Industrial software full check and one-time exception return method for ai call

Abstract

This invention discloses a method for full verification and one-time error return of industrial software for AI calls, relating to the field of automatic AI-industrial software integration technology. This invention employs a three-level full pre-verification and one-time error summary return mechanism. Through basic verification of empty tables, operation type validity verification, and full field verification by operation type, it achieves one-time submission, full verification, and one-time return of all errors. It establishes unified verification rules and error return paradigms, fully adapting to various development languages, compatible with both manual and AI calls, and can be used in conjunction with a unified data interaction interface architecture for industrial software. This invention solves the pain points of low AI integration efficiency and resource waste caused by traditional sequential error reporting, significantly reducing invalid calls, facilitating automatic AI parsing and correction, and has been fully implemented in actual industrial ERP and MES systems, possessing mature engineering value and industrial promotion capabilities.

Description

Technical Field

[0001] This invention belongs to the field of AI and industrial software automatic integration technology, specifically involving a method for full parameter verification and one-time exception return of industrial software adapted for batch automated AI calls. It is suitable for standardized and automated integration of large AI models with industrial ERP, MES, PLC, quality, equipment, inventory and other systems. It can be used with the unified data interaction interface architecture of industrial software and is fully compatible with all development languages / development systems such as Java, C#, .NET MAUI, Web development, and ARKTS. It is widely used in various manufacturing industries such as metallurgy, machinery, chemical industry, automobile, and electronics, as well as related fields of industrial digitalization and intelligentization. Background Technology

[0002] During the intelligent upgrade of industrial software, the demand for automated integration between AI large models and industrial ERP, MES and other systems has surged. However, the parameter verification methods of traditional industrial software are designed to adapt to manual operation, which is seriously incompatible with the characteristics of AI batch automated invocation. At the same time, the verification rules are not uniform across development languages / development systems, resulting in problems such as low AI integration efficiency and waste of system resources. The core pain points are as follows: Repeated error reporting results in extremely high AI interaction costs: Traditional validation stops validation and returns a single prompt when it encounters an error, requiring the caller to repeatedly modify and submit. While this is manageable for manual operations, AI generates a large number of invalid interactions during batch automated calls, resulting in extremely low integration efficiency. The lack of pre-emptive empty table interception wastes system resources: Without an empty table verification step, empty tables and invalid parameters without data will still enter subsequent business logic, performing meaningless calculations and data processing, resulting in a large waste of server resources; Inconsistent validation rules and high adaptation costs: Validation rules for different development languages / systems and different business segments are chaotic. There is no universal validation system for manual and AI calls, and customized validation logic needs to be developed separately for AI, resulting in high development and maintenance costs. The non-standard error return format makes it impossible for AI to automatically parse: Traditional error prompts are designed for manual reading, with arbitrary expressions and inconsistent formats. The AI ​​large model cannot accurately identify the error points and correction requirements, and cannot achieve automated parameter correction and re-entry. Poor compatibility with the unified interaction architecture of industrial software: Traditional verification logic is designed with a single interface and a single operation, which cannot be adapted to the unified interface architecture of industrial software with multiple operations for a single business. Verification logic needs to be developed repeatedly, which is contrary to the standardized interaction system.

[0003] Existing technologies lack a set of industrial software parameter verification methods that feature **"full pre-verification, one-time exception return, and universal adaptation across the entire system"**. This not only fails to meet the high-efficiency requirements of AI batch automated calls but also makes it difficult to seamlessly integrate with the unified data interaction architecture of industrial software, thus becoming a significant technical barrier to the intelligent implementation of industrial software. Summary of the Invention

[0004] This invention provides a method for full verification and one-time exception return of industrial software for AI calls. It employs a core design of **"three-level full pre-verification + one-time exception summary return"**, achieving full verification and one-time return of all errors with a single submission, eliminating the need for repeated interactions. It establishes unified verification rules, parameter passing specifications, and exception return formats, fully adapting to all development languages / systems such as Java, C#, .NET MAUI, Web development, and ARKTS. This single verification system is compatible with both manual operation and automated batch calls by AI. The verification rules and calling paradigm of this patent are adapted to the core design of unified encapsulation and calling of two or more operation types in the main patent. Conventional verification scenarios for single operation types can refer to this method but are not included in the exclusive protection scope of this patent. Furthermore, this method can seamlessly integrate with the unified data interaction interface architecture of industrial software, significantly reducing the cost of AI and industrial software integration, improving integration efficiency, and providing a standard verification foundation for the intelligent upgrade of industrial software. Technical solution

[0005] The core of this invention is **"tiered full-scale pre-verification, unified one-time exception return, and universal applicability across the entire system and all scenarios"**. It achieves pre-emptive interception of all vulnerabilities through a three-level tiered verification logic, eliminates repeated interactions by combining a unified exception aggregation and one-time return mechanism, and establishes fully unified rules and paradigms to adapt to all development languages / development systems and a unified data interaction architecture for industrial software. The core steps are as follows: Level 1 Basic Validation: Empty Table Direct Rollback After receiving the Table Value Parameter (TVP) data from the front-end / AI, an empty table validation is performed first. If the incoming TVP table is empty and contains no business data, a standardized exception message "Incoming table value parameter is empty, no valid business data" is output directly. The transaction is immediately rolled back and all subsequent validations and business logic are terminated, intercepting invalid calls at the source and avoiding waste of system resources. Empty table data from all development languages / development systems are validated and processed according to this unified rule.

[0006] Second-level core verification: After the empty table verification passes, the operation type legality verification is performed. According to the standardized identifiers of add (Z), subtract (J), delete (S), modify (G), and query (C) uniformly defined by the main patent, the following contents are fully verified: whether the operation type field is empty, whether multiple operation types are passed at the same time, and whether the operation type is in the system's legal list. This step does not interrupt the verification process and does not return exceptions separately. All exception information related to operation types is collected, concatenated and summarized to ensure that all legality issues are fully identified.

[0007] Three-level precise verification: After the operation type legality verification is completed, based on the current legal operation type (single type), a full and precise verification is performed on the core business fields corresponding to the operation. This fully covers the verification of required fields not being empty, the validity of numerical ranges, the standardization of data formats, and compliance with business rules. For example, the contract creation operation verifies required fields such as finished product volume number and customer information, and the unit price modification operation verifies that the unit price value is positive and the unit price precision meets industry standards. Similarly, the verification does not stop midway. All field-related abnormal information continues to be collected, spliced, and summarized to ensure that all data problems are identified without omission.

[0008] After the unified exception aggregation and one-time return mechanism completes the three-level full verification, it judges the exception information aggregation result: if the aggregation result is not empty (there is one or more errors), it immediately executes transaction rollback, and returns all collected exception information to the caller in a standardized format, clearly marking each error point, error type and correction requirements; if the aggregation result is empty (no errors), it does not return exception information, directly executes the subsequent business logic in sequence, and only returns a standardized "operation successful" prompt after the business execution is completed, without any redundant information.

[0009] A unified validation and adaptation rule is established across all development languages ​​and systems. This rule applies to all languages ​​and systems, including Java, C#, .NET MAUI, Web development, and ARKTS. For .NET framework languages, DataTable format data is passed according to the business-specific TVP table type, and the validation layer parses the DataTable fields for validation according to the unified rule. For Java / Web development (Vue / React), ARKTS, and other languages ​​and systems, JSON format data is encapsulated according to the business-specific TVP table type, and the validation layer parses the JSON key-value pairs for validation according to the unified rule. The underlying validation logic is fully compatible, only the format parsing is adapted, and the validation rules are not changed.

[0010] A unified exception return format specification for both human and AI establishes a standardized and structured exception return format that is compatible with both human reading and AI automatic parsing. The returned content includes the error code, the total number of errors, and a detailed error list. The detailed error list clearly indicates the field name, error description, and correction requirements for each error, using a fixed key-value pair / structured format to ensure that large AI models can accurately extract and automatically parse the error, enabling automated parameter correction and re-entry. When manually accessing the system, users can directly read the error details to quickly locate the problem.

[0011] Seamlessly integrated with the unified data interaction architecture of industrial software, this method's verification logic is fully compatible with the **"single business-specific table type + unified interaction rules"** architecture of the main patent. The verification layer directly connects to a unified standardized interface, receives business-specific TVP data transmitted by the interface, and performs precise verification according to the operation type field. There is no need to develop separate verification logic for different businesses and different interfaces. When adding a new business module, it is only necessary to add the field verification rules for that business in the third-level verification. There is no need to modify the overall architecture of the verification layer, which is completely consistent with the extended compatibility rules of the main patent, achieving non-intrusive integration.

[0012] Core Innovation Points The three-level hierarchical full-scale pre-verification design intercepts all invalid calls at the source: Through three-level hierarchical verification of empty tables, operation types, and business fields, it achieves full-scale vulnerability pre-interception, direct rollback of empty tables, and full identification of all errors, completely abandoning the traditional sequential error reporting mode, reducing invalid calls and system resource waste from the source, and adapting to the high-efficiency requirements of AI batch calls.

[0013] The one-time exception summary and return mechanism completely eliminates the pain point of repeated AI interactions: all errors are collected uniformly and returned completely at once. AI can obtain all error information with a single submission, eliminating the need for repeated modifications and submissions. This significantly reduces the cost of AI integration and interaction, reducing invalid calls by more than 90%, making it an ultimate adaptation for AI calling scenarios.

[0014] Universal rules across the entire system and all scenarios significantly reduce development and adaptation costs: A unified set of validation rules, adaptation rules, and return formats are established, fully adapting to all development languages / systems such as Java, C#, .NET MAUI, Web development, and ARKTS. A single validation system is compatible with both manual operation and AI-automated invocation, eliminating the need for repetitive development of customized logic and significantly reducing development and maintenance costs.

[0015] Standardized and structured exception returns enable AI to automatically analyze and correct errors: The exception return format is designed according to AI recognition requirements, presenting error information in a structured and standardized manner. AI can accurately extract error points, automatically correct parameters, and achieve re-entry without human intervention, truly realizing automated and unmanned integration between AI and industrial software.

[0016] Seamlessly integrated with the unified interaction architecture of industrial software to form a complete technical system: The verification logic and extension rules of this method are fully compatible with the unified data interaction architecture of the main patent, and can be directly connected to the unified standardized interface. New business extensions are non-intrusive, forming a complete industrial software + AI docking standard system of "unified interaction + unified verification" with the main patent. The patent protection boundary is clear. Conventional verification of single operation type is not restricted, and only verification paradigms that are uniformly encapsulated for multiple operation types are included in exclusive protection. Beneficial effects

[0017] Ultimately adaptable to AI batch automated calls, significantly improving integration efficiency: Completely solves the pain points of AI interaction in the traditional sequential error reporting mode. One-time submission, full verification, and one-time return of all errors reduce invalid calls between AI and industrial software by more than 90%, greatly improving integration efficiency and removing verification technology barriers for the intelligent implementation of industrial software.

[0018] Pre-emptive interception of all invalid parameters significantly reduces system resource consumption: Level 1 empty table verification intercepts calls without data at the source, and Level 3 full verification pre-emptively intercepts all illegal parameters, preventing invalid parameters from entering subsequent business logic, reducing meaningless calculations and data processing on the server, significantly reducing system resource consumption, and improving the overall operating efficiency of industrial software.

[0019] Universal across all development systems and scenarios, significantly reducing development and adaptation costs: Fully supports all development languages / development systems such as Java, C#, .NET MAUI, Web development, and ARKTS. A single verification system is compatible with both manual and AI calls, eliminating the need to develop separate verification logic for different languages ​​and different callers, reducing development workload by more than 80%. Later maintenance only requires monitoring one set of rules, greatly improving operational efficiency.

[0020] Standardized and structured anomaly information enables automated AI analysis and re-entry: The standardized anomaly return format is adapted to AI recognition requirements. AI can accurately analyze error information, automatically correct parameters, and achieve automated re-entry without human intervention. This truly realizes unmanned and automated integration between industrial software and large AI models, promoting the deeper implementation of industrial intelligence.

[0021] Seamlessly integrated with the unified data interaction architecture to form a complete standard system: This method is seamlessly integrated with the unified data interaction interface architecture of the main patent. The verification logic is connected to the unified interface and the extension rules are consistent with the main patent, forming a complete industrial software + AI docking standard system of "unified data interaction + unified AI friendly verification". It can be quickly implemented in industrial software in various industries such as metallurgy and machinery, with strong adaptability and replicability.

[0022] The entire operation is traceable and transaction is controllable, meeting industrial compliance requirements: all verification operations, exception returns, and business executions are guaranteed to be atomic through transaction control, and automatic rollback is performed in case of exceptions to avoid data chaos; at the same time, information such as the verification operator, operation time, and calling entity (human / AI) is recorded to meet the requirements of industrial software data traceability and financial compliance, and data processing is safe and reliable. Detailed Implementation

[0023] The specific implementation of this invention is based on a metallurgical industry ERP system, seamlessly integrated with the metallurgical business scenario of the main patent "Unified Data Interaction and Standardized Interface Method for Industrial Software", and fully compatible with development languages / development systems such as Java, C#, .NET MAUI, Web development, and ARKTS. The specific implementation steps are as follows: The underlying architecture for the three-level full verification is based on the industrial software backend development framework. An independent three-level full verification layer is built, which connects to the unified standardized interface / api / industrial / unifiedInterface of the main patent. It receives TVP data (DataTable / JSON format) from all business modules and all development languages / development systems. The verification layer has a built-in format parsing module that automatically completes the parsing and conversion of data of different formats. The underlying verification logic is unified and only performs format adaptation.

[0024] Configure a unified empty table validation rule in the validation layer: if the number of rows of parsed TVP data is 0 and there are no business records, immediately return the standardized exception message "ERROR: The passed table value parameter is empty, there is no valid business data, error code: NULL_TABLE", execute transaction rollback and terminate subsequent logic; this rule is universal for all businesses and all development languages / development systems, and requires no personalized modification.

[0025] Configure the secondary operation type validity verification rules according to the unified definition of Z (add), J (delete), S (delete), G (modify), and C (query) in the main patent. Configure the operation type verification rules to check whether the operation type field is not empty, whether only a single type is passed, and whether it is in the [Z,J,S,G,C] valid list. Collect all abnormal information of violations, such as "operation type field is empty", "both Z and G operation types are passed at the same time, only a single type is allowed", "operation type is X, non-valid identifier". The abnormal information is concatenated in the format of "field name - error description - error code".

[0026] Configure three levels of full field validation rules based on operation type for core businesses in the metallurgical industry, such as spot contracts, futures contracts, inventory, and equipment. Configure dedicated field validation rules based on operation type. Taking spot contract business as an example: Add (Z) operation: Verify that the fields such as contract number, customer code, steel coil code, quantity, and unit price are not empty, and verify that the quantity / unit price is positive and the contract number conforms to the metallurgical industry coding standard; Modify (G) operation: verify that the contract number is not empty and exists in the system, verify that the modified unit price / quantity is positive, and verify that the modified amount does not exceed the industry limit; all field error information is concatenated in the format of "field name - error description - error code", and summarized with the operation type exception information into a complete error detail list.

[0027] Configure unified exception return and business execution rules. Configure standardized exception return format: {"errorCode":"FAIL","errorCount":n,"errorList":[{"field":"field name","desc":"error description","code":"error code"}]}; Configure business execution rules: If errorCount=0, directly call the main patent's business-specific stored procedure to execute the corresponding logic, and return {"errorCode":"SUCCESS","msg":"operation successful","data":[]} after execution.

[0028] Practical verification implementation across development languages / development architectures The C#-developed metallurgical MES system calls the spot contract add (Z) operation: passing TVP data in DataTable format. If the steel coil code is empty and the unit price is negative, the verification layer fully identifies the two errors and returns a detailed list containing the two errors at once. The MES system front end displays the list directly and allows for quick manual correction. A Java-developed futures management system calls a contract modification (G) operation: JSON-formatted TVP data is passed without any errors, the validation layer directly calls the futures contract stored procedure to execute the modification logic, and returns a success message. ARKTS developed a HarmonyOS-based in-plant mobile system that calls an inventory lookup (C) operation: it transmits JSON-formatted TVP data, with the operation type field set to Y (invalid). After the validation layer identifies the error, it returns a standardized exception, and the HarmonyOS-based system automatically prompts the error, eliminating the need for repeated manual submissions. When a sales management platform in a web development (Vue) calls a contract deletion (S) operation, it passes empty TVP data, which is directly intercepted by the first-level validation, returns an empty table exception, and does not execute any subsequent logic.

[0029] The actual implementation of AI-automated invocation in the metallurgical industry requires batch execution of spot contract modification operations using a large-scale AI model. Data is encapsulated in JSON format according to the TVP table type specific to spot contracts in the main patent, and invoked through a unified interface. If the data input by AI contains some empty contract numbers or some negative unit prices, the verification layer will identify all errors and return all error details in a standardized format at once. The AI ​​model automatically parses the error details list, extracts error records such as empty contract numbers and negative unit prices, automatically completes parameter correction, repackages the data, and calls the unified interface again. The corrected data is error-free, and the verification layer directly executes the modification logic. The AI ​​can complete batch operations with a single correction without human intervention, achieving automated integration. In this embodiment, the AI ​​calls the business interface that is uniformly encapsulated for multiple operation types in the main patent. The verification paradigm of this patent is fully adapted to this scenario and falls within the exclusive protection scope of this patent.

[0030] The expansion of the new business segment to include the processing of incoming materials in the metallurgical ERP system can be achieved by adding a dedicated TVP table type and stored procedure design according to the main patent rules. Only field validation rules for each operation type of the processing of incoming materials (such as processing order number, incoming material weight, processing fee, etc.) need to be added to the three-level validation layer of this method. There is no need to modify the overall architecture of the validation layer or the validation rules of other businesses. This can achieve full validation of the processing of incoming materials. Both manual and AI calls are processed according to the same rules, achieving non-intrusive and seamless integration with the main patent.

Claims

1. A method for full verification and one-time exception return of industrial software for AI calls, characterized in that, It adopts a "three-level full pre-verification + one-time exception summary return" design, which can be used in conjunction with the unified data interaction interface architecture of industrial software, including the following steps: (1) Build a three-level full verification underlying architecture, connect to the unified standardized interface of industrial software, and receive table value parameter (TVP) data passed in by all development languages / development systems; (2) Perform first-level basic checks: If the input TVP table is empty, output standardized exception information directly, perform transaction rollback and terminate subsequent logic; (3) Perform secondary core verification: perform full verification of the legality of operation type fields without interrupting the verification process, and collect splicing exception information in a unified manner; (4) Perform three-level precise verification: Based on the legal operation type, perform full and precise verification on the corresponding core business fields, continue to collect splicing abnormal information in a unified manner, and identify all data problems without omission; (5) Perform unified exception summary and one-time return: If there is an exception, return all error information in a standardized structured format at once and perform transaction rollback; if there is no exception, directly execute the subsequent business logic and return operation success; (6) Formulate unified verification and adaptation rules for all development languages / development systems to achieve automatic parsing and compatibility of data in different formats; (7) Formulate a standardized exception return format that is consistent between human and AI, and adapt to AI automatic parsing and correction; the verification rules and calling paradigm of this patent are adapted to the core design of unified encapsulation and calling of two or more operation types in the main patent. The conventional verification scenarios of single operation types can refer to this method and are not included in the exclusive protection scope of this patent.

2. The method according to claim 1, characterized in that, The method is fully compatible with all development languages / systems, including Java, .NET MAUI, Web development, and ARKTS. .NET framework languages ​​transmit TVP data in DataTable format, while Java / Web development, ARKTS, and other languages / systems encapsulate TVP data in JSON format. The validation layer has a built-in format parsing module that automatically completes the parsing and conversion of data in different formats, and the underlying validation logic is unified.

3. The method according to claim 1, characterized in that, The validity check of the secondary operation type is performed according to the standardized identifiers of add (Z), subtract (J), delete (S), modify (G), and query (C). The operation type field is checked in its entirety to see if it is not empty, if only a single type is passed in, and if it is in the valid list. All abnormal information is collected uniformly and not returned separately.

4. The method according to claim 1, characterized in that, The three-level precise verification is a full field verification based on the operation type, covering verification of required fields not being empty, verification of numerical range validity, verification of data format standardization, and verification of business rule compliance. All field abnormal information and operation type abnormal information are summarized in a unified manner, and the verification does not exit midway.

5. The method according to claim 1, characterized in that, The standardized structured exception return format includes error codes, total number of errors, and a detailed error list. The detailed error list clearly indicates the field names, error descriptions, and correction requirements for each error. It adopts a fixed key-value pair / structured format and is compatible with both human reading and AI automatic parsing, enabling AI-automated parameter correction and re-entry.

6. The method according to claim 1, characterized in that, The method is seamlessly integrated with the "single business-specific table type + unified interaction rules" architecture of the main patent "Unified Data Interaction and Standardized Interface Method for Industrial Software". The verification layer directly connects to the unified standardized interface. When adding a new business module, only the field verification rules of that business need to be added. There is no need to modify the overall architecture of the verification layer, thus achieving non-intrusive expansion.

7. The method according to any one of claims 1-6, characterized in that, All verification operations and business executions are guaranteed to be atomic through transaction control. In case of an anomaly, the system will automatically roll back immediately to avoid data corruption. At the same time, the operator, operation time, and calling entity information of the verification operation are recorded to meet the data traceability and compliance requirements of industrial software.

8. The method according to any one of claims 1-6, characterized in that, The method is widely used in industrial software systems such as ERP, MES, and PLC in various manufacturing industries such as metallurgy, machinery, chemical, automotive, and electronics, as well as in all industrial software scenarios with AI automation call requirements, such as smart IoT and enterprise digital management. It can achieve seamless adaptation between human and AI, and greatly improve the efficiency of intelligent docking of industrial software.