A general service system AI model unified configuration and full life cycle management method

Abstract

The application belongs to the technical field of computer software architecture and artificial intelligence integration, and discloses a general business system AI model unified configuration and whole life cycle management and control method. The application designs a general artificial intelligence model configuration data table, centrally stores basic information, interface access addresses, hierarchical authorization keys, version parameters and whole-process operation trace data of multi-vendor and multi-purpose AI models; adopts database table value parameters as standardized interaction carriers of front and rear ends, combines centralized encapsulation storage processes, uniformly carries various operation requests of front ends, and integrally realizes AI model adding, modifying, deleting and default model setting business processing. The application breaks through the industry disadvantages of dispersed configuration, messy interface, disordered management and control and operation without traceability when traditional application software accesses artificial intelligence, the overall scheme is not limited by industries, can be adapted to all kinds of scenes such as desktop application software, informatization management system, commercial business platform and office software, and provides reliable bottom technology support for standardized, safe and standardized access of all kinds of programs to AI capability.

Description

Technical Field

[0001] This invention belongs to the fields of computer software architecture, database application and artificial intelligence integration technology, and specifically relates to a method for unified configuration, centralized management and full life cycle control of artificial intelligence models applicable to various general business systems. Background Technology

[0002] With the widespread adoption of artificial intelligence technology, various application software generally need to access intelligent services such as large models and third-party AI interfaces to realize extended functions such as voice interaction, intelligent recognition, data analysis, and office assistance.

[0003] Existing conventional development solutions have obvious technical flaws: Configuration information for various AI models is stored in a scattered manner, and there is no unified management entry point for interface addresses, authorization keys, and version parameters, resulting in high costs for later upgrades, replacements, and maintenance. There is a lack of a unified standard process for adding, modifying, and deleting models. Each business module writes its own database interaction logic, resulting in high code redundancy and poor reusability. The system lacks a robust data verification mechanism, which can easily lead to problems such as duplicate configurations, missing key parameters, and accidental deletion of core service models. Interface authorization keys are mostly stored and displayed in plaintext without any security protection measures, which can easily lead to the leakage of sensitive information. Operations such as configuration modifications and model adjustments leave no log traces, making it impossible to locate the source of the operation when the system malfunctions or the configuration is abnormal; Most AI access solutions are custom-developed for single projects, lacking versatility and unable to be reused across industries and software, thus limiting their applicability.

[0004] Therefore, there is an urgent need to design a universal, centralized, secure, and traceable AI model lifecycle management solution to meet the general needs of all types of software accessing artificial intelligence. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a unified configuration and full lifecycle management method for AI models in general business systems. It is implemented through the cooperation of a front-end interaction layer, a data interaction layer, a database logic layer, and a data storage layer. The overall architecture is designed to be universal and is not limited by industry or software type.

[0006] The front-end interaction layer includes a model overview interface, a new editing interface, and a default configuration management interface, enabling data loading and display, window adaptation, and custom list styles; it also sets up account permission interception to restrict unauthorized personnel from operating high-risk functions; and it de-identifies sensitive fields such as keys to avoid displaying them in plaintext.

[0007] The data interaction layer uniformly encapsulates the table-valued parameter structure and fixes the interaction field specifications. All configuration operations are carried out by batch passing data through table-valued parameters, eliminating the need for the front end to directly write business SQL statements, thus reducing code risk and maintenance difficulty.

[0008] The database logic layer is equipped with a dedicated integrated stored procedure to uniformly handle business operations of AI models across all categories. It has built-in complete logic for parameter validation, duplicate detection, transaction control, and exception handling to ensure that business processing rules are consistent and standardized.

[0009] The data storage layer establishes an independent AI model configuration data table, which centrally and persistently stores basic model information, authentication information, version information, status identifiers, and operation log data to ensure long-term stable data storage. Detailed Implementation

[0010] After logging into the system with an authorized account, you will enter the AI ​​model management interface. The system will automatically load all the entered model configuration data and display it visually. When adding a new AI model, fill in information such as the enterprise entity, model name, function category, interface access address, multi-level authorization key, and version number. After submission, the front end assembles the information into standardized table-value parameters and passes them into the database. The stored procedure sequentially performs non-empty checks, duplicate checks for the same category, and unique checks for the same name. After the checks pass, the data is inserted through a transaction. If there are no models in the current category, this data is automatically marked as enabled by default. When you need to modify the model configuration, select the target item to enter the editing interface, update the corresponding parameters and submit. The stored procedure updates the data according to the unique identifier and automatically records the modifier, operation address and modification time. When you need to switch the default model, go to the dedicated configuration interface, select the target models in batches, and submit them in batches. The database will automatically clear the original default identifiers of the same category and set the selected models as the global default calling objects. When performing a deletion operation, the stored procedure first determines whether the target model is the default service model. If so, it intercepts the deletion command. If there are no protection restrictions, it completes the data deletion through a transaction. All operations are automatically recorded and audited, and sensitive key fields are displayed in anonymized form for an extended period, ensuring system stability and information security. Beneficial effects

[0011] High versatility: The solution has no industry restrictions and can be adapted to various application software such as office software, commercial platforms, and information systems, with high reusability; Centralized management: Unified management of AI models from multiple vendors and of multiple types simplifies configuration and maintenance processes and reduces iteration and operation and maintenance costs; Security protection: Multiple protections including access control, key desensitization, default model to prevent accidental deletion, and transaction rollback enhance the system's security level; Logical uniformity: Standardized interaction is achieved by relying on table-valued parameters and centralized stored procedures, resulting in unified business rules and good scalability; Full traceability: All configuration operations are automatically logged, facilitating troubleshooting and operation auditing, and meeting standardized usage requirements; High degree of automation: New models are automatically initialized with default states, and the default switching process is simplified, reducing manual configuration costs.

Claims

1. A method for unified configuration and full lifecycle management of AI models in general business systems, characterized in that, Includes the following steps: (1) Construct a general AI model configuration data table and set common fields such as enterprise entity, model name, model category, interface address, multi-level key, model version, default identifier, operator, operation address, and operation time; (2) Customize the database table value parameter structure as the data interaction carrier between the front end and the database, and uniformly encapsulate operation type, data primary key, model parameters, operator and client address information; (3) The table-value parameter data is received uniformly through centralized stored procedures, and four types of business operations such as adding, modifying, deleting and configuring default models of AI models are completed in a centralized manner. (4) Configure multi-level verification logic inside the stored procedure, including parameter non-empty verification, record quantity verification, duplicate data uniqueness verification, and default model protection verification; (5) All data write and change operations are uniformly wrapped by database transactions, which captures operational exceptions and automatically rolls back transactions, and synchronously outputs standardized exception prompts. (6) The front-end program has built-in permission judgment logic, which allows only authorized accounts to perform AI model addition, modification and deletion operations, and directly intercepts and prompts unauthorized operations; (7) The front-end interface displays key-type sensitive data in an anonymized manner, hiding plaintext content to prevent the leakage of sensitive authentication information; (8) All configuration change operations automatically retain the operator, client address, and operation time, forming an operation audit ledger that can be traced for a long time.

2. The method according to claim 1, characterized in that: AI models are categorized by function into speech recognition, text processing, intelligent analysis, and visual recognition. Within each category, only a single model data item with a default identifier is retained.

3. The method according to claim 1, characterized in that: When adding a new AI model, the system automatically retrieves existing data for the corresponding category. If there are no existing models in the category, the newly added entry is automatically marked as enabled by default.

4. The method according to claim 1, characterized in that: When performing a model deletion operation, the default identifier of the target data is automatically detected, and models in the default service state are prohibited from being deleted to avoid interruption of business AI functions.

5. The method according to claim 1, characterized in that: The default model configuration supports batch selection of multiple data entries, which can be passed to the database in batches via table-valued parameters. This process clears the original default state of the same category and completes the binding of the new default model.