A task scheduling method, a storage medium and a device

Abstract

Embodiments of the present application provide a task scheduling method, a storage medium and equipment, belonging to the technical field of generation engine optimization. The method comprises: receiving a task instruction of a user, and determining a tenant identifier and a brand corresponding to the user; generating a task number and a time window based on the task instruction, and reading a version configuration; collecting original samples of the brand based on the time window and a sampling platform, performing data isolation verification on the original samples based on the tenant identifier and an isolation policy version, and obtaining an isolation verification result; if the isolation verification result is passed, calculating a task priority; scheduling a computing task corresponding to the task instruction based on the task priority; executing the computing task based on a scheduling result to obtain a task execution result; and generating a traceable task log based on the task number, the time window, the original sample, the isolation verification result, the task priority and the task execution result. The method is used to realize data isolation and permission management between various tenants.

Description

Technical Field

[0001] This invention relates to the field of generation engine optimization technology, and more specifically to a task scheduling method, storage medium, and device. Background Technology

[0002] If the generation engine optimizes the GEO service and is delivered on a project-by-project basis, it is difficult to establish unified metrics, templates, monitoring tasks, and subscription operation capabilities. When multiple customers share the platform, issues such as tenant data isolation, permission boundaries, resource quotas, functional hierarchy, and metric calculation task scheduling must be addressed.

[0003] However, the existing SaaS architecture is not designed for GEO business objects and AI answer monitoring processes. In particular, in tenant management, it is difficult to guarantee data isolation between tenants and to better manage permissions for each tenant. Summary of the Invention

[0004] The purpose of this invention is to provide a task scheduling method, storage medium, and device for achieving data isolation and access control between tenants.

[0005] To achieve the above objectives, in a first aspect, embodiments of the present invention provide a task scheduling method for a GEO industry SaaS platform. The method includes: receiving a user's task instruction and determining the tenant identifier and brand corresponding to the user; generating a task number and a time window based on the task instruction, and reading a version configuration, wherein the version configuration includes the sampling platform and isolation policy version of the original sample; collecting original samples of the brand based on the time window and the sampling platform; performing data isolation verification on the original samples based on the tenant identifier and the isolation policy version to obtain an isolation verification result; calculating a task priority if the isolation verification result is successful; scheduling the computation task corresponding to the task instruction based on the task priority; executing the computation task based on the scheduling result to obtain a task execution result; and generating a traceable task log based on the task number, time window, original sample, isolation verification result, task priority, and task execution result.

[0006] Optionally, the calculation task includes the calculation of indicator values ​​across multiple dimensions, the positioning of the brand's GEO (Global Health Index) and the weakness dimension, wherein the multiple dimensions include a combination of two or more of the following: mention rate, first recommendation rate, sentiment index, factual consistency, structured quality, brand trust index, and competitive pressure.

[0007] Optionally, based on the tenant identifier and the isolation policy version, data isolation verification is performed on the original sample to obtain the isolation verification result, including: determining the pre-configured independent data namespace and resource scope in the isolation policy version based on the tenant identifier; verifying the user, tenant identifier, and original sample to determine whether the user, tenant identifier, and original sample meet the permission requirements, and determining whether the original sample belongs to the independent data namespace and resource scope; and if the verification result is that it meets the permission requirements and belongs to the independent data namespace and resource scope, then the verification is deemed successful.

[0008] Optionally, the version configuration also includes a subscription level. If the isolation verification result is passed, the method further includes: determining the subscription level corresponding to the tenant identifier based on the tenant identifier; determining the package quota corresponding to the subscription level based on the subscription level; obtaining the current usage of the quota by the tenant identifier; calculating the remaining quota based on the package quota and the current usage; and determining whether the remaining quota meets the quota requirement of the task calculation task corresponding to the task instruction. If it does not meet the requirement, an over-quota limit is triggered, the task priority calculation is stopped, and an over-quota warning is sent to the user.

[0009] Optionally, the formula for calculating the task priority includes: TaskPriority = PlanWeight × Urgency × DataFreshness / Cost, Among them, TaskPriority is the task priority, PlanWeight is the subscription weight, Urgency is the urgency level, DataFreshness is the data freshness, and Cost is the task cost.

[0010] Optionally, the task cost is determined by matching the task cost from historical task time data based on the task type, sampling volume, number of target platforms, and tenant level corresponding to the computation task; or by inputting the task type, sampling volume, number of target platforms, and tenant level corresponding to the computation task into a pre-trained regression model to obtain the task cost, wherein the regression model is trained based on historical task time data.

[0011] Optionally, the method further includes: when the version configuration corresponding to the tenant identifier changes, generating a candidate version based on the changed version configuration; extracting a preset number of standardized test questions from historical tasks, and obtaining the return results of the sampling platform based on the extracted standardized test questions to form a sample replay set; executing computational tasks on the old version and the candidate version respectively based on the sample replay set; comparing the task execution results corresponding to the old version and the candidate version to obtain the differences between the old and new versions; and if the differences between the old and new versions do not exceed a preset tolerance range, then releasing the candidate version.

[0012] Optionally, the method further includes exception handling, which includes: during method execution, if a preset exception occurs, triggering corresponding exception code, wherein the preset exception includes one or more of the following: insufficient sample quantity, format error, inaccessible source, model refusal to answer, fact conflict, indicator mutation, threshold out-of-bounds, and permission mismatch, wherein insufficient sample quantity includes empty samples and duplicate samples, and format error includes missing fields, illegal field format, and unparsable structured code; writing the exception code, the original sample, the processing module that triggered the exception code, the trigger time, and the version configuration together into a record to be reviewed; and performing a review based on the record to be reviewed.

[0013] Optionally, the review based on the record to be reviewed includes: if the anomaly code of the record to be reviewed corresponds to one or more of the following anomalies: insufficient sample size, incorrect format, inaccessible source, and mismatched permissions, then automatic rule review is performed to obtain a review conclusion; if the anomaly code of the record to be reviewed corresponds to one or more of the following anomalies: model refusal to answer, factual conflict, indicator mutation, and threshold exceeding limits, then manual review is performed to obtain a review conclusion; if the review conclusion is that it is correct, then the review conclusion and the review version are saved, the review version including the current version configuration; and If the review conclusion indicates a problem, the corresponding original sample will be saved as abnormal evidence and will not be used to perform the computation task.

[0014] Optionally, the method further includes: repeatedly executing the computation task based on a preset period, reusing the version configuration of the previous execution each time, updating the time window and the original sample; comparing the task execution result obtained each time with the task execution result obtained the previous time to obtain a comparison result; and determining the source of difference between the two executions based on the comparison result and recording the attribution, wherein the source of difference includes at least one of subscription change, quota change, task cost change and task attribute change.

[0015] Secondly, embodiments of the present invention provide a machine-readable storage medium storing instructions that cause a machine to perform the method described thereon.

[0016] Thirdly, embodiments of the present invention provide a device for task scheduling, the device including a memory and a processor, the processor being used to run a program, wherein the program is executed to perform the method described thereon.

[0017] By employing the aforementioned technical solution, this invention addresses the lack of a dedicated architecture for multi-tenant isolation, subscription, and task scheduling on the GEO platform. It constructs a continuous processing chain from data collection, isolation verification, priority calculation, task scheduling, task execution, to result output. During scheduling, the user's corresponding tenant identifier and brand are first determined, thereby protecting tenant permission boundaries in the SaaS platform's multi-tenant subscription scheduling scheme and achieving general resource isolation and permission management. Furthermore, through the continuous processing chain and traceable task logs, this invention avoids merely focusing on abstract operational goals or display interfaces, providing a definite data entry point for subsequent repairs, recalculations, and audits.

[0018] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a task scheduling method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the subscription scheduling process provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the data isolation process provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the GEO industry SaaS platform structure provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the device provided in an embodiment of the present invention.

[0020] Explanation of reference numerals in the attached figures 101. Processor 102. Memory 103. Bus 10. Equipment Detailed Implementation The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0021] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0022] Figure 1 This is a flowchart illustrating a task scheduling method provided in an embodiment of the present invention. This method is used in a GEO industry SaaS platform, such as... Figure 1 As shown, the method includes steps S1 to S8.

[0023] Step S1: Receive the user's task instruction and determine the tenant identifier and brand corresponding to the user. After receiving the task instruction, the tenant identifier is determined first to protect the tenant's permission boundaries in subsequent processing.

[0024] Step S2: Based on the task instruction, generate a task number and time window, and read the version configuration, wherein the version configuration includes the sampling platform and isolation strategy version of the original sample.

[0025] This application involves various preset configurations during task scheduling and computation, such as weights, thresholds, time windows, sample validity rules, and risk level rules. All configurations are saved as version configurations. Each time the system executes a task, it reads the currently enabled version configuration and records the corresponding version number, preventing unexplained differences in the same batch of samples under different rules.

[0026] Step S3: Based on the time window and the sampling platform, collect the original samples of the brand.

[0027] The original samples include the results returned by a pre-built standardized test question bank on the target platform. The original samples include answer text, web page content, comment information, structured code, source data, event logs, template configuration, tenant configuration, and task configuration.

[0028] Specifically, a standardized test question bank covering selection, popular science, comparison, and consultation scenarios is pre-built; based on this standardized test question bank, complete answer texts are collected in batches according to platform, model version, account environment, and time window.

[0029] In some embodiments of the present invention, the original sample is first converted into a structured intermediate object, and then the computational task is performed based on the structured intermediate object.

[0030] Step S4: Based on the tenant identifier and the isolation policy version, perform data isolation verification on the original sample to obtain the isolation verification result.

[0031] As an example, data isolation can be achieved through shared library field isolation, independent schemas, or independent databases.

[0032] Step S5: If the isolation verification result is passed, then calculate the task priority.

[0033] Step S6: Based on the task priority, schedule the computing task corresponding to the task instruction.

[0034] As an example, task scheduling can employ queue priority, token bucket, or cost budget algorithms.

[0035] Step S7: Based on the scheduling result, execute the computation task to obtain the task execution result.

[0036] Step S8: Generate a traceable task log based on the task number, time window, original sample, isolation verification result, task priority, and task execution result.

[0037] In some embodiments of the present invention, after obtaining the task execution result, the brand GEO health status is determined based on the task execution result. The brand GEO health status includes a normal status, an abnormal status, a status pending review, or a status pending remediation. The brand GEO health status is also written to the task log when it is generated.

[0038] The method provided by this invention supports GEO SaaS platform systems with multiple brands, multiple agents, and multiple industry templates. Addressing the lack of a dedicated architecture for multi-tenant isolation, subscription, and task scheduling on GEO platforms, it constructs a continuous processing chain from data collection, isolation verification, priority calculation, task scheduling, task execution, to result output. During scheduling, the user's corresponding tenant identifier and brand are first determined, thereby achieving general resource isolation and access control within the multi-tenant subscription scheduling scheme of the SaaS platform.

[0039] Existing technologies typically only output page displays, manual judgment conclusions, or isolated statistical results, lacking unified sampling standards, structured intermediate objects, recalculated indicators, and anomaly evidence chains. This makes it difficult to verify and compare results across different platforms, time windows, and brands, and also hinders the transformation of results into subsequent governance tasks. This invention, through continuous processing links and traceable task logs, avoids merely remaining at the level of abstract operational goals or display interfaces, providing a definite data entry point for subsequent remediation, recalculation, and auditing.

[0040] The processing flow of this invention can be solidified into executable data objects, module interfaces, calculation rules, and review records. This allows users to reproduce processing results based on the same input and preset configuration, enabling each result to be traced back to its corresponding sample. This transforms related GEO processing from manual experience-based judgment into a process flow that is collectable, calculable, and verifiable. For example, a dedicated GEO multi-tenant data model can be established using tenants, brands, projects, tasks, and reports as core objects, forming a dedicated structure for GEO task scheduling and data isolation. The modules of the model do not transmit difficult-to-understand internal field names, but instead transmit clearly defined data objects in Chinese, such as brand identity profiles, business knowledge units, standard terminology sets, risk event records, stage result records, and governance task records.

[0041] Furthermore, the calculation task includes the calculation of indicator values ​​across multiple dimensions, the positioning of the brand's GEO overall health index (GHI) and the weakness dimension. The multiple dimensions include a combination of two or more of the following: mention rate, first recommendation rate, sentiment index, factual consistency, structured quality, brand trust index, and competitive pressure.

[0042] It should be noted that the computational tasks include various tasks already present in GEO processing, such as governance generation tasks. These tasks can be identified, normalized, calculated, judged, and output using existing dedicated modules. This invention does not limit the specific processing methods for the computational tasks. The task execution results are only used for output and generating traceable task logs, and are not used to directly modify the original samples. If subsequent reviews, objections, or model updates occur, the computation process at that time can be restored based on the task logs.

[0043] In some embodiments of the present invention, when performing computational tasks, the original samples are first processed through field cleaning, entity normalization, sample validity assessment, and anomaly marking, and then corresponding calculations are performed based on the processed data. For example, the system first writes the input materials into the original sample library and binds a task number, brand number, source number, time window, and version configuration to each batch of samples, so that each subsequent output can be traced back to the original sample along the task number. Subsequently, the multi-tenant management module completes the first round of structured processing, and the processing results are then passed to subsequent modules in module order until a verifiable result record is formed. During execution, each intermediate result saves the original sample, generation time, and version configuration.

[0044] Furthermore, based on the tenant identifier and the isolation policy version, data isolation verification is performed on the original sample to obtain the isolation verification result, including: determining the pre-configured independent data namespace and resource scope in the isolation policy version based on the tenant identifier; verifying the user, tenant identifier, and original sample to determine whether the user, tenant identifier, and original sample meet the permission requirements, and determining whether the original sample belongs to the independent data namespace and resource scope; and if the verification result is that it meets the permission requirements and belongs to the independent data namespace and resource scope, then the verification is deemed successful.

[0045] In some embodiments of the present invention, data isolation verification can be performed based on the following formula: TenantIsolationCheck = Policy(user, tenant, resource), Among them, TenantIsolationCheck is the verification result, including verification pass and verification fail; Policy is the preset rule of the isolation policy version, such as determining whether the user, tenant identifier and original sample meet the permission requirements, and determining whether the original sample belongs to the independent data namespace and resource scope; user is user; tenant is tenant identifier; resource is original sample.

[0046] It should be noted that all formulas in this invention are stored in a formula version database, which is a preset version configuration. Calculations can be performed by reading the formulas from this version configuration, thus satisfying subsequent recalculation requirements.

[0047] Furthermore, the version configuration also includes subscription levels. If the isolation verification result is passed, the method further includes: determining the subscription level corresponding to the tenant identifier based on the tenant identifier; determining the package quota corresponding to the subscription level based on the subscription level; obtaining the current usage of the quota by the tenant identifier; calculating the remaining quota based on the package quota and the current usage; and determining whether the remaining quota meets the quota requirements of the task calculation task corresponding to the task instruction. If it does not meet the requirements, an over-quota limit is triggered, the task priority calculation is stopped, and an over-quota warning is sent to the user.

[0048] The remaining quota, QuotaRemain, can be calculated using the following formula: QuotaRemain = Package quota - Current usage.

[0049] In some embodiments of the present invention, the package quota includes at least the number of AI platforms corresponding to the subscription level, query limit, monitoring frequency, report type, and API limit. In other embodiments of the present invention, it may also include other types of quota. The current usage is the used quota within the package.

[0050] Figure 2 This is a schematic diagram of the subscription scheduling process provided in an embodiment of the present invention, such as... Figure 2 As shown, after calculating the remaining quota, calculating the task priority, and allocating resources to schedule the computing tasks, the usage is recorded based on the allocated resources, thereby updating the current usage for calculating the remaining quota for the next time.

[0051] Furthermore, the formula for calculating the task priority includes: TaskPriority = PlanWeight × Urgency × DataFreshness / Cost, Among them, TaskPriority is the task priority; PlanWeight is the subscription weight, which can be obtained based on the subscription benefits corresponding to the subscription level; Urgency is the urgency level; DataFreshness is the data freshness, both of which can be obtained based on the user's task instructions or preset rules; and Cost is the task cost.

[0052] Furthermore, the task cost is determined in the following ways: based on the task type, sampling volume, number of target platforms, and tenant level corresponding to the computation task, the task cost is obtained by matching from historical task time data; or the task type, sampling volume, number of target platforms, and tenant level corresponding to the computation task are input into a pre-trained regression model to obtain the task cost, wherein the regression model is trained based on historical task time data.

[0053] Furthermore, the method also includes: when the version configuration corresponding to the tenant identifier changes, generating a candidate version based on the changed version configuration; extracting a preset number of standardized test questions from historical tasks, and obtaining the return results of the sampling platform based on the extracted standardized test questions to form a sample replay set; executing computational tasks on the old version and the candidate version respectively based on the sample replay set; comparing the task execution results corresponding to the old version and the candidate version to obtain the differences between the old and new versions; and if the differences between the old and new versions do not exceed a preset tolerance range, then releasing the candidate version.

[0054] The samples in the sample replay set store sample_hash and replay_task_id. In some embodiments of the present invention, the system also stores version_effective_time, old_version, new_version, rollback_pointer, and replay_result_hash.

[0055] In some embodiments of the present invention, when other models, templates, sampling platforms or data sources change, except for version configuration, candidate versions are also generated and compared with the old version.

[0056] Through the above steps, when changes occur, a small-sample replay verification is first performed to compare the differences between the old and new versions in key metrics, which are calculated by the computational task. If the difference exceeds the preset tolerance range, a version change review task is generated. After the review is completed, the official version is switched. It should be noted that various updates do not directly overwrite historical results, but should generate new versions; historical task logs continue to be bound to the old version, and new task logs are bound to the new version.

[0057] Furthermore, the method also includes exception handling, which includes: during the execution of the method, if a preset exception occurs, triggering the corresponding exception code. The preset exception includes one or more of the following: insufficient sample quantity, format error, inaccessible source, model refusal to answer, fact conflict, indicator mutation, threshold out-of-bounds, and permission mismatch. Insufficient sample quantity includes empty samples and duplicate samples. Format error includes missing fields, illegal field format, and unparsable structured code. The exception code, along with the original sample, the processing module that triggered the exception code, the trigger time, and the version configuration, are written into a record to be reviewed. Review is then performed based on the record to be reviewed.

[0058] Furthermore, the review based on the record to be reviewed includes: if the anomaly code of the record to be reviewed corresponds to one or more of the following anomalies: insufficient sample quantity, incorrect format, inaccessible source, and mismatched permissions, then automatic rule review is performed to obtain a review conclusion; if the anomaly code of the record to be reviewed corresponds to one or more of the following anomalies: model refusal to answer, factual conflict, indicator mutation, and threshold exceeding the limit, then manual review is performed to obtain a review conclusion; if the review conclusion is that it is correct, then the review conclusion and the review version are saved, the review version including the current version configuration; and if the review conclusion is that there is a problem, then the corresponding original sample is saved as anomaly evidence and is not used to perform the calculation task.

[0059] Specifically, if the anomaly can be automatically determined by the program, automatic rule review is performed; if the anomaly requires professional judgment, manual review is performed. When an anomaly occurs, the system generates a record to be reviewed, without directly overwriting the final result. This mechanism ensures a traceable implementation chain from input, processing, calculation, review to output.

[0060] Figure 3 This is a schematic diagram of the data isolation process provided in an embodiment of the present invention, such as... Figure 3 As shown, in some embodiments of the present invention, if the data isolation verification fails, it is determined to be an abnormal situation and written into the review record or a repair task is generated. During the review, the review can be based on automatic rules to obtain a review conclusion.

[0061] In some embodiments of the present invention, the record to be reviewed and the review conclusion are also written to the task log. Furthermore, if the current execution includes a governance task, the governance task is generated after the review is completed.

[0062] Furthermore, the method also includes: repeatedly executing the computation task based on a preset period, reusing the version configuration of the previous execution each time, updating the time window and the original sample; comparing the task execution result obtained each time with the task execution result obtained the previous time to obtain a comparison result; and determining the source of difference between the two executions based on the comparison result and recording the attribution, wherein the source of difference includes at least one of subscription change, quota change, task cost change and task attribute change.

[0063] In actual deployment, retesting tasks are triggered daily, weekly, or by project nodes. The retesting tasks reuse the task number mapping relationship and version configuration of the previous cycle, only updating the collection time window and the newly added sample index; then the results of this cycle, the results of the previous cycle, and the source of differences are output, so that the system can understand how the solution maintains reproducibility and comparability in continuous operation scenarios.

[0064] Example 1: The method provided by this invention unifies the modeling of tenants, brands, projects, industry templates, monitoring tasks, indicator services, reporting services, and subscription rights, and adopts multi-tenant isolation, hierarchical permissions, resource quotas, and subscription policy control. This embodiment describes the overall task scheduling process of the SaaS platform, starting from the establishment of tenants and their corresponding core objects, specifically including the following steps S10 to S80.

[0065] Step S10: Create core objects such as tenant, brand, project, user, role, and subscription plan.

[0066] Brand A and Brand B belong to different tenants. By establishing the above core objects for different tenants and forcing the tenant number to be included in all task, metric, and report queries, cross-tenant access can be blocked through access policies. Agent tenants create sub-projects for multiple clients. The system allows agents to view summary reports, but clients can only view data for their own brand.

[0067] Step S20: Configure an independent data namespace, logical isolation field, and access permission policy for each tenant.

[0068] Step S30: Encapsulate the GEO monitoring task, indicator calculation, template initialization, and report generation into a platform service.

[0069] Step S40: Configure the number of available industry templates, the number of AI platforms, query limits, monitoring frequency, report types, and API limits according to the subscription level.

[0070] The basic subscription allows monitoring of only 100 queries and 3 AI platforms per day, while the professional subscription allows for higher frequency and more reports. The task scheduler automatically limits excess tasks based on the remaining quota, QuotaRemain.

[0071] Step S50: The task scheduler allocates resources based on tenant quotas, task priorities, and platform call costs.

[0072] Step S60: The metrics calculation service reads data in isolation by tenant and outputs brand-level, project-level, and industry-level reports.

[0073] Step S70: Record the metered usage, subscription status, over-limit limit, and function switch.

[0074] Step S80: Provide agents or ecosystem partners with subtenant management and white label reporting capabilities.

[0075] Figure 4 This is a schematic diagram of the GEO industry SaaS platform structure provided in an embodiment of the present invention. The task scheduling method provided by the present invention can be implemented based on this platform. Please refer to... Figure 4 The platform includes at least a multi-tenant management module, a template service module, a monitoring task module, a central scheduling module, an indicator calculation module, a report output module, and a subscription billing module. Each module is connected via task number, object number, source number, version number, and time window fields, and the output results are all associated with the original sample index, calculation rules, and processing status.

[0076] When executing task scheduling on this platform, the input data objects include task number, brand number, industry or scenario tag, target platform, collection time window, source type, query task number, and rule version; the intermediate processing objects generate structured records around the "GEOSaaS platform ecosystem architecture," including at least the original sample index, entity recognition results, normalized fields, anomaly markers, review status, and stage results; the indicator calculation objects include variables involved in the formula calculation, version configuration, sample validity markers, and calculation time; the output result objects include monitoring results, diagnostic results, early warning results, remediation suggestions, scheduling results, or report results, and are bound to the original samples, version configuration, and task logs.

[0077] This invention also provides a machine-readable storage medium storing instructions that cause a machine to perform the method described thereon.

[0078] This invention also provides a device for task scheduling, such as... Figure 5 As shown, the device includes a memory and a processor, the processor being used to run a program, wherein the program is run to perform the method described.

[0079] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and task scheduling is achieved by adjusting kernel parameters.

[0080] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0081] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0082] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0087] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0088] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0089] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0090] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A task scheduling method, the method being used in a GEO industry SaaS platform, characterized in that, The method includes: Receive the user's task instructions and determine the tenant identifier and brand corresponding to the user; Based on the task instructions, a task number and time window are generated, and the version configuration is read, wherein the version configuration includes the sampling platform and isolation strategy version of the original sample; Based on the time window and the sampling platform, original samples of the brand are collected. Based on the tenant identifier and the isolation policy version, data isolation verification is performed on the original sample to obtain the isolation verification result; If the isolation verification result is passed, then the task priority is calculated; Based on the task priority, the computing tasks corresponding to the task instructions are scheduled; Based on the scheduling results, the computational task is executed to obtain the task execution results; and Based on the task number, time window, original sample, isolation verification result, task priority, and task execution result, a traceable task log is generated.

2. The method according to claim 1, characterized in that, The computational tasks include calculating indicator values ​​across multiple dimensions, identifying the brand's GEO (Global Health Index) and pinpointing its weaknesses. The multiple dimensions include a combination of two or more of the following: mention rate, first recommendation rate, sentiment index, factual consistency, structured quality, brand trust index, and competitive pressure.

3. The method according to claim 1, characterized in that, Based on the tenant identifier and the isolation policy version, data isolation verification is performed on the original sample to obtain the isolation verification results, including: Based on the tenant identifier, determine the pre-configured independent data namespace and resource scope in the isolation policy version; The user, tenant identifier, and original sample are validated to determine whether they meet the permission requirements, and whether the original sample belongs to the independent data namespace and resource scope; and If the verification result meets the permission requirements and belongs to the independent data namespace and resource scope, then the verification is deemed successful.

4. The method according to claim 1, characterized in that, The version configuration also includes subscription levels. If the isolation verification result is passed, the method further includes: Based on the tenant identifier, determine the subscription level corresponding to the tenant identifier; Based on the subscription level, determine the corresponding package quota for the subscription level; Obtain the current quota usage for the tenant identifier; Calculate the remaining quota based on the package quota and the current usage; and Determine whether the remaining quota meets the quota requirements of the indicator calculation task corresponding to the task instruction. If not, trigger the over-quota limit, stop calculating the task priority, and send an over-quota warning to the user.

5. The method according to claim 1, characterized in that, The formula for calculating the task priority includes: TaskPriority = PlanWeight × Urgency × DataFreshness / Cost, Among them, TaskPriority is the task priority, PlanWeight is the subscription weight, Urgency is the urgency level, DataFreshness is the data freshness, and Cost is the task cost.

6. The method according to claim 5, characterized in that, The cost of the task is determined in the following way: Based on the task type, sampling volume, number of target platforms, and tenant level corresponding to the computing task, the task cost is obtained by matching historical task time data; or The task type, sampling volume, number of target platforms, and tenant level corresponding to the computation task are input into a pre-trained regression model to obtain the task cost, wherein the regression model is trained based on historical task time data.

7. The method according to claim 1, characterized in that, The method further includes: When the version configuration corresponding to the tenant identifier changes, a candidate version is generated based on the changed version configuration; A predetermined number of standardized test questions are extracted from historical tasks, and the return results from the sampling platform are obtained based on the extracted standardized test questions to form a sample playback set; Based on the sample replay set, computational tasks are performed on the old version and the candidate version, respectively; The task execution results corresponding to the old version and the candidate version are compared to obtain the differences between the old and new versions; and If the differences between the old and new versions do not exceed a preset tolerance range, then the candidate version is released.

8. The method according to claim 1, characterized in that, The method further includes exception handling, which includes: If a preset exception occurs during the execution of the method, the corresponding exception code will be triggered. The preset exception includes one or more of the following: insufficient sample quantity, format error, inaccessible source, model refusal to answer, fact conflict, indicator mutation, threshold out of bounds, and permission mismatch. Among them, insufficient sample quantity includes empty samples and duplicate samples, and format error includes missing fields, illegal field format, and unparsable structured code. The abnormal code, along with the original sample, the processing module that triggered the abnormal code, the trigger time, and the version configuration, are written into the review record; and The review will be conducted based on the records to be reviewed.

9. The method according to claim 8, characterized in that, The review based on the records to be reviewed includes: If the abnormality code of the record to be reviewed corresponds to one or more of the following abnormalities: insufficient sample quantity, incorrect format, inaccessible source, and mismatched permissions, then automatic rule review is performed to obtain the review conclusion. If the abnormal situation corresponding to the abnormal code of the record to be reviewed is one or more of the following: model refusal to answer, factual conflict, indicator mutation, and threshold exceeding the limit, then manual review is performed to obtain the review conclusion. If the review conclusion is that there are no errors, then save the review conclusion and the review version, whereby the review version includes the current version configuration; and If the review conclusion indicates a problem, the corresponding original sample will be saved as evidence of anomalies and will not be used to perform the computation task.

10. The method according to claim 1, characterized in that, The method further includes: The computation task is repeatedly executed based on a preset period. Each time it is executed, the version configuration of the previous execution is reused, and the time window and the original sample are updated. The task execution result obtained each time is compared with the previous task execution result to obtain the comparison result; and Based on the comparison results, the sources of difference between the two executions are determined and the attribution is recorded. The sources of difference include at least one of subscription changes, quota changes, task cost changes, and task attribute changes.

11. A machine-readable storage medium having instructions stored thereon for causing a machine to perform the method as described in any one of claims 1-10.

12. A device for task scheduling, the device comprising a memory and a processor, characterized in that, The processor is used to run a program, wherein the program is run to execute the method as described in any one of claims 1-10.

Images