Business processing method and device based on large model, agent and electronic device

Abstract

The present disclosure provides a large model-based business processing method and device, an agent and electronic equipment, and relates to the technical field of artificial intelligence such as natural language model, large model and deep learning. The specific implementation scheme is: generating a first prompt text based on the business information of the to-be-processed business; performing business analysis on the to-be-processed business through a large model according to the business knowledge base and the first prompt text to obtain a business analysis result of the to-be-processed business; generating a second prompt text based on the business analysis result and the code template; generating a target code conforming to the specification for the to-be-processed business through the large model according to the second prompt text; allocating resources for the code execution task of the to-be-processed business through the large model based on the business characteristics and the system state of the to-be-processed business; and performing intelligent quality monitoring on the data processed by the large model on the code execution task. The present disclosure can significantly reduce the development cost of new business access, improve resource utilization and improve data quality.

Description

Technical Field

[0001] This disclosure pertains to the field of data processing, particularly to the fields of artificial intelligence technologies such as natural language models, large models, and deep learning. Specifically, it relates to business processing methods based on large models (multi-business unified processing methods driven by large models), devices, intelligent agents, and electronic devices. These methods can be widely applied in fields such as big data processing, data platform construction, enterprise data warehouses, and business intelligent analysis, providing enterprises with intelligent multi-business data processing capabilities. Background Technology

[0002] As enterprise-level data processing platforms continue to expand their business scale and the types of business they need to handle increase, the traditional method of manually writing ETL (Extract, Transform, Load) scripts can no longer meet the needs of rapid business iteration. Existing technical solutions suffer from problems such as high business understanding costs, low code development efficiency, unintelligent resource allocation, and insufficient data quality assurance, which seriously restrict the efficiency and quality of enterprise data processing. Summary of the Invention

[0003] This disclosure provides a business processing method, apparatus, intelligent agent, and electronic device based on a large model, which can solve problems such as high business understanding cost, low code development efficiency, unintelligent resource allocation, and insufficient data quality assurance in the prior art.

[0004] According to a first aspect of the present disclosure, a business processing method based on a large model is provided, including: Generate the first prompt text based on the business information of the pending business; Based on the business knowledge base and the first prompt text, the business to be processed is analyzed using a large model to obtain the business analysis results of the business to be processed. A second prompt text is generated based on the business analysis results and code template; Based on the second prompt text, generate compliant target code for the pending business using the large model; Based on the business characteristics and system status of the pending business, target partitioning strategy information and resource configuration scheme information are generated for the code execution task of the pending business through the large model. Intelligent quality monitoring is performed on the data after the code execution task is processed based on the large model.

[0005] According to a second aspect of the present disclosure, a business processing apparatus based on a large model is provided, comprising: The first generation module is used to generate the first prompt text based on the business information of the business to be processed. The business analysis module is used to perform business analysis on the business to be processed based on the business knowledge base and the first prompt text, and obtain the business analysis results of the business to be processed. The second generation module is used to generate a second prompt text based on the business analysis results and the code template; The code generation module is used to generate compliant target code for the business to be processed based on the second prompt text and the large model. The resource configuration generation module is used to generate target partitioning strategy information and resource configuration scheme information for the code execution task of the pending business based on the business characteristics and system status of the pending business and through the large model. The quality monitoring module is used to perform intelligent quality monitoring on the data after the code execution task is processed based on the large model.

[0006] According to a third aspect of the present disclosure, an intelligent agent is provided, comprising: The input module is used to receive input information; The processing module is configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and execute the method described in the first aspect by calling the large model to obtain output information; An output module is used to output the output information obtained by the processing module.

[0007] According to a fourth aspect of the present disclosure, an electronic 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 method described in the first aspect above.

[0008] According to a fifth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0009] According to a sixth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0010] According to the technical solution disclosed herein, intelligent business understanding and code generation driven by a large model can significantly reduce the development cost of new business integration. It enables intelligent processing from business analysis to code generation. Based on the business characteristics and system status, the large model intelligently generates resource configuration information for the business's code execution tasks, significantly improving resource utilization. It also enables intelligent data quality monitoring, which, compared to traditional methods relying on manual monitoring and rule verification, can greatly shorten discovery time and significantly improve discovery efficiency.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 This is a flowchart illustrating the business processing method based on a large model provided in an embodiment of this disclosure; Figure 2 This is a flowchart illustrating the business processing method based on a large model provided in an embodiment of this disclosure; Figure 3 This is a flowchart illustrating the business processing method based on a large model provided in an embodiment of this disclosure; Figure 4 This is a flowchart illustrating the business processing method based on a large model provided in this embodiment of the disclosure; Figure 5 This is a block diagram of a large-model-based service processing apparatus provided in an embodiment of this disclosure; Figure 6 This is a block diagram of a large-model-based service processing apparatus provided in an embodiment of this disclosure; Figure 7 This is a block diagram of a large-model-based service processing apparatus provided in an embodiment of this disclosure; Figure 8 This is a block diagram of an electronic device used to implement the large-model-based business processing method of the embodiments of this disclosure; Figure 9 A block diagram of an intelligent agent provided in an embodiment of this disclosure. Detailed Implementation

[0013] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0014] The embodiments disclosed herein relate to the fields of artificial intelligence technologies such as natural language processing, large models, and deep learning.

[0015] Artificial Intelligence (AI) is a new technological science that studies, develops, and applies theories, methods, technologies, and application systems to simulate, extend, and expand human intelligence.

[0016] Natural Language Processing (NLP) is an important field within computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. It is a discipline that uses computer technology to analyze, understand, and process natural language, treating computers as a powerful tool for language research. With computer support, it conducts quantitative research on language information and provides language descriptions that can be used by both humans and computers.

[0017] Large Language Models (LLMs) are deep learning models trained on large amounts of text data that can generate natural language text or understand the meaning of language text. LLMs can handle various natural language tasks, such as text classification, question answering, and dialogue, and are an important pathway to artificial intelligence.

[0018] Deep learning learns the inherent patterns and hierarchical representations of sample data. The information gained during this learning process greatly aids in interpreting data such as text, images, and sound. The ultimate goal of deep learning is to enable machines to possess analytical and learning capabilities similar to humans, allowing them to recognize data such as text, images, and sound.

[0019] An intelligent agent is a proxy capable of perceiving its environment and taking actions to achieve specific goals. It can be software, hardware, or a system, possessing autonomy, adaptability, and interactivity. By perceiving changes in the environment (e.g., through sensors or data input), an intelligent agent makes judgments and decisions based on its learned knowledge and algorithms, and then executes actions to influence the environment or achieve predetermined goals.

[0020] It should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0021] It should also be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0022] It should be noted that in the embodiments disclosed herein, certain software, components, models, and other existing solutions in the industry may be mentioned. These should be considered as exemplary and are intended only to illustrate the feasibility of implementing the technical solutions disclosed herein. However, they do not mean that the applicant has used or necessarily used such solutions.

[0023] In related technologies, the following three solutions are commonly used to address the unified processing of multiple business functions in enterprise-level data processing platforms: (1) Each business writes a dedicated ETL (Extract, Transform, Load) processing script. R&D and testing personnel share a set of processing logic, maintain multiple business branches in the code repository, write dedicated transformation logic for different business scenarios, and execute each business task uniformly through the scheduling system.

[0024] (2) Apply for multiple data processing resources. R&D and testing personnel go to the big data processing platform to configure their respective resource parameters, allocate independent computing and storage resources for different businesses, and maintain their own business processing code and data table structure.

[0025] (3) The data processing logic is encapsulated in a template, removing the dependence on specific business and changing it into a general data processing framework. Business rules and data mapping relationships are defined through configuration files to reduce the development work of repetitive code.

[0026] However, the three solutions mentioned above have different application scenarios, and therefore different problems and shortcomings: Solution (1) requires writing a dedicated ETL processing script for each business. Different business logics will cause development dependencies and development blockages. New business access requires in-depth understanding of business rules and reliance on the experience of business experts. Business rules are complex and changeable, and human understanding is prone to errors, resulting in a long new business access cycle. From requirements analysis to code deployment, it often takes a long time.

[0027] Although multiple sets of data processing resources can solve the business logic coupling problem of solution (1) in solution (2), it will inevitably lead to resource waste, and unreasonable resource configuration will easily cause performance problems. Some business resources are wasted, some business resources are insufficient, there is a lack of predictive optimization, and it is impossible to plan resource requirements in advance. Performance problems rely on manual investigation, which is inefficient.

[0028] Solution (3) can effectively solve the problems of solutions (1) and (2) above, but it is highly templated. Every time there is a business change, the template needs to be modified a lot, which increases the amount of ineffective code development and wastes manpower and financial resources. In addition, the general template cannot reflect the business characteristics, cannot effectively cover complex business scenarios, lacks a deep understanding of specific business logic, has limited business logic reasoning ability, and is difficult to handle complex scenarios.

[0029] Based on this, this disclosure proposes a business processing method based on a large model. Through intelligent business understanding and code generation driven by the large model, the development cost of new business integration can be significantly reduced. It enables intelligent processing throughout the entire process from business analysis to code generation. Based on the business characteristics and system status, the large model intelligently generates resource configuration information for the business's code execution tasks, significantly improving resource utilization. It also enables intelligent data quality monitoring, which, compared to traditional methods relying on manual monitoring and rule verification, can greatly shorten discovery time and significantly improve discovery efficiency.

[0030] It should be noted that the execution subject of the business processing method based on the large model in this embodiment can be a business processing device based on the large model. The device can be implemented by software and / or hardware. The device can be configured in an electronic device, which may include, but is not limited to, a terminal, a server, etc.

[0031] It is worth noting that the business processing method based on the large model in this embodiment can be implemented by an autonomous intelligent agent based on the large model. This autonomous intelligent agent based on the large model can achieve unified processing of multiple services by driving the large model, and can intelligently understand business logic, automatically generate processing code, intelligently optimize resource allocation, and comprehensively ensure data quality.

[0032] The following description, with reference to the accompanying drawings, outlines a business processing method, apparatus, intelligent agent, and electronic device based on a large model, representing embodiments of the present disclosure.

[0033] Figure 1 This is a flowchart illustrating the business processing method based on a large model provided in this embodiment of the disclosure. Figure 1 As shown, this business processing method based on a large model may include, but is not limited to, the following steps.

[0034] In step 101, a first prompt text is generated based on the business information of the business to be processed.

[0035] Optionally, business information of the pending business can be obtained, and a first prompt text can be constructed based on this business information and a preset first prompt template. In one possible implementation, the business information of the pending business can be filled into a preset first prompt template to obtain the first prompt text; wherein, the first prompt template may include, but is not limited to, business analysis dimensions, output format requirements, and large model call parameters, etc., and the output format requirements can be used to require that the business analysis results include the fields corresponding to the analysis results of each dimension.

[0036] In some embodiments, the service to be processed may be a new service to be accessed. The service information of the service to be processed may include, but is not limited to: basic service information of the service to be processed (e.g., including but not limited to service name, service description, service type, etc.), data samples, service rule information, data schema (pattern / graph) information, etc. Optionally, the basic service information, data samples, service rule information, data schema information, etc. of the service to be processed can be filled into a preset first prompt template to obtain the first prompt text.

[0037] In the embodiments disclosed herein, the first prompt text may include, but is not limited to, business analysis dimensions, output format requirements, and large model call parameters. The first prompt text may also include role settings (such as Spark Scala development experts and business analysts), task objectives (deep business understanding and analysis), and input information (such as basic business information, data samples, business rule information, and data schema information). This first prompt text can be structured text, which can be used to guide the large model in performing deep business understanding and analysis of the business to be processed.

[0038] In embodiments of this disclosure, the aforementioned business analysis dimensions may include, but are not limited to, one or more of the following: business pattern recognition; data feature analysis; business logic understanding; unified model mapping; performance and resource evaluation; and risk and challenge analysis. For example, the first prompt text may include analysis requirements, requesting the large model to perform business analysis from the following dimensions: Business model identification: Identifying business types, transaction patterns, revenue models, etc. Data feature analysis: analyzing data volume, data structure, data quality, etc. Business logic understanding: Understanding business rules, business processes, business constraints, etc.; Unified model mapping: Fields and rules mapped to a unified data model; Performance and resource assessment: Evaluate data volume, processing complexity, resource requirements, etc. Risk and Challenge Analysis: Identify potential risks and technical challenges.

[0039] Correspondingly, the first prompt text and template also include fields required for the business analysis results to include the corresponding analysis results for each dimension. For example, taking business pattern recognition as a business analysis dimension, the output analysis results should include fields such as business type, transaction model, and revenue model. Similarly, taking data feature analysis as a business analysis dimension, the output analysis results should include fields such as data volume, data structure complexity, and data quality assessment. Taking business logic understanding as a business analysis dimension, the output analysis results should include fields such as key business rules, business process descriptions, and business constraints. Taking unified model mapping as a business analysis dimension, the output analysis results should include fields such as source fields, target fields, and transformation rules. Taking performance and resource evaluation as a business analysis dimension, the output analysis results should include fields such as estimated data volume, processing complexity, and resource requirements. Taking risk and challenge analysis as a business analysis dimension, the output analysis results should include fields such as potential risks, technical challenges, and mitigation measures.

[0040] In the embodiments of this disclosure, the aforementioned large model calling parameters may include, but are not limited to, one or more of the following: temperature parameter, maximum number of output tokens or maximum output length (max_tokens), kernel sampling (top_p), frequency penalty (frequency_penalty), and presence penalty (presence_penalty). For example, the temperature parameter can be set to 0.3 to reduce randomness and improve output stability; max_tokens can be set to 2000 to ensure sufficient output detail; top_p can be set to 0.9 to control output diversity; frequency_penalty can be set to 0.0 to not reduce the probability of duplicate content; and presence_penalty can be set to 0.0 to discourage new topics.

[0041] In step 102, based on the business knowledge base and the first prompt text, the business to be processed is analyzed using a large model to obtain the business analysis results of the business to be processed.

[0042] It should be noted that, in some embodiments, the business knowledge base may include relevant knowledge information for multiple businesses, such as basic business information (e.g., business type, business name, business description, etc.), business characteristics (e.g., including but not limited to data field characteristics, business process characteristics, revenue model characteristics, etc.), business analysis results, business templates, etc. For example, data field characteristics may include, but are not limited to, one or more of the following: field quantity statistics, field type distribution, field complexity, field naming characteristics, field constraint characteristics, field association characteristics, field quality characteristics, field extension characteristics, etc. Business process characteristics may include, but are not limited to, one or more of the following: transaction type characteristics, settlement cycle characteristics, refund processing characteristics, revenue model characteristics, data flow characteristics, timeliness characteristics, data volume characteristics, concurrency characteristics, dependency relationship characteristics, error handling characteristics, monitoring characteristics, data quality characteristics, etc. Revenue model characteristics may include, but are not limited to, one or more of the following: revenue source characteristics, profit-sharing ratio characteristics, payment method characteristics, settlement delay characteristics, settlement method characteristics, reconciliation characteristics, risk control characteristics, compliance characteristics, tax characteristics, audit characteristics, report characteristics, analysis characteristics, etc. For numerical features, normalization can be used to map the values ​​to the [0,1] interval; for categorical features, one-hot encoding can be used, with each category corresponding to one binary bit; for Boolean features, 0 / 1 encoding can be used directly; for text features, TFIDF (termfrequency–inverse document frequency) or word vector encoding can be used.

[0043] In the embodiments of this disclosure, a first prompt text can be input into a large model. Based on this first prompt text and a business knowledge base, the large model can perform in-depth business understanding and analysis of the business to be processed. The large model can identify business patterns by analyzing the data characteristics and business descriptions of the business to be processed, collect data samples, business rule information, and data schema information of the business to be processed, extract data field features, business process features, and revenue model features, perform semantic matching with the business knowledge base, identify business similarities and differences, infer business type and classification, and thus output the business analysis results of the business to be processed.

[0044] Optionally, to improve processing efficiency, existing business templates can be directly reused for businesses with high similarity. For example, a cosine similarity algorithm can be used to calculate business similarity. As an example, the similarity can be calculated using the following formula:

[0045] Where: A is the multidimensional feature vector of the business to be processed, B is the multidimensional feature vector of the existing business in the business knowledge base, and A·B is the vector dot product, calculated using the following formula:

[0046] Where i = 1 to 128 (128 is the feature dimension), ||A|| is the magnitude of vector A, and ||B|| is the magnitude of vector B.

[0047] The similarity threshold can be determined as follows: High similarity threshold 0.85: Based on historical data analysis, when the similarity is ≥0.85, the business to be processed can directly reuse the existing business template, with an accuracy rate of over 90%; Medium similarity threshold 0.70: When 0.70 ≤ similarity < 0.85, the business to be processed needs to make partial modifications to the existing template, with an accuracy rate of over 80%; Low similarity threshold 0.50: When 0.50 ≤ similarity < 0.70, the business to be processed needs to be redesigned based on a general template, with an accuracy rate of over 70%; No similarity threshold 0.50: When the similarity is < 0.50, a completely new business analysis is required.

[0048] Optionally, in some embodiments, after outputting the business analysis results, the results can be parsed and validated using a large model. For example, regular expressions can be used to extract JSON-formatted output; a JSON parsing library can be used to parse structured data; missing fields can be filled with default values; and format errors can be automatically corrected. Validation of the business analysis results can include: field integrity checks: ensuring all required fields exist; data type checks: ensuring field values ​​conform to expected types; business logic checks: ensuring consistency of business rules; and data format checks: ensuring the correctness of the data format. Anomalies in the business analysis results can be handled as follows: when parsing fails, prompt the user to re-enter the data; when validation fails, return an error message and request correction; and timeout handling: for example, setting a 30-second timeout and automatically retrying after the timeout.

[0049] In step 103, a second prompt text is generated based on the business analysis results and the code template.

[0050] Optionally, the second prompt text can be generated based on the business analysis results and the code template, combined with the second prompt template. In one possible implementation, a code template is selected from the code template library based on the first business characteristic of the business to be processed; wherein, the first business characteristic can be obtained by feature extraction based on the business information of the business to be processed; the business analysis results and the selected code template are filled into the preset second prompt template to obtain the second prompt text.

[0051] For example, feature extraction can be performed on the business information of the business to be processed to obtain the first business feature of the business to be processed (such as including but not limited to data field features, business process features, revenue model features, etc.). Based on the first business feature of the business to be processed, a suitable code template can be selected from the code template library. The business analysis results of the business to be processed and the selected code template are filled into the preset second prompt template to obtain the second prompt text.

[0052] It should be noted that the code template library may be pre-built and may include multiple standardized code templates. In some embodiments, the code template library may include, but is not limited to, unified data processing workflow templates (such as main function templates, data reading templates, data cleaning templates, data calculation templates, data association templates), commonly used data transformation templates, performance optimization templates, data quality templates, etc.

[0053] In some possible embodiments, business similarity can be calculated based on the first business feature of the business to be processed and the business features of one or more reference businesses in the code knowledge base to obtain one or more business similarities; based on the highest business similarity among the one or more business similarities, a code template is selected from the code template library; wherein, if the highest business similarity is greater than or equal to a first threshold, a dedicated code template for the business type of the corresponding reference business is selected; if the highest business similarity is less than the first threshold but greater than or equal to a second threshold, a general code template is selected; and if the highest business similarity is less than the second threshold, a basic code template is selected.

[0054] It should be noted that the code knowledge base may include basic information about reference services (such as service type, service name, etc.), service characteristics, and code template identifiers corresponding to the reference services. Optionally, the similarity between services can be calculated using the first service characteristic of the service to be processed and the service characteristics of each reference service in the code knowledge base, resulting in the service similarity between the service to be processed and each reference service. The highest service similarity is then identified. If the highest service similarity is greater than or equal to a first threshold, a dedicated code template for the corresponding service type can be selected from the code template library. If the highest service similarity is less than the first threshold but greater than or equal to a second threshold, a general code template can be selected from the code template library. If the highest service similarity is less than the second threshold, a basic code template can be selected from the code template library. The selected code template is then used as the code template corresponding to the service to be processed, and code generation can be performed based on the selected code template. In other words, the code template selection logic is as follows: high-similarity services select a dedicated template corresponding to the service type; medium-similarity services select a general template with some modifications; and low-similarity services select a basic template and redesign it.

[0055] In some embodiments, the second prompt text may include, but is not limited to, generation requirement information, which may include, but is not limited to, one or more of the following: code style requirements, performance optimization requirements, error handling requirements, logging requirements, and data quality requirements. Optionally, the second prompt text may also include role settings (such as Spark Scala development expert), task objectives (generating Spark Scala code that conforms to the specifications), input information (business analysis report, data schema, business rules), output format requirements (complete Scala object class structure), etc. This second prompt text can be used to guide large models to automatically generate compliant target code for the business to be processed based on the business analysis results.

[0056] In step 104, based on the second prompt text, target code conforming to specifications is generated from the large model for the business to be processed.

[0057] In the embodiments of this disclosure, the second prompt text can be input into the large model. The large model can generate complete processing code based on the business analysis results and code template, parse the business analysis results to extract key information (such as including but not limited to data source information, field mapping relationships, conversion logic, performance requirements, etc.), fill the template parameters of the selected code template to generate basic code, apply performance optimization rules, add error handling and logging, and perform code quality checks and optimizations.

[0058] In step 105, based on the business characteristics and system status of the business to be processed, target partitioning strategy information and resource configuration scheme information are generated for the code execution task of the business to be processed through a large model.

[0059] In the embodiments of this disclosure, historical data processing records of the pending business can be collected. A large model is used to perform feature analysis on these historical data processing records to obtain the second business characteristics of the pending business. Based on these second business characteristics, the number of partitions is predicted. Combined with business priority and resource constraint information, a constraint optimization algorithm is applied to generate target partitioning strategy information and resource allocation scheme information. This facilitates the allocation of resources for the code execution tasks of the pending business using the target partitioning strategy information and resource allocation scheme information.

[0060] After generating the target code for the service to be processed, and generating target partitioning policy information and resource configuration scheme information for the service, the target code, target partitioning policy information, and resource configuration scheme information can be sent to the execution engine. The execution engine can allocate resources for the code execution task of the service to be processed based on the target partitioning policy information and resource configuration scheme information, and execute the target code. The execution engine can then return the data processed for the code execution task.

[0061] In step 106, intelligent quality monitoring is performed on the data after the code execution task is processed based on the large model.

[0062] In the embodiments of this disclosure, a normal data distribution model is established by combining a large model with statistical methods and machine learning to intelligently detect data anomalies. Multiple anomaly detection methods are used to detect anomalies in the data after code execution tasks. Intelligent diagnosis is performed on detected anomalies to provide remediation suggestions. The anomaly context is analyzed, including data context, business context, and technical context, to identify the root cause of the anomaly and generate remediation suggestion information and / or remediation impact assessment information.

[0063] In the above embodiments, intelligent business understanding and code generation driven by a large model can significantly reduce the development cost of new business integration. It enables intelligent processing from business analysis to code generation. Based on the business characteristics and system status, the large model intelligently generates resource configuration information for the business's code execution tasks, significantly improving resource utilization. It also enables intelligent data quality monitoring, which, compared to traditional methods relying on manual monitoring and rule verification, can greatly shorten discovery time and significantly improve discovery efficiency.

[0064] Figure 2 This is a flowchart illustrating the business processing method based on a large model provided in this embodiment of the disclosure. Figure 2 As shown, the optional implementation methods for generating compliant target code for the business to be processed through the large model based on the second prompt text include, but are not limited to, the following steps.

[0065] In step 201, based on the business analysis results and code specification requirements in the second prompt text, and in conjunction with the code knowledge base, the basic code is generated by filling the template parameters of the code template with the large model.

[0066] In the embodiments of this disclosure, template parameters may include, but are not limited to, one or more of the following: business parameters (such as business name, business description, package name, object name, application name, etc.), data source parameters (such as Hive table name, Parquet file path, SQL query statement, etc.), data processing parameters (such as field list, read logic, transformation logic, write logic, partition field, number of partitions, etc.), and performance parameters (such as number of Shuffle partitions, default parallelism, optimal number of partitions, etc.). In step 202, the basic code is subjected to performance tuning based on the performance optimization requirements in the second prompt text.

[0067] In step 203, the basic code that has undergone performance tuning is checked and optimized according to the data quality requirements in the second prompt text.

[0068] In some embodiments, the quality check may include, but is not limited to, one or more of the following: syntax check, performance check, specification check, logic check, and integrity check.

[0069] Optionally, if the quality check fails, the error message can be fed back to the large model for correction, for example, up to 3 retries can be made to optimize the target code of the pending business.

[0070] In step 204, after the quality check is passed, error handling and / or logging information is added according to the error handling requirements and / or logging requirements in the second prompt text, and target code that conforms to the specifications is generated for the business to be processed.

[0071] In the above embodiments, business analysis results can be used as input data. After code template selection and prompt text construction, the data is input into a large model for code generation. The generated results undergo quality checks and corrections before outputting the final, compliant target code. Throughout the process, the code template library and knowledge base serve as auxiliary data sources, providing reference and constraints for code generation. Intelligent business understanding and code generation driven by a large model significantly reduces the development cost of new business integration. Compared to traditional manual development methods, it can greatly shorten the new business integration cycle and significantly reduce development costs. It achieves full-process intelligence from business analysis to code generation. Compared to traditional methods relying on manual configuration and rule engines, the intelligence level of key aspects such as business understanding, code generation, and decision optimization is significantly improved, and key indicators such as business understanding accuracy, code generation usability, and anomaly detection accuracy are all significantly improved.

[0072] To further improve resource utilization, optionally, in some possible embodiments, during the execution of code execution tasks, resource scheduling is performed on the code execution tasks based on resource scheduling strategy information; wherein, the resource scheduling strategy information includes one or more of the following: prioritizing resource allocation for high-priority services, prioritizing scheduling for tasks nearing their deadlines, dynamically adjusting based on current resource usage, and predicting resource demand based on historical execution times.

[0073] For example, tasks in the task queue can be sorted according to priority, deadline, and estimated time. The sorting formula could be: Score = w1 × Priority + w2 × Urgency + w3 × Resource_efficiency, where w1 = 0.5, w2 = 0.3, and w3 = 0.2. For each task, resource requirements (e.g., CPU and memory requirements) are estimated, and resource availability is checked. If resources are sufficient, they are allocated and scheduled; if resources are insufficient, the task is added to a waiting queue. Task execution is monitored for dynamic adjustments. For example, if the task execution time is greater than 1.5 times the estimated time, resource allocation is increased; if the task execution time is less than 0.5 times the estimated time, resource allocation is decreased. Cluster load can also be checked periodically. If the load is unbalanced, tasks are redistributed. For example, the load balancing target is: CPU utilization difference between nodes < 10%. Using the above intelligent resource scheduling strategy, resource scheduling of code execution tasks for processing business can be achieved. By adopting an intelligent resource scheduling strategy, compared with the traditional static resource allocation method, resource allocation can be dynamically adjusted according to business characteristics, which can further improve resource utilization, significantly shorten task completion time, and significantly reduce overall processing costs.

[0074] Figure 3 This is a flowchart illustrating the business processing method based on a large model provided in this embodiment of the disclosure. Figure 3 As shown, the optional implementation methods for allocating resources for the code execution task of the pending business based on the business characteristics and system status of the pending business through the large model include, but are not limited to, the following steps.

[0075] In step 301, feature analysis is performed on the historical data processing records of the business to be processed based on the large model to obtain the second business features of the business to be processed.

[0076] In the embodiments of this disclosure, historical data processing records of the business to be processed can be collected. For example, historical data processing records of the business to be processed can be obtained from a historical data knowledge base. These historical data processing records may include, but are not limited to, one or more of the following: data volume, processing time, number of partitions, performance indicators, etc. For example, the time span of the data records may be no less than 30 days. A large model can be used to perform business characteristic analysis on the historical data processing records, including data growth trends, access patterns, performance bottlenecks, etc., to obtain the second business characteristics of the business to be processed.

[0077] In step 302, the second business feature is input into the partition number prediction model to predict the number of partitions.

[0078] Optionally, the partition number prediction model can be pre-built. For example, it can be constructed using a combination of linear regression and machine learning, comprehensively considering data volume prediction, time prediction, and business characteristic analysis. As an example, the formula for the partition number prediction model is expressed as follows: P = α × log(D) + β × T + γ × C + δ Where P is the predicted number of partitions, D is the data volume, T is the historical processing time, C is the business complexity coefficient, and α, β, γ, and δ are model parameters obtained through training on historical data. For example, the value range of α can be [0.5, 2.0], with a default value of 1.2. α represents the degree of influence of data volume on the number of partitions. This parameter α can be obtained through regression analysis of historical data. When the prediction error > 10%, the value of α can be adjusted. The value range of β can be [0.1, 0.5], with a default value of 0.3. β represents the degree of influence of processing time on the number of partitions. This parameter β can be analyzed based on the relationship between historical processing time and the number of partitions. When the processing time is abnormal, the value of β can be adjusted. The value range of γ can be [0.5, 3.0], with a default value of 1.5. γ represents the degree of influence of business complexity on the number of partitions. The complexity score γ can be calculated based on the business feature vector. The formula for calculating complexity C is as follows: , i = 1 to N (N is the feature dimension, for example, N can be 128), wi is the weight, fi is the feature value. δ is the bias term, the value range of δ can be [10, 10], the default value can be 5, δ can represent the adjustment value of the baseline partition number, the parameter δ can be obtained by least squares fitting, and the parameter δ can be dynamically adjusted according to the system resource status.

[0079] For example, the training of this partition number prediction model can be performed using the following steps 1 to 5: Step 1: Data Collection Collect historical data processing records, with a time span of no less than 30 days; the record fields include: data volume D, processing time T, number of partitions P, business complexity C, and performance indicators.

[0080] Data quality requirements: Remove outlier data points to ensure data accuracy.

[0081] Step 2: Data Preprocessing Take the logarithm of the data size D: .

[0082] Data normalization: Normalize each feature to the [0,1] interval; Dataset partitioning: 70% training set, 20% validation set, 10% test set.

[0083] Step 3: Model Training Linear regression was performed using the least squares method. Loss function: L = i = 1 to n, Pi: the actual number of partitions for the i-th historical data point. : The number of predicted partitions for the i-th historical data point; Optimization method: Gradient descent can be used to adjust model parameters, with a learning rate of 0.01 and a maximum of 1000 iterations.

[0084] Regularization: L2 regularization, with a regularization coefficient of 0.001.

[0085] Step 4: Model Validation The model performance is evaluated using a validation set; evaluation metrics: MAE (mean absolute error), RMSE (root mean square error), R² (coefficient of determination); the objective is, for example: MAE < 10, RMSE < 15, R² > 0.85.

[0086] Step 5: Model Update For example, the model is retrained weekly. When the prediction error continues to exceed 15%, the model is retrained using an incremental learning approach to retain historical knowledge.

[0087] In this step, a pre-trained partition number prediction model can be used in conjunction with the second business feature of the business to be processed to predict the number of partitions and obtain the predicted number of partitions.

[0088] In step 303, the predicted number of partitions is optimized by combining business priority and resource constraint information to obtain target partitioning strategy information and resource configuration scheme information.

[0089] In the embodiments of this disclosure, business priority and resource constraint information are combined and applied to a constraint optimization algorithm to optimize the predicted number of partitions, thereby obtaining target partitioning strategy information and resource configuration scheme information.

[0090] For example, the specific values ​​of the constraints are as follows: Minimum number of partitions (P_min): 40 (to avoid small file issues), maximum number of partitions (P_max): 200 (to avoid excessive task scheduling overhead), target CPU utilization: ≤95% (reserving 5% buffer), target memory utilization: ≤90% (reserving 10% buffer), target task completion time: ≤60 minutes (SLA requirement).

[0091] Constrained optimization algorithms may include the following steps 1 to 5: (wherein the input of the algorithm may include the predicted number of partitions and resource constraints (also called resource constraint information), and the output of the algorithm includes the optimal number of partitions) Step 1: Initialization P_current (current number of partitions) = P_pred (predicted number of partitions), P_min = 40, P_max = 200; Step 2: Constraint check If P_current < P_min, set P_current = P_min; if P_current > P_max, set P_current = P_max; Step 3: Resource evaluation Estimate the CPU usage using the current number of partitions P_current: CPU_usage = f_cpu(P_current); estimate the memory usage using the current number of partitions P_current: Memory_usage = f_memory(P_current); Step 4: Iterative optimization When CPU_usage > 95% or Memory_usage > 90%, the current number of partitions P_current may be adjusted, and the adjustment formula may be as follows: P_current = P_current - 5; Re-evaluate resource usage: if P_current < P_min, set P_current = P_min and exit the loop; Step 5: Output the result P_optimal (optimal number of partitions) = P_current, return P_optimal and the resource evaluation result.

[0092] Optionally, target partition policy information and resource configuration scheme information may be generated based on the optimal number of partitions and the resource evaluation result, which may include, but is not limited to, the number of partitions, the number of CPU cores, the memory size, and the like, for example.

[0093] In the above embodiments, the historical data processing records of the service to be processed may be used as input data, and after service feature analysis, the data is input into a partition number prediction model to predict the number of partitions. After the prediction result is subjected to constraint optimization, the optimal partition policy and resource configuration scheme are output, which can significantly improve resource utilization.

[0094] Figure 4 is a schematic flowchart of the large model-based service processing method provided by the embodiments of the present disclosure. As Figure 4 shown, the above optional implementation manner for performing intelligent quality monitoring on data after code execution task processing based on the large model includes but is not limited to the following steps: In step 401, anomaly detection is performed on the data after the code execution task is processed based on the large model.

[0095] In the embodiments of this disclosure, the anomaly detection may include, but is not limited to, one or more of the following: statistical anomaly detection, rule anomaly detection, and business logic anomaly detection.

[0096] Optionally, multiple anomaly detection methods (such as statistical anomaly detection, rule-based anomaly detection, and business logic anomaly detection) can be used to detect anomalies in the data after the code execution task is processed.

[0097] For statistical anomaly detection, statistical methods and machine learning can be combined to intelligently detect data anomalies, establish a normal data distribution model, calculate statistical indicators such as the mean, standard deviation, and quantiles of the data, and use statistical methods to detect data anomalies. These statistical methods may include, but are not limited to, the 3σ principle detection and quartile detection.

[0098] For rule-based anomaly detection, rule-based anomaly detection can be performed based on business rules to identify data that violates business rules, including null value detection, range detection, regular expression detection, and uniqueness detection.

[0099] For business logic anomaly detection, large models can be used to detect anomalies at the business level, including business logic anomalies, data quality anomalies, data distribution anomalies, relationship anomalies, and potential business risks.

[0100] In step 402, root cause analysis is performed on the detected anomalies to analyze the anomaly context and identify the root causes of the anomalies.

[0101] In the embodiments of this disclosure, root cause analysis is performed on the detected anomalies, analyzing the anomaly context, which may include, but is not limited to, data context, business context and technical context, to identify the root cause of the anomaly. The root cause may include, but is not limited to, data problems, code problems, configuration problems and external dependency problems.

[0102] For example, data context analysis mainly analyzes the temporal distribution, spatial distribution, and user distribution of abnormal data, identifies the central trends and patterns of abnormal data, and generates a data context analysis report. For business context analysis, it can be used to analyze the relationship between abnormal data and business rules, identify the impact of abnormal data on business processes, analyze the relationship between abnormal data and historical business data, and generate business context analysis reports. For technical context analysis, it can analyze system operating status and resource usage, and examine system logs and error logs. It can also analyze the status of dependent systems and generate a technical context analysis report.

[0103] In the embodiments of this disclosure, data context analysis reports, business context analysis reports, and technical context analysis reports can be combined to identify root causes using decision trees or rule engines. Root cause classifications include the following: (1) Data issues: source data quality issues, data transmission issues, and data format issues; (2) Code issues: code logic errors, code performance issues, code compatibility issues; (3) Configuration issues: incorrect parameter configuration, incorrect environment configuration, incorrect dependency configuration; (4) External dependency issues: third-party service failures, network problems, API changes.

[0104] In step 403, remediation recommendation information and / or remediation impact assessment information are generated based on the type of abnormal root cause.

[0105] In the embodiments of this disclosure, the repair suggestion information may include, but is not limited to, one or more of the following: immediate repair solution information, short-term repair solution information, and long-term optimization suggestion information. Immediate repair refers to repair measures that can be implemented immediately, short-term repair refers to repair measures that can be implemented in the short term, and long-term optimization refers to optimization measures that can be implemented in the long term.

[0106] In embodiments of this disclosure, the remediation impact assessment information may include, but is not limited to, one or more of the following: scope of impact, risk level, and remediation cost.

[0107] In the above embodiments, data processed by code execution tasks can be used as input data. After statistical anomaly detection, rule anomaly detection, and business logic anomaly detection, an anomaly report is generated. After root cause analysis, the anomaly report outputs remediation suggestions and impact assessments, which can improve the accuracy of root cause analysis, significantly enhance the effectiveness of remediation suggestions, and strengthen data quality assurance capabilities. Compared with traditional methods that rely on manual monitoring and rule verification, this approach can significantly shorten anomaly detection time and greatly improve detection efficiency.

[0108] To further improve the system's intelligence and adaptability, optionally, in some possible embodiments, if an anomaly is detected during the intelligent quality monitoring phase, the business knowledge base is updated based on the anomaly feedback information; and / or, if a code anomaly occurs during the code generation phase, the code knowledge base is updated based on the code anomaly feedback information.

[0109] It should be noted that this disclosure designs a multi-module collaborative feedback closed-loop mechanism. The forward collaborative process ensures that the output of each module serves as the input of the next module, forming a complete data processing pipeline: from the business understanding module to the code generation module to the decision optimization module to the quality monitoring module, outputting business analysis reports, generated code, resource configuration schemes, and anomaly detection results, respectively. The reverse feedback mechanism ensures that the system can learn from failures and continuously optimize the performance of each module. The quality monitoring module updates the business knowledge base of the business understanding module when anomalies are detected; the code generation module updates the technical knowledge of the code knowledge base module when generation failures occur; and the decision optimization module updates the historical data knowledge base when prediction deviations occur.

[0110] It is worth noting that, to ensure system security, this disclosure incorporates a robust security mechanism. The output verification mechanism rigorously verifies the output of the large model to ensure that the output content meets expectations. A security filtering mechanism checks whether the output of the large model contains sensitive information or malicious content. Code security checks the generated code to prevent issues such as SQL injection and excessively broad privileges. The key management scheme uses an enterprise-level key management system and rotates encryption keys periodically.

[0111] Figure 5 This is a block diagram of a large-model-based service processing apparatus provided in an embodiment of this disclosure. For example... Figure 5 As shown, the business processing device based on the large model may include, but is not limited to: a first generation module 501, a business analysis module 502, a second generation module 503, a code generation module 504, a resource configuration generation module 505, and a quality monitoring module 506.

[0112] The first generation module 501 is used to generate a first prompt text based on the business information of the business to be processed.

[0113] The business analysis module 502 is used to perform business analysis on the business to be processed based on the business knowledge base and the first prompt text, and obtain the business analysis results of the business to be processed.

[0114] The second generation module 503 is used to generate a second prompt text based on the business analysis results and code template.

[0115] The code generation module 504 is used to generate compliant target code for the business to be processed based on the second prompt text and the large model.

[0116] The resource configuration generation module 505 is used to generate target partitioning strategy information and resource configuration scheme information for the code execution task of the business to be processed based on the business characteristics and system status of the business to be processed through a large model.

[0117] The quality monitoring module 506 is used for intelligent quality monitoring of data after code execution tasks are processed based on a large model.

[0118] In some embodiments, the first generation module 501 is used to: fill the business information of the business to be processed into a preset first prompt template to obtain a first prompt text; wherein, the first prompt template includes at least business analysis dimensions, output format requirements and large model call parameters, and the output format requirements are used to require that the business analysis results include the fields corresponding to the analysis results of each dimension.

[0119] In some embodiments, business analytics dimensions include one or more of the following: business pattern recognition; data feature analysis; business logic understanding; unified model mapping; performance and resource assessment; and risk and challenge analysis.

[0120] In some embodiments, the second generation module 503 is used to: select a code template from a code template library based on a first business feature of the business to be processed; wherein the first business feature is obtained based on feature extraction of business information; and fill the business analysis result and the selected code template into a preset second prompt template to obtain a second prompt text.

[0121] In some embodiments, the second generation module 503 is configured to: calculate business similarity based on the first business characteristics of the business to be processed and the business characteristics of one or more reference businesses in the code knowledge base, to obtain one or more business similarities; select a code template from the code template library based on the highest business similarity among the one or more business similarities; wherein, if the highest business similarity is greater than or equal to a first threshold, a dedicated code template for the business type of the corresponding reference business is selected; if the highest business similarity is less than the first threshold but greater than or equal to a second threshold, a general code template is selected; and if the highest business similarity is less than the second threshold, a basic code template is selected.

[0122] In some embodiments, the second prompt text includes at least generation requirement information, which includes one or more of the following: code style requirements, performance optimization requirements, error handling requirements, logging requirements, and data quality requirements.

[0123] In some embodiments, the code generation module 504 is used to: generate basic code by filling template parameters of the code template with a large model based on the business analysis results and code specification requirements in the second prompt text and in conjunction with the code knowledge base; perform performance tuning on the basic code based on the performance optimization requirements in the second prompt text; perform quality checks and optimizations on the performance-tuned basic code based on the data quality requirements in the second prompt text; and after the quality check passes, add error handling and / or log recording information based on the error handling requirements and / or log recording requirements in the second prompt text to generate target code that conforms to the specifications for the business to be processed.

[0124] In some embodiments, the resource configuration generation module 505 is used to: perform feature analysis on the historical data processing records of the business to be processed based on the large model to obtain the second business feature of the business to be processed; input the second business feature into the partition number prediction model to predict the number of partitions; and combine business priority and resource constraint information to perform constraint optimization on the predicted number of partitions to obtain target partitioning strategy information and resource configuration scheme information.

[0125] In some possible embodiments, the quality monitoring module 506 is used to: perform anomaly detection on the data after code execution task processing based on a large model; wherein, the anomaly detection includes one or more of the following: statistical anomaly detection, rule anomaly detection, and business logic anomaly detection; perform root cause analysis on the detected anomalies, analyze the anomaly context, and identify the root cause of the anomaly; and generate repair suggestion information and / or repair impact assessment information according to the type of anomaly root cause.

[0126] In some possible embodiments, such as Figure 6 As shown, the business processing device based on the large model may further include a resource scheduling module 607. The resource scheduling module 607 is used to: schedule resources for code execution tasks based on resource scheduling strategy information during the execution of code execution tasks; wherein the resource scheduling strategy information includes one or more of the following: prioritizing resource allocation for high-priority tasks, prioritizing scheduling tasks nearing their deadlines, dynamically adjusting based on current resource usage, and predicting resource requirements based on historical execution times. Figure 6 601-606 and Figure 5 China 501-506 have the same function and structure.

[0127] In some possible embodiments, such as Figure 7 As shown, the business processing device based on the large model may further include an update module 708. The update module 708 is used to: update the business knowledge base based on the feedback information of the anomaly detected during the intelligent quality monitoring phase; and / or, update the code knowledge base based on the feedback information of the code anomaly occurring during the code generation phase. Figure 7 701-707 and Figure 6 China 601-607 have the same function and structure.

[0128] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0129] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0130] like Figure 8The diagram shown is a block diagram of an electronic device according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0131] like Figure 8 As shown, the electronic device includes one or more processors 801, a memory 802, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 8 Take the 801 processor as an example.

[0132] The memory 802 is the non-transitory computer-readable storage medium provided in this disclosure. The memory stores instructions executable by at least one processor to cause the at least one processor to perform the large-model-based business processing method provided in this disclosure. The non-transitory computer-readable storage medium of this disclosure stores computer instructions for causing a computer to perform the large-model-based business processing method provided in this disclosure.

[0133] Memory 802, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the large-model-based business processing method in the embodiments of this disclosure (e.g., appendix). Figure 5 The first generation module 501, business analysis module 502, second generation module 503, code generation module 504, resource configuration generation module 505, and quality monitoring module 506 are shown. The processor 801 executes various server functions and data processing by running non-transient software programs, instructions, and modules stored in the memory 802, thereby implementing the large-model-based business processing method in the above method embodiments.

[0134] The memory 802 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 802 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 802 may optionally include memory remotely located relative to the processor 801, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0135] The electronic device may also include an input device 803 and an output device 804. The processor 801, memory 802, input device 803, and output device 804 can be connected via a bus or other means. Figure 8 Taking the example of a connection between China and Israel via a bus.

[0136] Input device 803 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as touch screens, keypads, mice, trackpads, touchpads, joysticks, one or more mouse buttons, trackballs, joysticks, etc. Output device 804 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The display device may include, but is not limited to, liquid crystal displays (LCDs), light-emitting diode (LED) displays, and plasma displays. In some embodiments, the display device may be a touch screen.

[0137] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, application-specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0138] These computational programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0139] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0140] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0141] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0142] Figure 9 A block diagram of an intelligent agent provided in an embodiment of this disclosure. (See diagram below.) Figure 9 As shown, the intelligent agent may include an input module 901, a processing module 902, and an output module 903. The input module 901 is used to receive input information (e.g., business information of the service to be processed, etc.); the processing module 902 is used to determine the target task based on the input information received by the input module, determine a large model based on the target task, and execute the business processing method based on the large model in the above method embodiment by calling the large model to obtain output information; the output module 903 is used to output the output information obtained by the processing module (e.g., target code of the service to be processed, data after code execution task processing, etc.).

[0143] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0144] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A business processing method based on a large model, comprising: Generate the first prompt text based on the business information of the pending business; Based on the business knowledge base and the first prompt text, the business to be processed is analyzed using a large model to obtain the business analysis results of the business to be processed. A second prompt text is generated based on the business analysis results and code template; Based on the second prompt text, generate compliant target code for the pending business using the large model; Based on the business characteristics and system status of the pending business, target partitioning strategy information and resource configuration scheme information are generated for the code execution task of the pending business through the large model. Intelligent quality monitoring is performed on the data after the code execution task is processed based on the large model.

2. The method according to claim 1, wherein, The generation of the first prompt text based on the business information of the pending business includes: The business information of the business to be processed is filled into a preset first prompt template to obtain the first prompt text; The first prompt template includes at least business analysis dimensions, output format requirements, and large model call parameters. The output format requirements are used to require that the business analysis results include the fields corresponding to the analysis results of each dimension.

3. The method according to claim 2, wherein, The business analysis dimensions include one or more of the following: business pattern recognition; data feature analysis; business logic understanding; unified model mapping; performance and resource assessment; and risk and challenge analysis.

4. The method according to claim 1, wherein, The generation of the second prompt text based on the business analysis results and code template includes: Based on the first business characteristic of the business to be processed, a code template is selected from the code template library; wherein, the first business characteristic is obtained based on feature extraction of the business information; The business analysis results and the selected code template are filled into the preset second prompt template to obtain the second prompt text.

5. The method according to claim 4, wherein, The step of selecting a code template from the code template library based on the first business characteristic of the business to be processed includes: Based on the first business feature of the business to be processed and the business features of one or more reference businesses in the code knowledge base, business similarity is calculated to obtain one or more business similarities; Based on the highest business similarity among the one or more business similarities, a code template is selected from the code template library; wherein, if the highest business similarity is greater than or equal to a first threshold, a dedicated code template for the corresponding reference business business type is selected; if the highest business similarity is less than the first threshold but greater than or equal to a second threshold, a general code template is selected; and if the highest business similarity is less than the second threshold, a basic code template is selected.

6. The method according to claim 1, wherein, The second prompt text includes at least the generation requirement information, which includes one or more of the following: code style requirements, performance optimization requirements, error handling requirements, logging requirements, and data quality requirements.

7. The method according to claim 6, wherein, The step of generating compliant target code for the pending business based on the second prompt text using the large model includes: Based on the business analysis results and code specification requirements in the second prompt text, and in conjunction with the code knowledge base, the template parameters of the code template are filled into the large model to generate basic code; Based on the performance optimization requirements stated in the second prompt text, the basic code is subjected to performance tuning. Based on the data quality requirements stated in the second prompt text, perform quality checks and optimizations on the basic code that has undergone the performance tuning process; After the quality check is passed, error handling and / or logging information is added according to the error handling requirements and / or logging requirements in the second prompt text, and target code that conforms to the specifications is generated for the business to be processed.

8. The method according to claim 1, wherein, Based on the business characteristics and system status of the pending business, the large model generates target partitioning strategy information and resource configuration scheme information for the code execution task of the pending business, including: Based on the large model, feature analysis is performed on the historical data processing records of the business to be processed to obtain the second business feature of the business to be processed. The second business feature is input into the partition number prediction model to predict the number of partitions; By combining business priority and resource constraint information, the predicted number of partitions is constrained and optimized to obtain target partitioning strategy information and resource allocation scheme information.

9. The method according to claim 8, further comprising: During the execution of the code execution task, resource scheduling is performed on the code execution task based on resource scheduling strategy information; wherein, the resource scheduling strategy information includes one or more of the following: prioritizing resource allocation for high-priority services, prioritizing scheduling for tasks nearing their deadline, dynamically adjusting based on current resource usage, and predicting resource demand based on historical execution time.

10. The method according to claim 1, wherein, The intelligent quality monitoring of the data after the code execution task is processed based on the large model includes: Based on the large model, anomaly detection is performed on the data after the code execution task is processed; wherein, the anomaly detection includes one or more of the following: statistical anomaly detection, rule anomaly detection, and business logic anomaly detection; Root cause analysis is performed on detected anomalies to analyze the anomaly context and identify the root causes of the anomalies. Based on the type of the abnormal root cause, generate remediation suggestion information and / or remediation impact assessment information.

11. The method according to any one of claims 1-10, further comprising: If an anomaly is detected during the intelligent quality monitoring phase, the business knowledge base is updated based on the feedback information of the anomaly. And / or, In the event of a code exception during the code generation phase, the code knowledge base is updated based on the feedback information of the code exception.

12. A business processing device based on a large model, comprising: The first generation module is used to generate the first prompt text based on the business information of the business to be processed. The business analysis module is used to perform business analysis on the business to be processed based on the business knowledge base and the first prompt text, and obtain the business analysis results of the business to be processed. The second generation module is used to generate a second prompt text based on the business analysis results and the code template; The code generation module is used to generate compliant target code for the business to be processed based on the second prompt text and the large model. The resource configuration generation module is used to generate target partitioning strategy information and resource configuration scheme information for the code execution task of the pending business based on the business characteristics and system status of the pending business and through the large model. The quality monitoring module is used to perform intelligent quality monitoring on the data after the code execution task is processed based on the large model.

13. An intelligent agent, comprising: The input module is used to receive input information; The processing module is configured to determine a target task based on the input information received by the input module, determine a large model based on the target task, and obtain output information by calling the large model to execute the method of any one of claims 1-11. An output module is used to output the output information obtained by the processing module.

14. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-11.

15. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-11.

16. A computer program product comprising a computer program, wherein, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-11.

Images