical tasks, an efficiency fluctuation within 20% is acceptable, and the benchmark value is set to 0.20. For tasks with high performance requirements, if the fluctuation exceeds 10%, attention is required, and the benchmark value is set to 0.10. Here, the benchmark value is set to 0.20, indicating that tasks with an efficiency deviation exceeding 20% are marked as abnormal. Compare the fluctuation degree 0.231 of task T003a with the benchmark value 0.20. Since 0.231 > 0.20, task T003a is marked as abnormal. Compare the fluctuation degree 2.75 of task T003b with the benchmark value 0.20. Since 2.75 > 0.20. Task T003b is also marked as abnormal, obtaining the intelligent agent node task completion fluctuation group, which includes the task node numbers, task numbers marked as abnormal and their related efficiency fluctuation information. For example, {node number: N003, task number: T003a, efficiency deviation degree: 23.1%}, {node number: N003, task number: T003b, efficiency deviation degree: 275%}. The nodes in this fluctuation group are those task nodes whose own execution efficiency or rhythm is abnormal without direct tool interference.

[0033] Please refer to Figure 6 , and the specific steps for obtaining the task progress monitoring structure indicators are as follows: S511: Call the node numbers in the intelligent agent node task completion fluctuation group and the corresponding operation time intervals, identify the difference in task volume between adjacent task days, screen the nodes exceeding the threshold, record the time interval and the change range of task volume, and obtain the progress fluctuation abnormal identification set; Call the intelligent agent node task to complete the node numbers and corresponding job time intervals in the fluctuation group. For example, the fluctuation group contains node N003, and the actual execution time interval of its abnormal task T003a is [10:32, 10:45], and the actual execution time interval of task T003b is [10:46, 10:50]. Identify the difference in task volume between adjacent task days. For the nodes in the fluctuation group (such as N003), find the planned or actual task volume of the same type or related tasks executed on adjacent task days (such as yesterday 10-26 and today 10-27), and calculate the absolute value of the task volume difference. For example, node N003 contains another task T003c (planned to perform data cleaning every afternoon). The actual processed task volume on yesterday 10-26 was 500 pieces, and the actual processed task volume on today 10-27 was 450 pieces. The task volume difference is 500 - 450 = 50 pieces. Select the nodes with the exceeded value. The task volume difference threshold is used to define the significance of the task volume change, and its setting refers to the stability of the business requirements or the empirical value of the original task volume fluctuation range. For example, for tasks with an average daily fluctuation of the original task volume not exceeding 20 pieces, the threshold can be set to 30 pieces, or set to 10% of the task volume average value (if the task volume average value is 500 pieces, then the threshold is 500 x 10% = 50 pieces). Here, the threshold is set to 40 pieces, indicating that if the task volume difference exceeds 40 pieces, it is considered a significant change. Compare the task volume difference of 50 pieces of task T003c with the threshold of 40 pieces. Since 50 > 40, node N003 is screened out due to the significant change in task volume. Record the time interval and the amplitude of the task volume change, record the date interval when the task volume change occurs (such as from 10-26 to 10-27) and the specific task volume change value or percentage (such as a change of -50 pieces or a change of -10%). Obtain the progress fluctuation abnormal identification set. This set contains the task nodes marked due to the significant change in task volume between adjacent dates and their change details. For example, the set contains {node: N003, related task: T003c, change area: [10-26, 10-27], task volume change: -50 pieces}.

[0034] S512: Based on the knowledge base resources corresponding to the nodes in the progress fluctuation abnormal identification set, identify the sequence of node resource input ratios, extract the abnormal distribution interval and compare it with the critical coefficient, record the ratio deviation direction and the node number, and form a task resource matching deviation index group; Based on the knowledge base resources corresponding to the nodes identified by the progress fluctuation anomalies, for example, the anomaly identification set includes node N003 and its associated task T003c. The knowledge base resources associated with this task (data cleaning) are "Data Cleaning Rule Document v2.0" and "Abnormal Data Processing Guide". Identify the resource input ratio sequence of the nodes. The resource input ratio can be defined as the frequency (such as the number of accesses per minute) or duration of the Agent accessing or using specific knowledge base resources during the execution of the node task, or the ratio sequence of the resource update frequency to the task execution frequency. For example, when node N003 executes task T003c, the access frequency of "Data Cleaning Rule Document v2.0" on October 26 yesterday was 0.5 times per minute, and the access frequency when executing the same task on October 27 today was 0.2 times per minute, forming a resource input ratio sequence..., 0.5 (yesterday), 0.2 (today),.... Extract the abnormal distribution interval and compare it with the critical coefficient. The abnormal distribution interval refers to the period when significant changes occur in the resource input ratio sequence. For example, the ratio dropped from 0.5 to 0.2 from yesterday to today, and the change amplitude is 0.5 - 0.2 = 0.3. The critical coefficient is used to define the significance of the change in the resource input ratio, and its setting refers to the importance of the resource and the stability of the original ratio. For example, for the core rule document, its access frequency should not have large fluctuations, and the critical coefficient is set relatively low (such as 0.1). For reference documents, the coefficient with a high tolerance for fluctuations is set relatively high (such as 0.3). Here, the critical coefficient is set to 0.25, indicating that if the absolute value of the ratio change is greater than 0.25, it is considered that the resource input is abnormal. Here, the ratio change amplitude 0.3 > 0.25, and the abnormal distribution interval is from yesterday to today. Compare the critical coefficient to confirm the significance of the change, record the ratio deviation direction and the node number. The ratio dropped from 0.5 to 0.2, and the deviation direction is "down" or "negatively deviated". Record node number N003 and its associated task T003c to form a task-resource matching deviation index group. This group includes the task nodes marked due to abnormal changes in the resource input ratio and their resource matching deviation situations. For example, the index group includes {node: N003, associated task: T003c, associated resource: Data Cleaning Rule Document v2.0, ratio deviation direction: down, ratio change: -0.3}.

[0035] S513: According to the task-resource matching deviation index group, extract the knowledge base resource allocation and the task execution time period, analyze the difference between the resource allocation cycle and the operation cycle, identify the task time synchronization deviation distribution table, and sort and mark the difference according to the progress benchmark to generate the task progress monitoring structure index; Match the offset index group according to the task resources. For example, the index group shows that there is a resource matching decline offset in task T003c of node N003. Extract the knowledge base resource delivery and task execution time periods. Obtain the planned update, release, or available time period of the knowledge base resources from resource management, and obtain the planned or actual execution time period of the relevant tasks from the task plan or execution records. For example, the "Data Cleaning Rule Document v2.0" is scheduled for routine update at 10:00 am every Monday, and an emergency ad-hoc update was performed at 15:00 on October 27 (Tuesday) today. The task T003c of node N003 is scheduled to be executed from 2:00 pm to 4:00 pm every day, and the actual execution period today is from 2:05 pm to 4:10 pm. Analyze the difference between the resource delivery cycle and the job cycle, and compare the matching degree and timing relationship between the resource update or available cycle and the task execution cycle. For example, the resource was updated at 15:00 in the middle of the task execution today, while the task started at 2:05 pm, which means that the old version of the rule was used in the beginning stage of the task, or the Agent needs to interrupt the current execution to adapt to the new rule, resulting in a decrease in efficiency (associated with the efficiency fluctuation of T003c in S413), or combined with the task volume change identified in S511 (such as a 50-item reduction in the task volume), analyze whether it is due to resource update lag or resource content change that some data is marked as abnormal and not processed. Identify the task time synchronization offset distribution table, construct a table to record the task node, associated resources, planned resource delivery time, actual resource delivery time, planned task execution time period, actual task execution time period, as well as the timing difference and potential synchronization problems, and sort and mark the differences according to the progress benchmark. According to the planned progress benchmark of the entire project or phase, such as the total task completion time, key milestone attainment rate, etc., sort the entries in the task time synchronization offset distribution table, for example, sort by the degree of impact on the total progress, the task volume involved, the deviation duration, or the node criticality, and mark the degree or type of deviation (such as "resource lag type synchronization offset", "resource content change impact"); Table 3: Task Progress Synchronization Offset Distribution Table; 。

[0036] As shown in Table 3, the task progress synchronization offset distribution table summarizes various progress deviation situations related to time synchronization, covering aspects such as document updates, task execution, and resource allocation. By analyzing the structured deviation data, task progress monitoring structure indicators can be generated. These indicators quantify and summarize the structural problems in the synchronization of task progress with knowledge base resources, tools, and the execution efficiency of the Agent itself. For example, the indicators may include "the number of nodes affected by resource synchronization" (there is 1 node N003 in Table 3), "the average resource lag duration" (such as the resource of N003 lags nearly 2 hours), "the expected rework rate caused by resource mismatch" (evaluated based on the impact of resource changes on task results), "the proportion of delayed nodes caused by Agent efficiency fluctuations" (such as N003 due to the efficiency fluctuations of T003a and T003b), and "the cumulative delay duration caused by tool call interference" (calculated based on the intervention intensity and duration in Table 2), which are used for high-level progress monitoring, problem diagnosis, and optimization suggestions.

[0037] The multi-model agent collaboration system is used to execute the above multi-model agent collaboration method. The system includes: The task status extraction module obtains the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extracts the intelligent agent task node numbers and time intervals, compares the real-time execution status with the task node identifiers, marks the node tasks with inconsistent times, and generates a node status offset label group; The node trend classification module, based on the node status offset label group, locates the Agent-type intelligent assistant with reasoning - planning - execution capabilities, extracts the task decomposition table and planned time, identifies the difference between the real-time task time and the planned time, classifies and labels the difference types, and generates a task node progress trend identification group; The tool call identification module, based on the task node progress trend identification group, filters out the fluctuating and lagging task nodes, locates the corresponding partitions, extracts the tool call frequency interference time periods, determines the coincidence with the task time periods, and filters out the frequently interfered sections to generate a task progress tool call interference mapping set; The task efficiency diagnosis module, based on the task progress tool call interference mapping set, eliminates the tasks in the interference sections, extracts the intelligent agent task records and operation logs, matches the task assignment and operation time periods, calculates the ratio of the operation volume to the task, identifies the intensive but inefficient node tasks, and obtains a node task execution deviation set; The resource allocation analysis module, based on the node task execution deviation set, locates the intelligent agent resource input records, extracts the ratio of the task volume to the resource configuration, compares the difference between the execution cycle and the input cycle, maps the resource usage and task progress status, and forms task progress monitoring structure indicators.

[0038] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A multi-model agent collaboration method, characterized in that, It includes the following steps: S1: Obtain the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extract the execution plan numbers and time intervals of the intelligent agent task nodes, compare the real-time execution status of the intelligent agent with the planned node identifiers, identify the task node numbers deviating from the plan, and generate a task node synchronization offset list; S2: Based on the task node synchronization offset list, screen the Agent-type intelligent assistants, extract the start and end times of task decomposition in combination with the task decomposition table, perform difference calculation and trend classification with the planned time point, and generate a task progress trend label group; S3: Invoke the task progress trend label group, extract the numbers of fluctuating task partitions, identify the frequency data of regional tool calls, compare the tool call time period with the fluctuating task time period, record the number of overlapping time periods, and generate a list of sections affected by tool calls for task nodes; S4: Based on the list of sections affected by tool calls for task nodes, screen the task of partition nodes not affected, extract the segmented task records and operation log data of the intelligent agent, map the daily task distribution volume to the operation period of the intelligent agent, and determine whether there is an inefficient distribution of tasks to obtain a task completion fluctuation group for intelligent agent nodes.

2. The multi-model agent collaboration method according to claim 1, wherein, The task node synchronization offset list includes task offset numbers, execution status labels, time offsets, and task node classifications. The task progress trend label group includes progress offset levels, trend change types, planned comparison results, and task association numbers. The list of sections affected by tool calls for task nodes includes tool call types, affected time sections, the number of overlapping time periods, and affected task numbers. The task completion fluctuation group for intelligent agent nodes includes task distribution uneven numbers, operation duration records, task completion deviations, and intelligent agent operation matching degrees.

3. The multi-model agent collaboration method according to claim 1, wherein The specific steps for obtaining the task node synchronization offset list are as follows: S111: Obtain the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extract the planned times and numbers of intelligent agent task nodes, match the update time and coordinates of the knowledge base documents, compare the matching document time range with the node planned time, and generate a document recording period for partition nodes; S112: Based on the document recording period of the partition nodes, perform a coincidence judgment with the planned time interval of the task nodes, extract the ratio of the coincidence time to the total duration, screen the node numbers with a ratio lower than the benchmark value, and update the node document status annotation to obtain the progress coverage deviation rate of the partition nodes; S113: According to the progress coverage deviation rate of the partition nodes, determine the deviation status of the task node numbers, identify the task node numbers with a deviation rate higher than the node synchronization threshold, integrate the node numbers, document coverage information, and deviation rate values, and generate a task node synchronization offset list.

4. The multi-model agent collaboration method according to claim 3, wherein, The specific steps for obtaining the task progress trend label group are as follows: S211: Based on the task node synchronization offset list, identify the Agent-type intelligent assistants with reasoning-planning-execution capabilities and the task decomposition table, extract the start and end times of real-time task decomposition, calculate the difference in start and end times of the task section, compare it with the time difference of the planned task node, and obtain the task time offset value; S212: Call the task time offset value, combine the segment distribution, offset trend and adjustment frequency, collect task node offset data, identify and calculate the trend offset degree according to the segment number, determine the trend direction according to the adjustment frequency, and obtain the task progress trend label group.

5. The multi-model agent collaboration method according to claim 4, wherein The steps for obtaining the list of sections of the task node affected by the tool call are specifically as follows: S311: calling the task progress trend tag group, filtering the fluctuation offset task partition number, extracting the task time period according to the node association table, processing the segment task time period according to the time dimension, identifying the task time period index table, and obtaining the fluctuation task activity time period set; S312: According to the set of fluctuating task activity periods, collect tool call frequencies of the same period sections, identify the impact table, determine interference based on the interference threshold, match it with the fluctuating task construction period, analyze tool intervention intensity, determine abnormal associations with progress monitoring, and generate a list of sections of the task node affected by the tool call.

6. The multi-model agent collaboration method according to claim 5, characterized in that The steps for obtaining the agent node task completion fluctuation group are specifically as follows: S411: Based on the list of sections affected by the tool call of the task node, filter out the partition nodes that have not been marked, extract and record the job task list of the node, including the task number, planned process, start time and end time, and obtain the task set of the node that is not affected by the tool call; S412: calling the node task set that is not disturbed by the tool call, matching the daily segmented task records of the agent with the operation log data, extracting the planned task volume according to the task number, counting the operation time and the real-time operation time period, and generating the agent task execution matching data set; S413: Based on the intelligent agent task execution matching data set, determine the matching degree between the task execution efficiency and the operation distribution, identify the efficiency fluctuation nodes and evaluate the task deviation degree, analyze the task efficiency fluctuation degree of the task nodes, mark the tasks whose fluctuation degree exceeds the set benchmark value as abnormal task nodes, and obtain the intelligent agent node task completion fluctuation group.

7. The multi-model agent collaboration method according to claim 1, wherein The method further comprises step S5: S5: calling the task completion fluctuation group of the agent node, extracting the abnormal fluctuation task group, calculating the ratio of resource input to task volume in the knowledge base, recording the difference distribution between the resource input cycle and the task execution cycle, and generating task progress monitoring structure indicators; The task progress monitoring structure indicators include resource allocation ratio, task intensity level, execution cycle deviation, and task efficiency index.

8. The multi-model agent collaboration method according to claim 7, characterized in that The steps for obtaining the task progress monitoring structure indicator are specifically as follows: S511: calling the node number and the corresponding operation time interval in the task completion fluctuation group of the agent node, identifying the difference in task volume of adjacent task days, filtering out nodes that exceed the threshold, recording the time interval and task volume change range, and obtaining a progress fluctuation abnormality identification set; S512: Based on the knowledge base resources corresponding to the nodes in the progress fluctuation abnormality identification set, identify the node resource input ratio sequence, extract the abnormal distribution interval and compare the critical coefficient, record the ratio deviation direction and the node number, and form a task resource matching deviation indicator group; S513: Match the offset index group according to the task resources, extract the knowledge base resource placement and task execution time period, analyze the difference between the resource placement cycle and the operation cycle, identify the task time synchronization offset distribution table, and sort and label the difference according to the progress benchmark to generate the task progress monitoring structure index.

9. A multi-model agent collaboration system, characterized in that, The system is used to implement the multi-model agent collaboration method according to any one of claims 1-8, and the system includes: The task status extraction module obtains the knowledge base documents, tool interface configuration information, and external API call records in the task scenario, extracts the intelligent agent task node numbers and time intervals, compares the real-time execution status with the task node identifiers, marks the node tasks with inconsistent times, and generates a node status offset label group; The node trend classification module, based on the node status offset label group, locates the Agent-type intelligent assistant with reasoning-planning-execution capabilities, extracts the task decomposition table and the planned time, identifies the difference between the real-time task time and the planned time, classifies and labels the difference types, and generates a task node progress trend identification group; The tool call identification module, based on the task node progress trend identification group, filters out the fluctuating and lagging task nodes, locates the corresponding partitions, extracts the tool call frequency interference time periods, determines the coincidence with the task time periods, filters out the frequently interfering sections, and generates a task progress tool call interference mapping set; The task efficiency diagnosis module, based on the task progress tool call interference mapping set, eliminates the tasks in the interference sections, extracts the intelligent agent task records and operation logs, matches the task assignment and the operation time periods, calculates the ratio of the operation volume to the task, identifies the intensive but inefficient node tasks, and obtains a node task execution deviation set; The resource allocation analysis module, based on the node task execution deviation set, locates the intelligent agent resource input records, extracts the ratio of the task volume to the resource configuration, compares the difference between the execution cycle and the input cycle, maps the resource usage and the task progress status, and forms a task progress monitoring structure index.

Images