Site inspection methods, devices, and equipment based on AI agent-based collaborative scheduling

Abstract

This invention discloses a site inspection method, apparatus, and device based on collaborative scheduling of artificial intelligence agents. The method includes: a data preparation module truncating Markdown text to a set character length, compressing screenshots and converting them to base64 format to generate a structured message containing execution steps and prompts; a language analysis module using a large-scale language model to process the structured message and output a structured inspection report; a scheduling engine writing event information from each inspection stage to a database and synchronously updating the memory cache, pushing events to the visualization layer via an SSE mechanism; a columnar database storing the structured inspection report and event records for each stage; a visualization layer displaying the inspection timeline in real time; and retrieving the structured inspection report and displaying it in card format after inspection is completed. This approach achieves real-time performance, cost-effectiveness, traceability, and ease of use for multi-agent collaborative site inspection.

Description

Technical Field

[0001] This invention relates to the field of aviation management technology, and in particular to a site inspection method, apparatus and equipment based on collaborative scheduling of artificial intelligence agents. Background Technology

[0002] In existing site inspection technologies, most solutions rely on manual spot checks or single-function scripts, only capable of basic checks on site accessibility. They cannot automate analyses of page content legality and compliance, sensitive words, or NSFW (Not Safe For Work) risks. Furthermore, existing scripts are mostly monolithic, exhibiting four core flaws: First, concurrent inspections cannot record the execution stage and time of each subtask, leading to difficult and inefficient fault localization. Second, they fail to assess whether site content has changed, repeatedly calling the algorithm model even when content remains unchanged, resulting in significant waste of GPU / CPU (Graphics Processing Unit) / Central Processing Unit computing power. Third, inspection results are stored in fragmented logs, hindering querying, statistics, and historical comparison, creating data silos. Fourth, a unified user interface is lacking; inspection plan configuration, execution status viewing, and data visualization are disconnected, preventing integrated scheduling. These issues render existing inspection solutions inadequate in terms of real-time performance, cost-effectiveness, traceability, and ease of use for practical applications. Summary of the Invention

[0003] In view of this, the purpose of this invention is to propose a site inspection method, device and equipment based on the collaborative scheduling of artificial intelligence agents, which can realize the real-time performance, economy, traceability and ease of use of multi-agent collaborative site inspection.

[0004] According to one aspect of the present invention, a site inspection method based on collaborative scheduling of artificial intelligence agents is provided, comprising: S1: a visualization layer receiving at least one target Uniform Resource Locator (URL) and corresponding inspection task type input by a user, wherein the inspection task type includes one or more of compliance detection, timeliness detection, and availability detection; S2: the visualization layer triggers a scheduling engine through a Representational State Transition (REST) ​​interface, the scheduling engine assigns an independent task to each target URL and manages the overall inspection progress through a flow mechanism; S3: a headless crawling module responds to the scheduling engine's instructions, uses an automated browser to access the target URL, performs page scrolling and waits for Document Object Model (DOM) rendering operations as needed, and outputs the target page in a lightweight markup language (Markdown). S4: The data preparation module truncates the lightweight Markdown text to a set character length, compresses the screenshot and converts it to base64 format, generating a structured message containing execution steps and prompts; S5: The language analysis module calls a large-scale language model to process the structured message and outputs a structured inspection report containing status, site name, latest date, sensitive words, and summary; S6: The scheduling engine writes the event information of each inspection stage to the database and updates the memory cache synchronously. It pushes events to the visualization layer through the server-side event sending (SSE) mechanism; S7: The columnar database stores the structured inspection report and event records of each stage. The visualization layer displays the inspection progress timeline in real time. After the inspection is completed, the structured inspection report is retrieved and displayed in card format.

[0005] After step S3 is completed, the content fingerprint caching step is also included: calculating the content hash value of the target page to form a cache identifier of Uniform Resource Locator URL-Inspection Task Type-Content Hash, and querying whether there is a matching cache identifier in the cache; if there is a matching identifier, the historical structured inspection report is retrieved directly from the cache, skipping steps S4-S5; if there is no matching identifier, steps S4-S5 are executed, and the current cache identifier is associated with the structured inspection report and stored in the cache.

[0006] In step S6, the columnar database includes an inspection record table, an inspection configuration table, an inspection task table, a task execution association table, a process flow mechanism process operation event table, and a site baseline data table. The site baseline data table records the first crawl time of each target's Uniform Resource Locator (URL). In step S7, if the latest date field of the structured inspection report is unknown, the visualization layer retrieves the first crawl time from the site baseline data table and displays a prompt message indicating that no obvious updates have been identified since YYYY-MM-DD HH:MM.

[0007] The process also includes a scheduled inspection configuration step: The user inputs the target Uniform Resource Locator (URL), inspection task type, and scheduled task configuration Cron expression through the task settings page in the visualization layer; the scheduling engine generates a corresponding Deployment based on the input information, and establishes the association between the URL, inspection task type, and Deployment through the Deployment_id field of the inspection_tasks table; the scheduling engine triggers the scheduled inspection according to the scheduled task configuration Cron expression, executing steps S3-S7; the inspection dashboard in the visualization layer aggregates and displays the task success rate, recent running status, and historical difference data of the scheduled inspection.

[0008] In step S6, the event push logic of the server sending event SSE mechanism is as follows: The degree engine monitors the status changes of each stage of the inspection in real time, and generates various events such as Flow.start (start of process flow mechanism), inspect_start (start of inspection), crawl_start (start of collection), crawl_complete (complete of collection), analysis_start (start of analysis), analysis_complete (complete of analysis), inspect_complete (complete of inspection), and Flow.complete (complete of process flow mechanism). These events are pushed to the visualization layer through the server sending event SSE interface. After pushing the inspect_complete or Flow.complete event, the event retry push is automatically stopped.

[0009] This includes a step for comparing inspection results: receiving a user-initiated comparison request and specifying two inspection records from different time points; retrieving two corresponding structured inspection reports from the inspection record table and comparing the site name, latest date, sensitive words, and summary fields one by one to generate a list of differences and display it in the visualization layer.

[0010] In step S4, the aggregation display logic of the inspection dashboard also includes: real-time monitoring of the status field in the structured report of the scheduled inspection; if the status field is critical, the preset inter-system real-time communication webhook interface is triggered to push alarm information to the administrator. The alarm information includes the Uniform Resource Locator URL, inspection time, anomaly type, and difference list.

[0011] In step S2, the scheduling engine's flow mechanism also includes: assigning a unique flow run ID to each independent task, recording the task's execution order, dependencies, and resource usage; and displaying the event information, time consumption data, and exception prompts of each stage of the corresponding task through the flow run ID.

[0012] According to another aspect of the present invention, a site inspection device based on collaborative scheduling of artificial intelligence agents is provided, comprising: a visualization layer, a headless crawling module, a data preparation module, a language analysis module, a scheduling engine, and a columnar database; the visualization layer is used to receive at least one target Uniform Resource Locator (URL) and corresponding inspection task type input by a user, wherein the inspection task type includes one or more of compliance detection, timeliness detection, and availability detection; and triggers the scheduling engine through a Representational State Transition (REST) ​​interface, wherein the scheduling engine assigns an independent task to each target URL and manages the overall inspection progress through a flow mechanism; the headless crawling module is used to respond to instructions from the scheduling engine, use an automated browser to access the target URL, perform page scrolling and wait for Document Object Model (DOM) rendering operations as needed, and output the target page. The system includes: a lightweight Markdown text processing module and optional screenshots; a data preparation module that truncates the Markdown text to a set character length, compresses screenshots and converts them to base format, generating structured messages containing execution steps and prompts; a language analysis module that uses a large-scale language model to process the structured messages and outputs a structured inspection report containing status, site name, latest date, sensitive words, and a summary; a scheduling engine that writes event information from each stage of the inspection to the database and updates the memory cache synchronously, and pushes events to the visualization layer via the Server-Send-Event (SSE) mechanism; and a columnar database that stores the structured inspection report and event records for each stage. The visualization layer displays the inspection progress timeline in real time, and retrieves the structured inspection report after the inspection is completed and displays it in card format.

[0013] According to another aspect of the present invention, a computer device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the site inspection method based on collaborative scheduling of artificial intelligence agents as described in any of the preceding claims.

[0014] According to another aspect of the present invention, a computer-readable storage medium is provided, storing a computer program that, when executed by a processor, implements the site inspection method based on collaborative scheduling of artificial intelligence agents as described in any one of the preceding claims.

[0015] It can be seen that the above solutions solve the problems of invisible inspection progress and difficult fault location in existing technologies: By recording events at each stage through the flow mechanism and combining them with real-time push of events sent by the server via SSE, the delay in inspection progress is effectively shortened. The visualization layer can accurately display the status of capturing / analyzing / completing. Compared with existing solutions, the average time spent on fault location is reduced. Multi-task parallel scheduling is achieved: each Uniform Resource Locator (URL) is independently assigned tasks, breaking through the serial limitation of traditional monolithic scripts and greatly improving the efficiency of simultaneous inspection of multiple sites. Standardized inspection output results are achieved: structured reports are generated through the language analysis module, avoiding the fragmented problems of existing log formats and laying the foundation for subsequent querying and statistics. Through the above methods, the real-time, economical, traceable, and easy-to-use multi-agent collaborative site inspection can be achieved.

[0016] Furthermore, the above solution addresses the problem of wasted computing power caused by repeated inference when the content remains unchanged in existing technologies. It achieves intelligent deduplication of repeated inspections through content fingerprint caching: in sites with stable content, the hit rate of repeated inspections can be effectively improved, saving GPU computing power, significantly reducing hardware resource consumption for inspections, and improving the economic efficiency of inspections.

[0017] Furthermore, the above solutions address the problem of isolated inspection data in existing technologies: the columnar database supports centralized storage of tens of thousands of inspection records, with controllable query time, enabling rapid retrieval, statistics, and historical comparison of inspection results; and resolve the auditing challenges of sites lacking timestamps: by restoring the initial capture time through the site baseline data table, the system avoids missing audit information due to unknown latest dates, thereby improving the completeness and audit availability of inspection reports.

[0018] Furthermore, the above solution addresses the disconnect between inspection plan configuration and visualization in existing technologies, achieving the integration of scheduled inspections and the visualization system: users can complete scheduled task configuration, status viewing, and data statistics without switching interfaces. The dashboard's aggregated display function effectively improves inspection efficiency and provides a standardized solution for routine inspections of batch sites.

[0019] Furthermore, compared to the existing technology that relies on front-end polling to obtain progress, the above solution uses the server-sent event SSE mechanism to actively push events, effectively shortening the progress delay control time and significantly improving the real-time performance of the visualization layer. At the same time, by automatically terminating and retrying, it reduces network bandwidth consumption and front-end resource usage, thereby improving system stability.

[0020] Furthermore, the above solution addresses the difficulty of comparing historical records in existing technologies. By comparing fields, it quickly locates changes in page content, such as the addition of sensitive words or changes in update dates, effectively improving the efficiency of difference detection. This provides direct data support for compliance auditing and content change tracking, while reducing the workload and error rate of manual comparison.

[0021] Furthermore, the above solution addresses the issue of untimely response to anomalies in existing scheduled inspections: when a site experiences a critical state such as compliance risks or availability failures, administrators can receive alerts in real time, reducing the anomaly response time from the traditional hours to minutes, significantly lowering compliance risks and business losses.

[0022] Furthermore, the above solution addresses the issues of chaotic task trajectories and difficulty in locating anomalies in existing concurrent inspections: the Flow Run ID mechanism allows for quick association and querying of the complete execution logs of a single task, including information such as the time consumed at each stage and the anomaly trigger point, effectively improving the efficiency of single-task fault diagnosis and providing accurate performance data support for system optimization. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an embodiment of the site inspection method based on collaborative scheduling of artificial intelligence agents according to the present invention;

[0024] Figure 2 This is a schematic diagram of a site inspection device based on the collaborative scheduling of artificial intelligence agents according to the present invention.

[0025] Figure 3 This is a schematic diagram of the structure of an embodiment of the computer device of the present invention. Detailed Implementation

[0026] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be particularly noted that the following embodiments are for illustrative purposes only and do not limit the scope of the invention. Similarly, the following embodiments are only some, not all, embodiments of the present invention, and all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] This invention provides a site inspection method based on collaborative scheduling of artificial intelligence agents, which can achieve real-time, economical, traceable and easy-to-use site inspection through multi-agent collaboration.

[0028] Please see Figure 1 , Figure 1This is a flowchart illustrating an embodiment of the site inspection method based on collaborative scheduling of artificial intelligence agents according to the present invention. It should be noted that if substantially the same result is obtained, the method of the present invention is not necessarily identical. Figure 1 The illustrated process sequence is limited. For example... Figure 1 As shown, the method includes the following steps:

[0029] S1: The visualization layer receives at least one target URL (Uniform Resource Locator) and the corresponding inspection task type input by the user. The inspection task type includes one or more of compliance detection, timeliness detection, and availability detection.

[0030] S2: The visualization layer triggers the scheduling engine through the REST (Representational State Transfer) interface. The scheduling engine assigns an independent task to each target Uniform Resource Locator (URL) and manages the overall inspection progress through the Flow mechanism.

[0031] S3: The headless crawling module responds to the scheduling engine's instructions, uses an automated browser to access the target Uniform Resource Locator (URL), performs page scrolling and waits for DOM (Document Object Model) rendering operations as needed, and outputs the Markdown text of the target page and optional screenshots.

[0032] S4: The data preparation module truncates the lightweight markup language Markdown text to a set character length, compresses the screenshot and converts it to base64 format, and generates a structured message containing execution steps and prompts.

[0033] S5: The language analysis module calls a large-scale language model to process structured messages and outputs a structured inspection report containing status, site name, latest date, sensitive words, and summary.

[0034] S6: The scheduling engine writes the event information of each stage of the inspection to the database and updates the memory cache synchronously. It pushes the events to the visualization layer through the SSE (Server-Sent Events) mechanism.

[0035] S7: The columnar database stores structured inspection reports and event records for each stage. The visualization layer displays the inspection progress timeline in real time. After the inspection is completed, the structured inspection report is retrieved and displayed in card format.

[0036] In this embodiment, the solution achieves integrated inspection of multiple Uniform Resource Locators (URLs) and multiple task types through a closed-loop architecture consisting of a visualization layer, a scheduling engine, a multi-functional module, data storage, and real-time push. The core of this solution lies in using the scheduling engine's flow mechanism to achieve parallel management of multiple tasks, a headless crawling module to ensure complete acquisition of page content, a data preparation module to standardize data formats to adapt to language models, a server-side event sending (SSE) mechanism to achieve real-time event push, and a columnar database to centrally store results.

[0037] It can be observed that, in this embodiment, the technical effects of this solution are as follows: It solves the problems of invisible inspection progress and difficult fault location in the prior art: By recording events at each stage through the process flow mechanism and combining it with the server to send events in real time via SSE, the delay in inspection progress is effectively shortened. The visualization layer can accurately display the status of capturing / analyzing / completing. Compared with the existing solutions, the average time for fault location is reduced. At the same time, it realizes multi-task parallel scheduling: Each Uniform Resource Locator (URL) is independently assigned tasks, breaking through the serial limitation of traditional single scripts and greatly improving the efficiency of simultaneous inspection of multiple sites. Furthermore, it generates structured reports through the language analysis module to achieve standardized inspection output results, avoiding the fragmented problem of existing log formats and laying the foundation for subsequent querying and statistics. Through the above methods, the real-time performance, economy, traceability, and ease of use of multi-agent collaborative site inspection can be achieved.

[0038] After step S3 is completed, the content fingerprint caching step is also included: calculating the content hash value of the target page to form a cache identifier of Uniform Resource Locator URL-Inspection Task Type-Content Hash, and querying whether there is a matching cache identifier in the cache; if there is a matching identifier, the historical structured inspection report is retrieved directly from the cache, skipping steps S4-S5; if there is no matching identifier, steps S4-S5 are executed, and the current cache identifier is associated with the structured inspection report and stored in the cache.

[0039] In this embodiment, the solution adds a content fingerprint verification mechanism after page crawling. It uses a hash algorithm to uniquely identify the core content of the page, establishes a cache mapping relationship, and realizes the logic of reusing historical results if the content has not changed.

[0040] It can be seen that in this embodiment, in order to address the problem of wasted computing power caused by repeated inference when the content has not changed in the prior art, intelligent deduplication of repeated inspection is achieved by using content fingerprint caching: in sites with stable content, the hit rate of repeated inspection can be effectively improved, saving GPU computing power, significantly reducing the hardware resource consumption of inspection, and improving the economic efficiency of inspection.

[0041] In step S6, the columnar database includes an inspection record table, an inspection configuration table, an inspection task table, a task execution association table, a process flow mechanism process operation event table, and a site baseline data table. The site baseline data table records the first crawl time of each target's Uniform Resource Locator (URL). In step S7, if the latest date field of the structured inspection report is unknown, the visualization layer retrieves the first crawl time from the site baseline data table and displays a prompt message indicating that no obvious updates have been identified since YYYY-MM-DD HH:MM.

[0042] In this embodiment, the solution achieves structured storage of inspection data through a multi-table design of a columnar database. It focuses on recording the first capture time through the site baseline data table, providing backup data support for scenarios where the latest date is unknown.

[0043] It can be seen that, in this embodiment, the technical effects of this solution are as follows: It solves the problem of isolated inspection data in the prior art: the columnar database supports centralized storage of tens of thousands of inspection records, with controllable query time, enabling rapid retrieval, statistics, and historical comparison of inspection results; it solves the auditing problems of sites lacking timestamps: by restoring the initial capture time through the site baseline data table, it avoids the loss of audit information caused by the unknown latest date, improving the completeness and audit availability of inspection reports.

[0044] This also includes the steps for configuring scheduled inspections:

[0045] The task settings page in the visualization layer receives user input of the target Uniform Resource Locator URL, inspection task type, and scheduled task configuration Cron expression.

[0046] The scheduling engine generates corresponding deployment instances (Deployments) based on the input information, and establishes the association between the Uniform Resource Locator (URL), the inspection task type, and the deployment instance (Deployment) through the Deployment_id field of the inspection_tasks table.

[0047] The scheduling engine triggers a scheduled inspection according to the Cron expression configured for the scheduled task, and executes steps S3-S7.

[0048] The inspection dashboard in the visualization layer aggregates and displays the success rate of scheduled inspection tasks, recent running status, and historical difference data.

[0049] In this embodiment, the solution uses Cron expressions to configure scheduled tasks to achieve flexible configuration of inspection cycles, associates tasks and scheduling rules with Deployment instances, and combines dashboards to achieve centralized management and data visualization of scheduled tasks.

[0050] It can be seen that in this embodiment, the problem of the separation between inspection plan configuration and visualization display in the prior art is solved, and the integration of timed inspection and visualization system is realized: users can complete the configuration of timed tasks, status viewing and data statistics without switching interfaces. The aggregated display function of the dashboard effectively improves the inspection efficiency, and at the same time provides a standardized solution for the routine inspection of batch sites.

[0051] In step S6, the event push logic of the server sending event SSE mechanism is as follows: The degree engine monitors the status changes of each stage of the inspection in real time and generates various events such as Flow.start, inspection start, crawl_start, crawl_complete, analysis start, analysis complete, inspection complete, and Flow.complete. These events are then pushed to the visualization layer through the server sending event SSE interface. After pushing the inspection complete (inspect_complete) or Flow.complete event, the event retry push is automatically stopped.

[0052] In this embodiment, the solution defines standardized event types for the entire inspection process, achieves event synchronization through a one-way real-time push mechanism of the server sending events SSE, and clarifies the push termination conditions to avoid invalid retries.

[0053] It can be observed that, in this embodiment, compared with the prior art's reliance on front-end polling to obtain progress, the server-sent event SSE mechanism enables proactive event pushing, effectively shortening the progress delay control time and significantly improving the real-time performance of the visualization layer; at the same time, by automatically terminating and retrying, it reduces network bandwidth consumption and front-end resource occupation, thereby improving system operational stability.

[0054] This includes a step for comparing inspection results: receiving a user-initiated comparison request and specifying two inspection records from different time points; retrieving two corresponding structured inspection reports from the inspection record table and comparing the site name, latest date, sensitive words, and summary fields one by one to generate a list of differences and display it in the visualization layer.

[0055] In this embodiment, the solution is based on the historical inspection records stored in a columnar database to achieve accurate comparison of structured fields at a specified time point and output a standardized difference list.

[0056] It can be seen that in this embodiment, the problem of difficulty in comparing historical records in the prior art is solved. By comparing fields, changes in page content such as the addition of sensitive words or changes in update dates can be quickly located, which effectively improves the efficiency of difference discovery, provides direct data support for compliance auditing and content change tracking, and reduces the workload and error rate of manual comparison.

[0057] In step S4, the aggregation display logic of the inspection dashboard also includes: real-time monitoring of the status field in the structured report of the scheduled inspection; if the status field is critical, a preset inter-system real-time communication (webhook) interface is triggered to push alarm information to the administrator. The alarm information includes a Uniform Resource Locator URL, inspection time, anomaly type, and a list of differences.

[0058] In this embodiment, the solution adds an abnormal state monitoring and alarm triggering mechanism on the basis of timed inspection, and realizes the instant push of alarm information through the real-time communication webhook interface between systems.

[0059] It can be seen that in this embodiment, the problem of not being able to respond to abnormal situations in a timely manner in the existing timed inspection is solved: when a site experiences a critical state such as compliance risk or availability failure, the administrator can receive alarms in real time, and the abnormal response time is shortened from the traditional hour level to the minute level, which significantly reduces compliance risk and business losses.

[0060] In step S2, the scheduling engine's flow mechanism also includes: assigning a unique flow run ID to each independent task, recording the task's execution order, dependencies, and resource usage; and displaying the event information, time consumption data, and exception prompts of each stage of the corresponding task through the flow run ID.

[0061] In this embodiment, the solution uses the Flow Run ID mechanism to uniquely identify tasks and establishes a mapping between tasks, events, and logs, enabling precise traceability of the execution trajectory of each task.

[0062] It can be seen that in this embodiment, the problems of chaotic task trajectories and difficulty in locating anomalies in existing concurrent inspections are solved: the Flow Run ID mechanism can quickly associate and query the complete execution log of a single task, including the time consumed at each stage, anomaly trigger points and other information, which effectively improves the efficiency of single task fault diagnosis and provides accurate performance data support for system optimization.

[0063] The present invention also provides a site inspection device based on the collaborative scheduling of artificial intelligence agents, which can realize the real-time, economical, traceable and easy-to-use nature of multi-agent collaborative site inspection.

[0064] Please see Figure 2 , Figure 2 This is a schematic diagram of an embodiment of the site inspection device based on collaborative scheduling of artificial intelligence agents according to the present invention. In this embodiment, the site inspection device 20 based on collaborative scheduling of artificial intelligence agents includes a visualization layer 21, a headless crawling module 22, a data preparation module 23, a language analysis module 24, a scheduling engine 25, and a columnar database 26.

[0065] The visualization layer 21 is used to receive at least one target Uniform Resource Locator (URL) and the corresponding inspection task type input by the user. The inspection task type includes one or more of compliance detection, timeliness detection, and availability detection. The scheduling engine is triggered through the Representational State Transition (REST) ​​interface. The scheduling engine assigns an independent task to each target Uniform Resource Locator URL and manages the overall inspection progress through the Flow mechanism.

[0066] The headless crawling module 22 is used to respond to the scheduling engine's instructions, use an automated browser to access the target Uniform Resource Locator URL, perform page scrolling and wait for Document Object Model (DOM) rendering operations as needed, and output the target page's lightweight Markdown text and optional screenshots.

[0067] The data preparation module 23 is used to truncate the lightweight markup language Markdown text to a set character length, compress the screenshot and convert it to base64 format, and generate a structured message containing execution steps and prompts.

[0068] The language analysis module 24 is used to call a large-scale language model to process structured messages and output a structured inspection report containing status, site name, latest date, sensitive words, and summary.

[0069] The scheduling engine 25 is used to write event information of each stage of inspection into the database and update the memory cache synchronously. It also pushes events to the visualization layer through the server-sent event SSE mechanism.

[0070] This columnar database 26 is used to store structured inspection reports and event records at each stage. The visualization layer displays the inspection progress timeline in real time. After the inspection is completed, the structured inspection report is retrieved and displayed in card format.

[0071] Each unit module of the site inspection device 20 based on the collaborative scheduling of artificial intelligence agents can execute the corresponding steps in the above method embodiment. Therefore, the details of each unit module will not be elaborated here. Please refer to the description of the corresponding steps above for details.

[0072] This invention also provides a computer device, such as... Figure 3 As shown, it includes: at least one processor 31; and a memory 32 communicatively connected to at least one processor 31; wherein the memory 32 stores instructions that can be executed by at least one processor 31, and the instructions are executed by at least one processor 31 to enable at least one processor 31 to execute the above-described site inspection method based on collaborative scheduling of artificial intelligence agents.

[0073] The memory 32 and processor 31 are connected via a bus, which may include any number of interconnecting buses and bridges, connecting various circuits of one or more processors 31 and memory 32. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 31 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 31.

[0074] Processor 31 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 32 can be used to store data used by processor 31 during operation.

[0075] The present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the above-described method embodiments.

[0076] It can be seen that the above solutions solve the problems of invisible inspection progress and difficult fault location in existing technologies: By recording events at each stage through the flow mechanism and combining them with real-time push of events sent by the server via SSE, the delay in inspection progress is effectively shortened. The visualization layer can accurately display the status of capturing / analyzing / completing. Compared with existing solutions, the average time spent on fault location is reduced. Multi-task parallel scheduling is achieved: each Uniform Resource Locator (URL) is independently assigned tasks, breaking through the serial limitation of traditional monolithic scripts and greatly improving the efficiency of simultaneous inspection of multiple sites. Standardized inspection output results are achieved: structured reports are generated through the language analysis module, avoiding the fragmented problems of existing log formats and laying the foundation for subsequent querying and statistics. Through the above methods, the real-time, economical, traceable, and easy-to-use multi-agent collaborative site inspection can be achieved.

[0077] Furthermore, the above solution addresses the problem of wasted computing power caused by repeated inference when the content remains unchanged in existing technologies. It achieves intelligent deduplication of repeated inspections through content fingerprint caching: in sites with stable content, the hit rate of repeated inspections can be effectively improved, saving GPU computing power, significantly reducing hardware resource consumption for inspections, and improving the economic efficiency of inspections.

[0078] Furthermore, the above solutions address the problem of isolated inspection data in existing technologies: the columnar database supports centralized storage of tens of thousands of inspection records, with controllable query time, enabling rapid retrieval, statistics, and historical comparison of inspection results; and resolve the auditing challenges of sites lacking timestamps: by restoring the initial capture time through the site baseline data table, the system avoids missing audit information due to unknown latest dates, thereby improving the completeness and audit availability of inspection reports.

[0079] Furthermore, the above solution addresses the disconnect between inspection plan configuration and visualization in existing technologies, achieving the integration of scheduled inspections and the visualization system: users can complete scheduled task configuration, status viewing, and data statistics without switching interfaces. The dashboard's aggregated display function effectively improves inspection efficiency and provides a standardized solution for routine inspections of batch sites.

[0080] Furthermore, compared to the existing technology that relies on front-end polling to obtain progress, the above solution uses the server-sent event SSE mechanism to actively push events, effectively shortening the progress delay control time and significantly improving the real-time performance of the visualization layer. At the same time, by automatically terminating and retrying, it reduces network bandwidth consumption and front-end resource usage, thereby improving system stability.

[0081] Furthermore, the above solution addresses the difficulty of comparing historical records in existing technologies. By comparing fields, it quickly locates changes in page content, such as the addition of sensitive words or changes in update dates, effectively improving the efficiency of difference detection. This provides direct data support for compliance auditing and content change tracking, while reducing the workload and error rate of manual comparison.

[0082] Furthermore, the above solution addresses the issue of untimely response to anomalies in existing scheduled inspections: when a site experiences a critical state such as compliance risks or availability failures, administrators can receive alerts in real time, reducing the anomaly response time from the traditional hours to minutes, significantly lowering compliance risks and business losses.

[0083] Furthermore, the above solution addresses the issues of chaotic task trajectories and difficulty in locating anomalies in existing concurrent inspections: the Flow Run ID mechanism allows for quick association and querying of the complete execution logs of a single task, including information such as the time consumed at each stage and the anomaly trigger point, effectively improving the efficiency of single-task fault diagnosis and providing accurate performance data support for system optimization.

[0084] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or units may be electrical, mechanical, or other forms.

[0085] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0086] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0087] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0088] The above description is only a part of the embodiments of the present invention and does not limit the scope of protection of the present invention. Any equivalent device or equivalent process transformation made based on the content of the present invention specification and drawings, or direct or indirect application in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A site inspection method based on collaborative scheduling of artificial intelligence agents, characterized in that, include: S1: The visualization layer receives at least one target Uniform Resource Locator (URL) and the corresponding inspection task type input by the user. The inspection task type includes one or more of compliance testing, timeliness testing, and availability testing. S2: The visualization layer triggers the scheduling engine through the expressive state transition REST interface. The scheduling engine assigns an independent task to each target Uniform Resource Locator (URL) and manages the overall inspection progress through the flow mechanism. S3: The headless crawling module responds to the scheduling engine's instructions, uses an automated browser to access the target Uniform Resource Locator URL, performs page scrolling and waits for Document Object Model (DOM) rendering operations as needed, and outputs the target page's lightweight Markdown text and optional screenshots. S4: The data preparation module truncates the lightweight markup language Markdown text to a set character length, compresses the screenshot and converts it to base64 format, and generates a structured message containing execution steps and prompts. S5: The language analysis module calls a large-scale language model to process structured messages and outputs a structured inspection report containing status, site name, latest date, sensitive words, and summary. S6: The scheduling engine writes the event information of each stage of the inspection into the database and updates the memory cache synchronously. It also pushes events to the visualization layer through the server-side event sending (SSE) mechanism. S7: The columnar database stores structured inspection reports and event records for each stage. The visualization layer displays the inspection progress timeline in real time. After the inspection is completed, the structured inspection report is retrieved and displayed in card format.

2. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, After step S3 is completed, the following steps are also included: calculating the content hash value of the target page to form a cache identifier of Uniform Resource Locator URL-inspection task type-content hash, and querying whether there is a matching cache identifier in the cache. If a matching identifier exists, the historical structured inspection report is retrieved directly from the cache, skipping steps S4-S5; if no matching identifier exists, steps S4-S5 are executed, and the current cache identifier is associated with the structured inspection report and stored in the cache.

3. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, In step S6, the columnar database includes an inspection record table, an inspection configuration table, an inspection task table, a task execution association table, a process flow mechanism process operation event table, and a site baseline data table. Among them, the site baseline data table records the first crawl time of each target's Uniform Resource Locator (URL). In step S7, if the latest date field of the structured inspection report is unknown, the visualization layer retrieves the first crawl time from the site baseline data table and displays a prompt message indicating that no obvious updates have been identified since YYYY-MM-DD HH:MM.

4. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, It also includes the following steps for configuring scheduled inspections: The user inputs the target Uniform Resource Locator (URL), inspection task type, and scheduled task Cron expression through the task settings page in the visualization layer. The scheduling engine generates a corresponding Deployment based on the input information and establishes a relationship between the URL, inspection task type, and Deployment through the Deployment_id field in the inspection_tasks table. The scheduling engine triggers the scheduled inspection according to the scheduled task Cron expression, executing steps S3-S7. The inspection dashboard in the visualization layer aggregates and displays the task success rate, recent running status, and historical difference data for the scheduled inspections.

5. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, In step S6, the event push logic of the server sending event SSE mechanism is as follows: The degree engine monitors the status changes of each stage of the inspection in real time, and generates various events such as Flow.start (start of process flow mechanism), inspect_start (start of inspection), crawl_start (start of data collection), crawl_complete (complete of data collection), analysis_start (start of analysis), analysis_complete (complete of analysis), inspect_complete (complete of inspection), and Flow.complete (complete of process flow mechanism). These events are pushed to the visualization layer through the server sending event SSE interface. After pushing the inspect_complete or Flow.complete event, the event retry push is automatically stopped.

6. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, It also includes a step for comparing inspection results: receiving a user-initiated comparison request and specifying inspection records at two different time points; retrieving the corresponding two structured inspection reports from the inspection record table, comparing the site name, latest date, sensitive words, and summary fields one by one, generating a list of differences and displaying it in the visualization layer.

7. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, In step S4, the aggregation display logic of the inspection dashboard also includes: real-time monitoring of the status field in the structured report of the scheduled inspection; if the status field is critical, the preset inter-system real-time communication webhook interface is triggered to push alarm information to the administrator. The alarm information includes the Uniform Resource Locator URL, inspection time, anomaly type, and difference list.

8. The site inspection method based on collaborative scheduling of artificial intelligence agents as described in claim 1, characterized in that, In step S2, the scheduling engine's flow mechanism also includes: assigning a unique flow run ID to each independent task, recording the task's execution order, dependencies, and resource usage; and displaying the event information, time consumption data, and exception prompts of each stage of the corresponding task through the flow run ID.

9. A site inspection device based on collaborative scheduling of artificial intelligence agents, characterized in that, include: The system includes a visualization layer, a headless crawling module, a data preparation module, a language analysis module, a scheduling engine, and a columnar database. This visualization layer is used to receive at least one target Uniform Resource Locator (URL) and the corresponding inspection task type input by the user. The inspection task type includes one or more of compliance detection, timeliness detection, and availability detection. The scheduling engine is triggered through the Representational State Transition (REST) ​​interface. The scheduling engine assigns an independent task to each target Uniform Resource Locator URL and manages the overall inspection progress through the Flow mechanism. This headless crawling module is used to respond to scheduling engine instructions, automatically access the target Uniform Resource Locator URL using a browser, perform page scrolling and wait for Document Object Model (DOM) rendering operations as needed, and output the target page's lightweight Markdown text and optional screenshots. This data preparation module is used to truncate lightweight markup language Markdown text to a set character length, compress screenshots and convert them to base format, and generate structured messages containing execution steps and prompts. This language analysis module is used to call a large-scale language model to process structured messages and output a structured inspection report containing status, site name, latest date, sensitive words, and summary. This scheduling engine is used to write event information of each stage of inspection into the database and update the memory cache synchronously. It also pushes events to the visualization layer through the server-sent event SSE mechanism. This columnar database is used to store structured inspection reports and event records at each stage. The visualization layer displays the inspection progress timeline in real time. After the inspection is completed, the structured inspection report is retrieved and displayed in card format.

10. A computer device, characterized in that, include: At least one processor; And a memory communicatively connected to at least one processor; wherein the memory stores instructions executable by at least one processor, the instructions being executed by at least one processor to enable at least one processor to perform the site inspection method based on collaborative scheduling of artificial intelligence agents as described in any one of claims 1 to 8.

Images