Structured data construction method, task processing method and system

Abstract

The embodiment of the present specification provides a structured data construction method, a task processing method and a system, wherein the structured data construction method comprises: determining business logic information and data dependency information according to global business information of a target business; constructing a business operation graph structure according to the business logic information, and constructing a data description graph structure according to the data dependency information; establishing an execution mapping relationship of the business operation graph structure and the data description graph structure, an access mapping relationship of a business data source of the target business and the data description graph structure, and a configuration execution constraint condition for the business operation graph structure, the data description graph structure and the business data source, to obtain a target data structure; and constructing an agent guided structured data corresponding to the target business based on an action executable file corresponding to the target data structure, the business data source and the business operation graph structure.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to methods for constructing structured data, methods for processing tasks, and systems. Background Technology

[0002] With the development of computer and internet technologies, intelligent agent tools can now support long-term planning and autonomous execution. Users only need to provide macro-level business objectives, and the tool can automatically complete multi-step cyclical operations such as file reading, tool invocation, command execution, result verification, and error correction within minutes to hours. However, intelligent agents generally suffer from divergent thinking and logical illusions when handling complex business processes. For example, when faced with multi-step, long-cycle financial business operations, intelligent agents struggle to accurately understand and break down standard business rules, lacking the ability to decompose business processes. This can lead to their behavior deviating from the established business logic and failing to reliably complete task objectives. Furthermore, existing technologies impose overly simplistic constraints on intelligent agents, primarily relying on assigning a limited number of skills or tools for control. In business scenarios involving cross-domain collaboration, multi-node interaction, and complex data flow, this constraint method cannot fully describe the topology and necessary constraints of the business process. Consequently, intelligent agents often get trapped in ineffective loops, repeatedly invoking unrelated tools, resulting in a significant waste of computing resources, especially when handling long-process business scenarios. Therefore, an effective solution is urgently needed to address these issues. Summary of the Invention

[0003] In view of this, embodiments of this specification provide a method for constructing structured data. One or more embodiments of this specification also relate to task processing methods, task processing systems, structured data construction apparatus, task processing devices, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, a method for constructing structured data is provided, comprising: Based on the global business information of the target business, determine the business logic information and data dependency information; A business operation graph structure is constructed based on the business logic information, and a data description graph structure is constructed based on the data dependency information; Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configure execution constraints for the business operation graph structure, the data description graph structure and the business data source to obtain the target data structure; Based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure, construct the intelligent agent guidance structured data corresponding to the target business.

[0005] According to a second aspect of the embodiments of this specification, a task processing method is provided, applied to a business intelligence agent, comprising: Based on the business processing request submitted for the target business, the target sub-business flowchart structure is determined in the intelligent agent-guided structured data, wherein the intelligent agent-guided structured data is constructed using the above method; A business task sequence is generated based on the target sub-business process diagram structure, and the pending business task corresponding to the current task execution cycle is determined in the business task sequence. In the structured data guided by the intelligent agent, at least one business action unit corresponding to the business task to be processed is invoked, and each business action unit is executed in sequence as the task execution operation of the business task to be processed. If all business tasks included in the business task sequence have been completed, task execution information is determined as a response to the business processing request.

[0006] According to a third aspect of the embodiments of this specification, a task processing system is provided, including an operation terminal and a processing terminal, comprising: The operation terminal is used to construct a business processing request in response to the business processing information submitted by the business user for the target business, and send the business processing request to the processing terminal; The processing terminal is configured to: invoke a business agent to determine the target sub-business process flow diagram structure in the agent-guided structured data according to the business processing request, wherein the agent-guided structured data is constructed using the method described above; generate a business task sequence based on the target sub-business process flow diagram structure, and determine the pending business task corresponding to the current task execution cycle in the business task sequence; invoke at least one business action unit corresponding to the pending business task in the agent-guided structured data, and execute each business action unit sequentially as the task execution operation of the pending business task; and, if all business tasks included in the business task sequence are completed, determine the task execution information and feed it back to the operation terminal.

[0007] According to a fourth aspect of the embodiments of this specification, a structured data construction apparatus is provided, comprising: The information determination module is configured to determine business logic information and data dependency information based on the global business information of the target business. The graph structure building module is configured to build a business operation graph structure based on the business logic information, and to build a data description graph structure based on the data dependency information; The relationship establishment module is configured to establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and to configure execution constraints for the business operation graph structure, the data description graph structure and the business data source to obtain the target data structure; The data construction module is configured to construct intelligent agent-guided structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure.

[0008] According to a fifth aspect of the embodiments of this specification, a task processing apparatus is provided, applied to a business intelligence agent, comprising: The determination module is configured to determine the target sub-business flowchart structure in the agent-guided structured data based on the business processing request submitted for the target business, wherein the agent-guided structured data is constructed using the method described above. The generation module is configured to generate a business task sequence based on the target sub-business process diagram structure, and to determine the pending business task corresponding to the current task execution cycle in the business task sequence. The execution module is configured to call at least one business action unit corresponding to the business task to be processed in the intelligent agent-guided structured data, and execute each business action unit in sequence as the task execution operation of the business task to be processed. The determination module is configured to determine task execution information as a response to the business processing request, provided that all business tasks included in the business task sequence have been completed.

[0009] According to a sixth aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-described structured data construction method or task processing method.

[0010] According to a seventh aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described structured data construction method or task processing method.

[0011] According to an eighth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described structured data construction method or task processing method.

[0012] The structured data construction method provided in this embodiment determines business logic information and data dependency information based on the global business information of the target business; constructs a business operation graph structure based on the business logic information and a data description graph structure based on the data dependency information; establishes the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configures execution constraints for the business operation graph structure, the data description graph structure, and the business data source to obtain the target data structure; and constructs the intelligent agent guidance structured data corresponding to the target business based on the action executable files corresponding to the target data structure, the business data source, and the business operation graph structure. By systematically constructing the business operation graph structure, the data description graph structure, and their mapping relationships, and configuring multi-level constraints, it achieves precise constraints and guidance on the intelligent agent's behavior, effectively solving the problems of behavioral deviation and resource waste of intelligent agents in complex business environments, while preventing logical illusions and bypassing levels of operation, ensuring the accurate execution of business processes, and improving system reliability and resource utilization efficiency. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating a structured data construction method provided in one embodiment of this specification; Figure 2 This is a flowchart illustrating a task processing method provided in one embodiment of this specification; Figure 3 This is a schematic diagram of an agent guiding structured data in a task processing method provided in one embodiment of this specification; Figure 4 This is a schematic diagram of an agent performing a task in a task processing method provided in one embodiment of this specification; Figure 5 This is a flowchart illustrating the processing procedure of a task processing method provided in one embodiment of this specification. Figure 6 This is a schematic diagram of the structure of a task processing system provided in one embodiment of this specification; Figure 7 This is a schematic diagram of a structured data construction apparatus provided in one embodiment of this specification; Figure 8 This is a schematic diagram of the structure of a task processing device provided in one embodiment of this specification; Figure 9 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0014] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0015] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0016] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0017] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0018] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0019] In the field of artificial intelligence, an intelligent agent is an entity capable of perceiving its environment, making autonomous decisions, and taking actions to achieve a specific goal. It can be a software program, a hardware device (such as a robot), or a complex system.

[0020] This specification provides a method for constructing structured data. One or more embodiments of this specification also relate to task processing methods, task processing systems, structured data construction apparatus, task processing devices, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0021] See Figure 1 , Figure 1 A flowchart of a structured data construction method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0022] Step S102: Determine business logic information and data dependency information based on the global business information of the target business.

[0023] Step S104: Construct a business operation graph structure based on the business logic information, and construct a data description graph structure based on the data dependency information.

[0024] Step S106: Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configure execution constraints for the business operation graph structure, the data description graph structure and the business data source to obtain the target data structure.

[0025] Step S108: Based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure, construct the intelligent agent guidance structured data corresponding to the target business.

[0026] The structured data construction method provided in this embodiment can be applied to the processing of structured data for intelligent agents in any business scenario, such as financial business scenarios, intelligent customer service scenarios, R&D automation programming scenarios, text writing scenarios, and image editing scenarios. It is used to construct structured data that plays a constraining and guiding role. After configuring it to the corresponding intelligent agent, it can effectively solve the problems of behavioral deviation and resource waste of intelligent agents in complex business environments, while preventing logical illusions and bypassing levels of operation, ensuring the accurate execution of business processes, and improving system reliability and resource utilization efficiency.

[0027] This embodiment uses the application of structured data construction methods in financial business scenarios as an example for illustration. For the same or corresponding descriptions in other scenarios, please refer to the same or corresponding descriptions in this embodiment. This embodiment will not elaborate further here.

[0028] Specifically, the target business refers to a specific business scenario or process, such as loan approval, transaction clearing, or customer service in the financial sector. This target business has its unique business logic, data flow, and compliance requirements. Global business information refers to the complete and accessible collection of information related to the target business, including but not limited to business rule documents, operation manuals, system interface specifications, data dictionaries, and historical transaction records. This information is the foundation for understanding and modeling the target business. Business logic information refers to the information extracted from the global business information that describes the core business rules and processes, such as the order of operations, conditional judgments, and decision paths of the target business. It defines how the business is executed, and the operations performed in the target business are recorded in the business operation diagram and associated with the atomic capabilities of the nodes in the diagram. Data dependency information refers to the information extracted from the global business information that describes the relationships between data entities in the target business, data attributes, data flow, and data access permissions. It defines what data is required for business execution.

[0029] Correspondingly, a business operation graph structure refers to a model that graphically represents business logic information. Nodes can represent business processes, sub-processes, or specific business actions, while edges represent the sequence or dependencies between these operations. This structure aims to provide a macro-to-micro execution sequence of business processes. A data description graph structure refers to a model that graphically represents data dependency information. Nodes can represent business entities (such as customers or accounts) or data attributes (such as names or balances), while edges represent relationships between entities and attributes, or between entities themselves. This structure aims to provide an AI-friendly representation of business data and its associations.

[0030] Correspondingly, the execution mapping relationship refers to the association established between operation nodes in the business operation graph structure and data nodes in the data description graph structure. This relationship specifies which data entities or attributes a particular business operation needs to access or modify. The business data source refers to the physical or logical database system that stores data related to the target business. It is the carrier of the actual data read and written by the business operation, such as... Figure 3The diagram illustrates that in a financial business scenario, the business data source can include resource-related databases, API interfaces (business data interfaces), and dialogue-related databases. This means that any data required for the execution of a business-related task can be obtained from the business data source. Access mapping relationships refer to the associations established between specific data tables, fields, or business data interfaces in the business data source and data nodes in the data description graph structure. This relationship specifies how abstract data in the data description graph corresponds to the actual database storage location and business data interface. It can be understood that the access mapping relationship establishes a mapping between the data description graph structure and the data tables / fields recorded in the business database of the business data source, as well as between the data description graph structure and the business data interfaces in the business data source. This is used to obtain relevant data and call corresponding business data interfaces to support the agent's work during the application phase. Execution constraints refer to the rules or restrictions set to regulate the execution behavior of the business operation graph structure, data description graph structure, and business data source. These can include operation permissions, data access permissions, restrictions on bypassing levels, execution order restrictions, etc., aiming to ensure the compliance and security of business execution. In one optional implementation, execution constraints can be further divided into global axiomatic constraints and node-local constraints (data compliance or process verification triggered only when accessing specific actions or data nodes) based on their scope, thereby achieving finer-grained rule engine scheduling. In practice, the setting of execution constraints can be freely adjusted according to actual needs; this embodiment will not make any such adjustments. The target data structure refers to the comprehensive data model used for agent guidance, formed after establishing the business operation graph structure, data description graph structure, various mapping relationships, and configuring execution constraints. This structure encapsulates all the logic, data, and rules required for business execution.

[0031] Correspondingly, action executable files refer to program or script files that correspond to specific business action nodes in the business operation graph structure and can be invoked in the computing environment. These files encapsulate the underlying code logic for executing specific business actions. Agent-guided structured data refers to the final constructed structured data file that can be directly loaded and used by the business agent. This data file contains the target data structure, access paths to the business data source, and references to action executable files, providing the agent with a complete and controlled navigation and execution framework for executing the target business.

[0032] Based on this, the first step is to determine the business logic information and data dependencies according to the overall business information of the target business. In practice, manual analysis can be used, such as manually identifying the various stages, decision points, and required data elements and their interrelationships in the business process from business documents, regulations, and system interface specifications. For example, for a loan approval business, the approval manual can be used to identify business stages such as "application submission," "qualification review," "risk assessment," and "approval decision," and the data required for each stage, such as user identity information, income verification, and credit reports, can be identified.

[0033] Secondly, a business operation diagram structure is constructed based on the aforementioned business logic information, and a data description diagram structure is constructed based on the aforementioned data dependency information. For the business operation diagram structure, the identified business process steps and actions are connected using basic graphical symbols such as sequence, branch, and loop to form an intuitive business operation diagram structure. For example, in loan approval, "Submit Application" can be used as the starting node, "Qualification Review" and "Risk Assessment" as parallel nodes, and "Approval Decision" as the convergence node, with arrows indicating the process direction. For the data description diagram structure, an entity-attribute list format can be used to list the identified business entities (such as "Customer" and "Loan") and their associated attributes (such as "Name" and "ID Number") in a flat manner, and simply indicate the direct relationships between them. For example, the "Customer" entity is listed, with attributes including "Name" and "ID Number"; the "Loan" entity is listed, with attributes including "Amount" and "Term," and it is noted that there is an "Application" relationship between "Customer" and "Loan."

[0034] Next, establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the target business's business data source and the data description graph structure, and configure execution constraints for the business operation graph structure, data description graph structure, and business data source to obtain the target data structure. Specifically, the execution mapping relationship can be established by adding connections between each action node in the business operation graph structure and the data entities or attributes in the data description graph structure. For example, the "qualification review" action node connects to customer identity information and income verification. The access mapping relationship can be established by maintaining a separate table, linking each data entity or attribute name in the data description graph structure to the corresponding table name, field name, and business data interface in the business data source. For example, the table records "customer identity information" corresponding to the "ID_CARD" field of the database table "T_CUSTOMER," or the table records "information verification" corresponding to the "enterprise information verification service interface API" in the business data interface. Execution constraints can take the form of a simple rule list, such as listing "approval decision actions can only be executed by users with 'approval permissions'" or "loan amount cannot exceed 1 million," etc. These mapping relationships and constraints can be integrated into a configuration file to form a preliminary target data structure.

[0035] Finally, based on the target data structure, business data source, and the corresponding action executable files of the business operation graph structure, the intelligent agent guidance structured data corresponding to the target business can be constructed. Specifically, the aforementioned target data structure can be directly serialized into a JSON or XML file, and the connection string of the business data source and the path information of the action executable files can be embedded as metadata into this structured file. For example, a JSON file can contain various parts of the target data structure, storing database connection information in one field and a list in another field listing the names of all action executable files and their relative paths in the file system. When the intelligent agent executes, it can parse the JSON file and perform operations based on the information contained therein.

[0036] This embodiment uses the structured data construction process corresponding to the "personal microloan approval" business in a financial scenario as an example. Based on the global business information of the "personal microloan approval" business, such as the loan approval manual, customer information management system interface documents, risk control rule sets, etc., the business logic information and data dependency information are determined. For example, the business logic information may include a series of processes and actions such as "customer submits application", "system automatically conducts preliminary review", "manual qualification review", "risk score calculation", and "approval result notification". The data dependency information may include data entities and their attributes such as "customer basic information (name, ID number, contact information)", "financial status information (income, liabilities)", and "credit history information (credit report)". Furthermore, a business operation graph structure can be constructed based on the business logic information, and a data description graph structure can be constructed based on the above data dependency information. The business operation graph structure can be constructed as a directed graph, where "customer submits application" is the starting node, "system automatically conducts preliminary review" is the subsequent node, and so on, with each node representing a specific business action or process. The data description graph structure can be constructed as an entity-attribute graph. For example, the "customer" entity is associated with attributes such as "name" and "ID number"; the "loan application" entity is associated with attributes such as "application amount" and "application period".

[0037] Based on this, execution mapping relationships can be established between the business operation graph structure and the data description graph structure, as well as access mapping relationships between the target business's business data source and the data description graph structure. Execution constraints can also be configured for the business operation graph structure, data description graph structure, and business data source to obtain the target data structure. For example, the execution mapping relationship can map the "Manual Qualification Review" action node in the business operation graph structure to the "Customer Basic Information" and "Financial Status Information" entities in the data description graph structure. The access mapping relationship will map the "Customer Basic Information" entity in the data description graph structure to fields such as "NAME" and "ID_NUMBER" in the "T_CUSTOMER" table of the business data source. Execution constraints will be configured; for example, the "Approval Result Notification" action cannot be executed before the "Risk Score Calculation" is completed. Furthermore, data access constraints can be configured; for example, the agent cannot directly modify the "ID Number" field in "Customer Basic Information" when performing any operation, unless through a specific authentication process. Finally, the target data structure is serialized into a unified JSON file. This file not only contains the business operation diagram, data description diagram, and their mappings and constraints, but also embeds the connection configuration information of the business data source, as well as references to the actual executable programs corresponding to each business action (such as "system automatic preliminary review" and "risk score calculation"). When the business agent receives the task of "processing personal microloan applications," it loads and parses this agent guidance structured data. The agent will strictly follow the business operation diagram structure defined in this data to execute the task step by step.

[0038] The structured data construction method provided in this embodiment determines business logic information and data dependency information based on the global business information of the target business; constructs a business operation graph structure based on the business logic information and a data description graph structure based on the data dependency information; establishes the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configures execution constraints for the business operation graph structure, the data description graph structure, and the business data source to obtain the target data structure; and constructs the intelligent agent guidance structured data corresponding to the target business based on the action executable files corresponding to the target data structure, the business data source, and the business operation graph structure. By systematically constructing the business operation graph structure, the data description graph structure, and their mapping relationships, and configuring multi-level constraints, it achieves precise constraints and guidance on the intelligent agent's behavior, effectively solving the problems of behavioral deviation and resource waste of intelligent agents in complex business environments, while preventing logical illusions and bypassing levels of operation, ensuring the accurate execution of business processes, and improving system reliability and resource utilization efficiency.

[0039] Furthermore, directly constructed business operation graph structures are often logically chaotic and unevenly granular, making it difficult for intelligent agents to accurately locate business process nodes when executing tasks. Therefore, in this embodiment, constructing the business operation graph structure based on the business logic information includes: The business logic information is hierarchically divided and structurally transformed according to preset hierarchical division information and operational constraint information to obtain a local business process diagram structure and a local business action diagram structure. The local business process diagram structure includes at least one sub-business process diagram structure corresponding to a business process. Based on the local business process diagram structure and the local business action diagram structure, a business operation diagram structure corresponding to the target business is generated.

[0040] Specifically, the pre-defined hierarchical information refers to information used to define the decomposition hierarchy and structured rules of business logic. For example, this information could be a set of predefined hierarchical relationship templates for business domains, business processes, sub-business processes, and business actions, used to guide the progressive refinement of business logic. Alternatively, this information could be configuration rules stored in XML or JSON format, based on a business classification system developed according to industry standards or internal enterprise specifications. Operational constraint information refers to a set of rules used to regulate business operations and ensure compliance and security. For example, this information could include the execution order of business actions, preconditions, postconditions, concurrency limits, and permission requirements to ensure the standardization of business operations. Alternatively, this information could be operational specifications set based on business risk control requirements. Hierarchical division and structural transformation of business logic information refers to decomposing and reconstructing the original, potentially unstructured, business logic description according to pre-defined rules, giving it an AI-friendly hierarchical structure and standardized operational definitions. For example, semantic analysis of business documents can be performed using natural language processing (NLP) technology. By combining pre-defined hierarchical information and operational constraints, business processes and actions can be automatically identified and extracted, and then mapped to predefined structured templates.

[0041] Correspondingly, local business process diagram structures and local business action diagram structures refer to the results of hierarchical division and structural transformation, and are the basic components for constructing a complete business action diagram structure. For example, a local business process diagram structure can be represented as a directed acyclic graph (DAG), where nodes represent sub-business processes and edges represent the sequential or parallel relationships between processes; a local business action diagram structure can be represented as a more fine-grained graph structure, where nodes represent specific business actions. A local business process diagram structure includes at least one sub-business process diagram structure corresponding to a business process, emphasizing the hierarchical nature of business processes. A large business process can be decomposed into multiple smaller, manageable sub-business processes. For example, a local business process diagram structure for "customer onboarding" can include "identity verification sub-business process diagram structure," "document submission sub-business process diagram structure," and "account opening sub-business process diagram structure," etc. This inclusion relationship can be reflected through nested nodes or parent-child relationships in the graph structure, ensuring the modularity and reusability of the business process.

[0042] Based on this, a standardized processing framework is provided for the structured construction of business logic by introducing pre-defined hierarchical division information and operational constraint information. When constructing the business operation diagram structure, the complex business logic information is first decomposed from macro to micro levels using the pre-defined hierarchical division information, breaking it down into locally defined business process diagram structures with clear hierarchies. Each locally defined business process diagram structure can contain multiple sub-business process diagram structures, effectively solving the problems of vague business process descriptions and uneven granularity. Simultaneously, the business logic is structurally transformed using operational constraint information, encapsulating detailed business actions into locally defined business action diagram structures, embedding constraints such as execution order and preconditions within them. This limits the scope of the agent's actions from the source, preventing logical divergence. Finally, by organically integrating these hierarchical locally defined business process diagram structures and detailed locally defined business action diagram structures, the business operation diagram structure corresponding to the target business is generated. This business operation diagram structure not only provides a complete view of the business process but also embeds detailed action execution logic and mandatory constraint mechanisms, forming an "operation map." It also stipulates that leaf nodes must and can only be "business actions." For example, in financial business scenarios, all financial compliance constraints and execution prerequisites are encapsulated within the leaf nodes. This structured construction approach enables subsequent steps in establishing the execution mapping relationship between the business operation diagram structure and the data description diagram structure, configuring execution constraints, and ultimately building the intelligent agent to guide structured data. This is based on a highly standardized, logically clear, and AI-friendly business operation model, significantly enhancing the intelligent agent's logical execution capabilities and compliance in complex financial scenarios, and improving the accuracy of the intelligent agent's task execution.

[0043] Taking a bank's "account opening business" as an example, its business logic may involve a series of complex steps, including customer identity verification, information entry, account creation, and card issuance. To construct its business operation diagram structure, predefined hierarchical information is first defined. For example, the "account opening business" is divided into two main processes: "personal account opening process" and "corporate account opening process." Each main process includes sub-processes such as "identity verification," "document submission," and "account opening." Simultaneously, operational constraints are defined, such as requiring "identity verification" to be completed before "document submission," and "account opening" to proceed only after both "identity verification" and "document submission" are successful. Furthermore, sensitive information entry requires double verification. Next, the system performs hierarchical division and structural transformation of the original "account opening business" logic based on this predefined information, generating a complete and hierarchically defined "bank account opening business operation diagram structure." This structure displays all the details from the macro-level business process to the micro-level business actions, as well as their dependencies and compliance constraints, providing precise navigation and execution guidance for the intelligent agent to perform the "account opening business."

[0044] In summary, by embedding pre-defined hierarchical division information and operational constraint information into the business operation diagram structure, the intelligent agent can strictly follow predefined business processes and operational norms when executing tasks, significantly improving the accuracy, compliance, and efficiency of task execution. Especially in highly compliant scenarios such as finance, it can effectively reduce operational risks.

[0045] Furthermore, when facing complex business scenarios, there is a lack of clear implementation paths for accurately and automatically extracting the key business elements required to construct a data description graph structure from massive and messy data dependency information. Therefore, in this embodiment, the construction of the data description graph structure based on the data dependency information includes: According to the preset object description information, extract business entity information and business attribute information from the data dependency information; based on the business entity information and the business attribute information, construct the data description graph structure corresponding to the target business.

[0046] Specifically, the pre-defined object description information can be a structured metadata definition, such as XML SchemaDefinition (XSD) or JSON Schema, which contains the structured definitions, data types, constraint rules, and semantic descriptions of core entities in the business domain (such as "customer," "product," and "transaction") and their associated attributes (such as "customer ID," "product name," and "transaction amount"). Data dependency information can refer to the raw data table structures, field definitions, data dictionaries, and data lineage documents stored in an internal data warehouse (EDW), operational data store (ODS), or data lake. Furthermore, data dependency information can also be business system interface documents, API specifications, database design documents (such as entity-relationship diagrams (ER diagrams), and data flow diagrams). Extracting business entity information and business attribute information can be done through a parser or data extraction tool, scanning the data dependency information according to the patterns or rules defined in the pre-defined object description information, identifying and extracting business entities (such as table names and concept names) and business attributes (such as field names and attribute names) that meet the conditions. In addition, natural language processing (NLP) technology can be used to perform semantic analysis on unstructured or semi-structured data-dependent information (such as business documents), and automatically identify and extract business entities and attributes by combining keywords and concepts in preset object description information.

[0047] Based on this, by introducing pre-defined object description information as a benchmark for filtering and identification, accurate extraction of core business elements from raw data dependency information is achieved. During the construction of the data description graph structure, filtering data dependency information using pre-defined object description information ensures that the extracted business entity information and business attribute information have a high degree of business relevance, avoiding the inclusion of irrelevant data in the model, thus guaranteeing the purity and logical rigor of the data description graph structure. By organically combining the extracted business entity information and business attribute information, a data description graph structure corresponding to the target business is constructed. This approach not only clarifies the logical relationships between data but also provides a standardized data foundation for subsequent mapping to the business operation graph structure.

[0048] In summary, the data description graph structure precisely constructed through the above processing provides AI-friendly and unambiguous data access and logical support for intelligent agents when performing business processing tasks. This greatly reduces the risk of intelligent agents exhibiting divergent thinking or logical illusions in complex financial transactions, and improves the efficiency and compliance of business processing.

[0049] Furthermore, the lack of refined control over the business execution process leads to issues such as unauthorized operations, tool abuse, or chaotic execution logic in intelligent agents, making it impossible to meet the stringent constraints on business processes in highly compliant scenarios such as finance. To address this, this embodiment includes establishing the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the target business's business data source and the data description graph structure, and configuring execution constraints for the business operation graph structure, the data description graph structure, and the business data source to obtain the target data structure, comprising: Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, and establish the access mapping relationship between the business data source of the target business and the data description graph structure. Generate a business topology structure based on the mapping relationship. Add cross-level restriction constraints and tool mounting constraints to the business operation graph structure in the business topology structure, and add execution direction constraints and execution permission constraints to the business topology structure. Generate the target data structure based on the constraints.

[0050] Specifically, establishing the execution mapping relationship between the business operation graph structure and the data description graph structure aims to clarify the correlation between business operations (e.g., process steps, atomic actions) and business data (e.g., entities, attributes). This correlation enables the agent to clearly identify the required data input and the potential data output or impact of the operation when performing a specific operation. This mapping relationship can be achieved by establishing connecting edges between nodes in the business operation graph structure and specific entities or attributes in the data description graph structure. Establishing the access mapping relationship between the target business's business data source and the data description graph structure serves to associate the data elements in the abstract data description graph structure with the physical data tables, fields, and business data interfaces actually stored in the business data source. This allows the agent to access and manipulate the actual data in the underlying database by understanding the data description graph structure. This can be achieved through a metadata management system that binds each data entity or attribute in the data description graph structure to the specific table name, field name, data type, and business data interface in the business data source.

[0051] Correspondingly, a business topology structure is generated based on the mapping relationship. This business topology structure is a unified structure that integrates business operations, data descriptions, and database access mappings. It provides a comprehensive and structured business model, serving as the foundation for adding various constraints subsequently. Generating the business topology structure can be achieved by integrating the two mapping relationships mentioned above into a single data structure, where nodes can represent business operations, data entities, data attributes, or database tables / fields, and edges represent the execution, access, or dependency relationships between them; alternatively, it can be achieved by generating an XML, JSON, or other structured file containing all these related information, forming a parsable business topology model.

[0052] Adding hierarchical constraints to the business operation graph structure within the business topology aims to prevent agents from skipping necessary intermediate steps or compliance nodes when executing business operations, thereby ensuring the sequentiality and compliance of the business process. This can be achieved by defining strict execution orders and preconditions between nodes in the business operation graph structure, requiring agents to meet these preconditions before executing subsequent operations; alternatively, a "hierarchy" or "permission level" can be set for each business operation node, stipulating that agents can only operate between adjacent levels, prohibiting direct calls across levels.

[0053] To address this, tool mounting constraints are added to the business operation graph structure within the business topology. These constraints aim to strictly limit the entry points and methods by which agents can call external tools (e.g., APIs, microservices, large model skills), preventing tool abuse or invocation at inappropriate times. Implementation methods include explicitly specifying callable external tool interfaces on specific leaf nodes (i.e., atomic operations) of the business operation graph structure and restricting other nodes from directly calling these tools; alternatively, a tool registry can be used to bind tools to specific operations within the business operation graph structure, allowing agents to call tools only through this registry.

[0054] Adding execution direction constraints to the business topology clarifies the data flow (input or output) between business operations and data, ensuring that data is correctly used and modified within the business process. This can be achieved by adding "Input" or "Output" markers to the edges between business operation nodes and data entity or attribute nodes in the business topology, indicating whether data is used as input or output for the operation; alternatively, it can be achieved by defining data flow rules that specify certain operations can only read certain data, while other operations can modify that data.

[0055] Adding execution permission constraints to the business topology aims to restrict agents' access to and modification permissions for business data and operations, ensuring that agents can only execute authorized operations and access authorized data. This can be achieved through a Role-Based Access Control (RBAC) model, assigning roles to agents and defining the operations and data permissions (e.g., CRUD permissions) that each role can execute within the business topology; alternatively, it can be achieved through an Attribute-Based Access Control (ABAC) model, dynamically evaluating execution permissions based on agent attributes, operation attributes, data attributes, and environmental attributes.

[0056] Based on this, by establishing the execution mapping relationship between the business operation graph structure and the data description graph structure, as well as the access mapping relationship between the business data source and the data description graph structure, and generating a business topology structure on this basis, the discrete business logic, data model, and underlying data source are organically integrated, providing the agent with a unified and comprehensive business execution view. Furthermore, by adding cross-level constraint conditions and tool mounting constraints to the business operation graph structure, and adding execution direction constraints and execution permission constraints to the business topology structure, multi-dimensional and refined control over the agent's behavior can be achieved.

[0057] In other words, the bypass constraint effectively prevents the agent from skipping key business processes, ensuring the compliance and integrity of the business workflow; the tool mounting constraint strictly limits the timing and method of calling external tools, avoiding arbitrary tool calls and abuse; the execution direction constraint clarifies the data flow path between operations, ensuring the accuracy of data use; and the execution permission constraint fundamentally limits the agent's access to data and operations, eliminating unauthorized operations and logical inconsistencies. The introduction of these constraints ensures that the agent's behavior is strictly confined within preset compliance boundaries when executing tasks, significantly improving the security, standardization, and controllability of business execution. This is particularly suitable for scenarios with extremely high compliance requirements, such as finance, effectively solving problems such as logical divergence and unauthorized operations that may occur in complex business scenarios.

[0058] Continuing with the previous example, based on the global business information of the target business, determine the business logic information and data dependency information, and construct the business operation diagram structure and data description diagram structure. For example, the business operation diagram structure may include steps such as "customer information entry," "credit assessment," "risk approval," "contract signing," and "loan disbursement." The data description diagram structure may include entities such as "customer," "loan application," "credit report," and "contract," as well as attributes such as "customer name," "ID number," "loan amount," and "approval status." The business data source includes the actual Customers table, LoanApplications table, CreditReports table, and DATA API (external interface). Based on this, further establish execution mapping relationships. For example, the "customer information entry" operation maps to the "customer" entity and the "loan application" entity, the "credit assessment" operation maps to the "credit report" entity, and the "risk approval" operation maps to the "approval status" attribute of the "loan application." At the same time, access mapping relationships are established. For example, the "Customer" entity is mapped to the Customers table, the "Loan Application" entity is mapped to the LoanApplications table, the "Credit Report" entity is mapped to the CreditReports table, and information verification is mapped to the "Enterprise Information Verification Service Interface API". Based on these mapping relationships, a unified business topology structure is generated, which integrates operations, data entities and their corresponding relationships in the database.

[0059] Furthermore, refined constraints are added to this business topology. For example, a skip-level restriction constraint is added, stipulating that the agent cannot directly execute "risk approval" without completing "customer information entry," nor can it skip "credit assessment" and directly proceed with "loan disbursement." Specifically, when the agent executes the "risk approval" operation, it must first ensure that the "credit assessment" operation has been completed and its result has been written into the specific attributes of the "loan application" entity. Tool attachment constraints are added; for example, the external "credit inquiry API" tool is attached to the leaf node of the "credit assessment" operation, and the "electronic signature service API" tool is attached to the leaf node of the "contract signing" operation, restricting the agent from calling the "electronic signature service API" during the "customer information entry" stage. Add execution direction constraints. For example, the "Customer Information Entry" operation has an "Output" direction (write data) for the "Customer" and "Loan Application" entities; the "Credit Assessment" operation has an "Input" direction (read data) for the "Credit Report" entity; and the "Risk Approval" operation has an "Output" direction (modify data) for the "Approval Status" attribute of the "Loan Application" entity. Add execution permission constraints. For example, when the agent executes "Customer Information Entry," it has "Create" and "Update" permissions on the Customers and LoanApplications tables; when executing "Credit Assessment," it has "Read" permission on the CreditReports table, but no "Create, Delete, or Update" permissions; and when executing "Loan Disbursement," it has "Update" permission on the "Loan Status" field of the LoanApplications table, but cannot modify other fields. Finally, based on these constraints, a structured data set containing all the above graph structures, mapping relationships, and constraints is generated, serving as guidance and restrictions for the agent when performing loan approval tasks.

[0060] In summary, by designing refined control measures during business execution, intelligent agents can strictly adhere to operational constraints when performing tasks, preventing issues such as unauthorized operations, tool abuse, or chaotic execution logic. This ensures that constraints are strictly followed in any business scenario, effectively guaranteeing business compliance.

[0061] Based on this, how to construct a data structure that can directly support the refined operation of intelligent agents in complex business scenarios is a problem that urgently needs to be solved. In this embodiment, the step of generating the target data structure by adding results based on constraints includes: An initial data structure is generated by adding results based on constraints; the sub-data structures associated with the business operation graph structure in the initial data structure are processed to obtain a business operation structure, wherein the business operation structure includes an action sub-data structure and an orchestration sub-data structure, the action sub-data structure being associated with business actions and the orchestration sub-data structure being associated with business scenarios; the sub-data structures associated with the data description graph structure in the initial data structure are processed to obtain a data description structure, and the sub-data structures associated with the business data source in the initial data structure are processed to obtain a data source registration structure; based on the business operation structure, the data description structure, and the data source registration structure, a target data structure corresponding to the target business is constructed.

[0062] Specifically, the process of generating an initial data structure based on constraints aims to initially structure and integrate the business topology determined in the preceding processing, along with various execution constraints configured on it (such as out-of-level restrictions, tool mounting constraints, execution direction constraints, and execution permission constraints), forming a raw data set containing all necessary metadata and rules. This can be understood as bringing together all business rules and structural information into a unified data model that requires further refinement. For example, the business topology and its associated constraints can be serialized into JSON, XML, or other structured data formats, or stored in a temporary database table as input for subsequent processing.

[0063] The sub-data structures associated with the business operation graph structure in the initial data structure are processed to obtain the business operation structure. This processing aims to identify and extract the parts related to the business operation graph structure from the initial data structure and reconstruct them into a data structure specifically for describing business operations. This business operation structure will serve as the core basis for the agent to execute business actions and orchestrate processes. For example, by parsing the initial data structure, all elements corresponding to the nodes of the business operation graph structure can be identified and organized into an independent operation model that can be directly understood and invoked by the agent according to a predefined pattern. The business operation structure includes an action sub-data structure and an orchestration sub-data structure. The action sub-data structure is the part of the business operation structure used to carry the definition of specific business actions. It abstracts and encapsulates each executable business action, making it the smallest functional unit that the agent can invoke. For example, the action sub-data structure may contain a reference to an external tool or API, or contain a piece of directly executable script code, thereby realizing the invocation of a specific business action. The orchestration sub-data structure is the part of the business operation structure used to describe the business process and scenario logic. It defines the execution order of business actions, conditional dependencies, and how to combine them to complete a complete business scenario.

[0064] The sub-data structures associated with the data description graph structure in the initial data structure are processed to obtain a data description structure. This process aims to extract and reconstruct the parts related to the data description graph structure from the initial data structure, forming a data structure specifically for describing business data. This data description structure will provide the agent with semantic information and a structured view of the business data. For example, by parsing the initial data structure, all elements corresponding to the nodes of the data description graph structure can be identified and organized into an independent data model that the agent can use to understand and manipulate the data according to a predefined pattern. The sub-data structures associated with the business data source in the initial data structure are processed to obtain a data source registration structure. This process aims to extract and reconstruct the parts related to the business data source from the initial data structure, forming a data structure specifically for managing and accessing the underlying physical data source. This data source registration structure will serve as a unified interface and security barrier for the agent to access the business data source. For example, by parsing the initial data structure, all elements related to business data source connection information, table structure, field definitions, etc., can be identified and organized into an independent registry containing data source metadata and access credentials.

[0065] Based on this, by modularly decomposing and reconstructing the initial data structure, a deep transformation from general constraints to business execution units is achieved, providing the agent with a structured and executable business guidance framework. Specifically, after generating the initial data structure by adding results based on constraints, this initial data structure carries the business topology and various execution constraints configured on it, laying the foundation for subsequent structure refinement. Building on this, the sub-data structures associated with the business operation graph structure in the initial data structure are first processed to obtain the business operation structure. This business operation structure is further subdivided into action sub-data structures and orchestration sub-data structures. The action sub-data structure is responsible for encapsulating atomic-level business actions, while the orchestration sub-data structure is responsible for defining the combination logic and execution order of these actions in specific business scenarios. This "dual-axis separation" design completely decouples business actions from process orchestration. When executing tasks, the agent can understand the business process through the orchestration sub-data structure and call specific execution capabilities through the action sub-data structure, effectively avoiding divergent thinking and logical illusions.

[0066] Simultaneously, the sub-data structures associated with the data description graph structure in the initial data structure are processed to obtain the data description structure. This data description structure provides a semantic view of business data, enabling the agent to accurately understand and manipulate the data. Furthermore, the sub-data structures associated with business data sources in the initial data structure are processed to obtain the data source registration structure. This structure serves as a unified access interface for physical data sources, ensuring that the agent can securely and efficiently access the required data when executing tasks, and achieving data traceability. Finally, based on the obtained business operation structure, data description structure, and data source registration structure, the target data structure corresponding to the target business is constructed. This target data structure organically unifies business logic, data description, and data source access, forming a highly structured, hierarchical, and execution-constrained overall model. Through this layered construction approach, the agent operates in a controlled and clear environment, with its decisions and operations strictly guided and constrained, thus solving the problems of traditional agents easily deviating from standards and having limited constraint methods in complex scenarios. This construction method not only improves the accuracy and efficiency of the agent in handling complex business processes but also greatly facilitates subsequent business rule adjustments and auditing.

[0067] Continuing with the previous example, in a financial business scenario, when constructing the target data structure for processing "user loan application approval," the system first generates an initial data structure based on the business topology and configured execution constraints (e.g., restrictions on bypassing approval processes, access permissions for specific tools, etc.). This initial data structure might be a JSON document containing all business process nodes, data entities, database tables, and their relationships and constraints. Next, the sub-data structures associated with the business operation graph structure within this initial data structure are processed to obtain the business operation structure. Specifically, the system identifies business actions related to "loan application approval," such as "identity verification," "credit assessment," "risk check," and "approval decision," and encapsulates them into independent action sub-data structures. Each action sub-data structure can be defined as `type: action` and contains a `metadata.tools` field, pointing to a specific external interface (such as an identity verification API or credit scoring model). Simultaneously, the system constructs and orchestrates sub-data structures based on the business process definition, such as defining them as type: process or module. These structures do not directly mount tools but instead use the hasAction attribute array as pointers to reference the aforementioned action sub-data structures, thereby describing the process sequence and conditional branches of "identity verification -> credit assessment -> risk check -> approval decision". Furthermore, the system processes the sub-data structures associated with the data description graph structure in the initial data structure to obtain the data description structure. For example, the system identifies business entities such as "user", "loan application", and "credit report", as well as their attributes (such as user ID, name, loan amount, credit score, etc.) and constructs an object structure. This structure can be hierarchically divided using the type field (e.g., root -> domain:loan -> entity:user -> attributes:userId) and a tree graph topology is constructed using subNodeOf and edges. On each node, metadata anchoring can be configured. For example, on the "Loan Amount" attribute node, the `flows` attribute can record its corresponding `db_table` and `db_columns` (such as `loan_applications.amount`), achieving physical data anchoring; the `rules` attribute can attach financial compliance constraints (such as "Loan amount shall not exceed 1 million") to intercept illegal data writing. Furthermore, the sub-data structures associated with business data sources in the initial data structure are processed to obtain the data source registration structure. This can be a unified data source registry, containing metadata such as connection information, table structure, and primary / foreign key relationships of the business data sources.For example, it can include detailed definitions of "User Table," "Loan Application Form," and "Credit Report Form," but as a foundation, it only accepts addressing requests from the `flows` identifier in the data description structure, rejecting all direct access requests from the agent, thus achieving physical foundation isolation and unified description. Ultimately, based on the business operation structure, data description structure, and data source registration structure obtained above, the system will construct the target data structure corresponding to "User Loan Application Approval." This target data structure will be a comprehensive structured data, in which the flow and actions of business operations, the semantics and structure of business data, and the access paths of physical data sources are defined and linked in a way that is easier for AI to understand. When the agent executes the loan approval task, it will be guided according to this target data structure.

[0068] In summary, by refining the business operation graph structure into action sub-data structures and orchestration sub-data structures, a complete decoupling of business actions and process orchestration is achieved. This allows the agent to clearly understand atomic-level actions and their combination logic in complex business scenarios, effectively preventing the agent from experiencing divergent thinking and logical illusions when executing complex processes. Simultaneously, by processing the sub-data structures of the data description graph structure and the business data source, data description structures and data source registration structures are obtained respectively. This not only provides the agent with a semantic view of business data but also establishes a secure and traceable physical data access mechanism, avoiding invalid data access and resource waste. Ultimately, the target data structure built based on these three types of structures enables the agent to operate in a highly structured, controlled, and transparent environment, greatly improving the accuracy, compliance, and efficiency of the agent in processing complex financial businesses, and providing a solid foundation for flexible adjustment and auditing of business processes.

[0069] Furthermore, in practical applications, intelligent agents often lack a direct association mechanism with underlying database files and specific action executable files. This results in the agent being unable to accurately locate resources when executing tasks and lacking a clear deployment location, making it difficult for the agent to achieve automated and standardized task processing in complex business environments. To address this, in this embodiment, the construction of the intelligent agent guidance structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure includes: The process involves: determining the database file corresponding to the business data source and the action executable file corresponding to the business operation graph structure; constructing agent-guided structured data corresponding to the target business based on the target data structure, the database file, and the action executable file; wherein, after the step of constructing agent-guided structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure, the process further includes: determining the file addition location corresponding to the business agent of the target business, and adding the agent-guided structured data to the file addition location; wherein, the agent-guided structured data is used by the business agent to execute business processing tasks associated with the target business.

[0070] Specifically, determining the database file corresponding to the business data source refers to identifying and locating the physical storage location or access path of the business database in the file system. This is the foundation for establishing a connection between the agent and the actual data source. Determining the executable file corresponding to the action in the business operation graph structure refers to identifying and locating the physical path of the executable program or script corresponding to each business action defined in the business operation graph structure. This step ensures that the agent can accurately find and call the underlying code that executes the business logic. Determining the file addition location corresponding to the business agent of the target business refers to identifying the specific directory or storage area used by the business agent in the deployment environment to load its configuration, instructions, or tools. This location is crucial for the agent to discover and use structured data. Implementation methods may include, but are not limited to: obtaining its plugin directory or data loading path by reading the business agent's configuration file (such as agent.conf, settings.ini); querying the specific mount point or storage volume allocated to the agent in the operating system or container environment (such as Docker, Kubernetes); or placing the agent's structured data in a specific folder that is automatically scanned when the agent starts, according to predefined deployment specifications. Adding the agent's bootstrap structured data to the file addition location refers to copying, deploying, or linking the constructed agent's bootstrap structured data to a specified directory or storage area that the business agent can access and load. This step enables the agent to obtain and utilize this structured data to perform business processing tasks.

[0071] Based on this, by integrating abstract business logic, specific data storage, and actual execution actions, a comprehensive and executable "brain" is provided for the business agent. First, before constructing the agent's guiding structured data, the system precisely determines the physical database file corresponding to the business database in the business data source and the executable files corresponding to each action in the business operation graph structure. This process ensures that the agent can directly and accurately locate the required data source and execution tools when executing tasks, avoiding execution deviations caused by inaccurate resource lookups or incorrect call paths. On this basis, the previously constructed target data structure (which already includes business logic, data models, execution mappings, and constraints) is deeply integrated with the determined database file and executable files to construct the agent's guiding structured data. This integration is not a simple data stacking, but rather a physical binding of the abstract business processes and data dependencies in the target data structure with the actual database access paths and action call interfaces. For example, a specific business action defined in the target data structure will be explicitly associated with its corresponding executable file in the agent's guiding structured data. Similarly, data entities involved in the target data structure will be associated with specific tables or fields in the business database of the business data source, and access paths will be provided through database files. This close association allows the agent to directly translate high-level business intents into actionable low-level instructions. Furthermore, to ensure that the agent can successfully load and utilize this guiding information, the process includes determining the file addition location corresponding to the business agent and deploying the constructed agent guiding structured data to that location. This deployment step solves the problem of the agent "not finding its way" in complex business environments, ensuring that the agent can automatically discover and load this structured data upon startup or runtime.

[0072] In summary, by identifying the database file corresponding to the business data source and the action executable file corresponding to the business operation graph structure, the agent can accurately locate and invoke the required underlying resources when executing tasks, avoiding execution deviations caused by improper resource matching. Constructing structured data to guide the agent based on the target data structure, database file, and action executable file transforms abstract business logic into a set of instructions that the agent can directly read and execute, greatly enhancing the agent's ability to handle complex business. Furthermore, by determining the file addition location corresponding to the business agent and deploying the constructed structured data, it ensures that the agent can accurately load and execute related business processing tasks, solving the problem of the agent "not finding its way" in the business environment. This mechanism of encapsulating and deploying logical structure, data source, and execution actions in an integrated manner provides the agent with clear business execution boundaries and resource access paths, thereby ensuring the accuracy and security of business processing at the underlying architecture level. Especially in highly compliant and heavily regulated scenarios such as finance, it can effectively improve the execution certainty and standardization of the agent, significantly reducing the risk of decision-making deviations and violations. See Figure 2 , Figure 2 A flowchart of a task processing method according to an embodiment of this specification is shown. The method is applied to a business intelligence agent and specifically includes the following steps.

[0073] Step S202: Based on the business processing request submitted for the target business, determine the target sub-business flowchart structure in the intelligent agent guidance structured data, wherein the intelligent agent guidance structured data is constructed using the above method.

[0074] Step S204: Generate a business task sequence based on the target sub-business process flow diagram structure, and determine the pending business task corresponding to the current task execution cycle in the business task sequence.

[0075] Step S206: In the intelligent agent-guided structured data, call at least one business action unit corresponding to the business task to be processed, and execute each business action unit in sequence as the task execution operation of the business task to be processed.

[0076] Step S208: If all business tasks included in the business task sequence have been completed, determine the task execution information as a response to the business processing request.

[0077] The task processing method provided in this embodiment is a method for forcibly guiding a business intelligence agent to perform task processing using structured data generated by the above-described structured data construction method. The process of constructing structured data guided by the intelligence agent can be found in the same or corresponding descriptions in the above embodiments, and will not be elaborated upon here. It should be noted that the descriptions of the target business, the structured data guided by the intelligence agent, and the business intelligence agent in this task processing method can all be found in the descriptions in the above embodiments. The business intelligence agent is deployed on the server side, and after the client submits a request to the server according to its needs, it invokes the business intelligence agent to execute the corresponding business task.

[0078] Specifically, a business processing request refers to a processing request submitted by a business user when participating in a target business activity. After this request is submitted to the business intelligence agent, it triggers the agent to execute the corresponding business processing task according to the agent-guided structured data and provides feedback on task execution information. The target sub-business process flow diagram structure refers to the sub-business process flow diagram structure extracted from the agent-guided structured data that relates to the processing requirements of the current business processing request. Since the agent-guided structured data contains all sub-business process flow diagram structures covering all business scenarios, the target sub-business process flow diagram structure needs to be determined through matching to drive the agent to execute business tasks related to the associated business processing request.

[0079] Accordingly, a business task sequence specifically refers to a sequence of tasks composed of multiple business tasks constructed based on the target sub-business process flowchart structure. Since the agent-guided structured data is used to force the business agent to execute relevant business processing operations, each business task in the sequence has a strict execution order, ensuring that the business agent executes each task strictly in sequence. The current task processing cycle is the execution cycle corresponding to each business task, and the pending business tasks specifically refer to the business tasks that need to be executed in the current task processing cycle.

[0080] Correspondingly, a business action unit specifically refers to the business action associated with each business task in the intelligently guided structured data. It can be understood as the atomic capability that executes each business task. For example, if a subtask of a pending business task is system login, the corresponding business action could be `submitLogin`, used to automatically complete the system login operation. Another example is if a subtask of a pending business task is confirming login status, the corresponding business action could be `confirmLoginStatus`, used to determine the current account's login status. Executing these action units allows the business intelligence agent to automatically complete the pending tasks. It should be noted that in different business scenarios, a pending business task can contain different numbers of business action units. Each business action unit can be understood as the execution operation of a sub-business task within the pending business task. The pending business task is considered complete when all sub-tasks corresponding to multiple business action units are executed.

[0081] Correspondingly, task execution information specifically refers to the execution result information obtained when all business tasks in the business task sequence have been completed. It is used to feed back to the end side that submitted the business processing request to determine the operation result completed by the intelligent agent.

[0082] The task processing method provided in this embodiment can be applied to any business scenario, such as performing writing tasks through a business intelligence agent, performing financial management tasks through a business intelligence agent, performing intelligent customer service tasks through a business intelligence agent, or performing R&D tasks through a business intelligence agent. Specific application scenarios can be set according to actual needs, and this embodiment does not impose any limitations here.

[0083] Based on this, by combining the agent's guided structured data with the generated business task sequence in a predefined manner, and introducing a decoupled execution mechanism for business action units, the behavioral path of the agent in complex business processes is standardized. Specifically, based on the business processing request submitted for the target business, the target sub-business process diagram structure can be determined in the agent's guided structured data. This enables the agent to accurately locate compliant business logic branches through a pre-built structured data as a navigation framework. Since the agent's guided structured data encapsulates complete business rules and constraints, the agent does not need to infer the business path itself when processing specific business requests, thus effectively avoiding the logical illusion problem caused by blind searching. Furthermore, a business task sequence is generated based on the target sub-business process diagram structure, and the pending business tasks corresponding to the current task execution cycle are determined in the business task sequence, thereby forcibly decomposing the macro-business process into ordered serialized task units. Given that business processes in general business scenarios have strict temporal dependencies, by clarifying the task execution cycle, it is ensured that the agent strictly follows the preset logical order, preventing skipping or invalid loops in long-process business.

[0084] Furthermore, the agent invokes at least one business action unit corresponding to the business task to be processed within the structured data guided by the intelligent agent, and executes each business action unit sequentially as the task execution operation for the business task to be processed. This achieves deep decoupling between business actions and the agent's execution logic. Each business action unit adheres to the structured data specifications, and its execution scope is limited to the constraints defined by the structured data, thus ensuring the compliance of the operation. For example, when performing operations involving sensitive data, the business action unit automatically triggers access control checks to ensure that the intelligent agent can only access data entities within the authorized scope. Finally, once all business tasks in the business task sequence have been completed, the task execution information can be determined as a response to the business processing request. This closed-loop mechanism, through real-time tracking and verification of task status, ensures that the business intelligent agent strictly executes tasks according to the structured data, thereby effectively saving resource consumption while ensuring compliance.

[0085] For example, when a securities company teller is processing a securities account opening for a client of Institution A, the teller can submit a natural language description to the Agent stating, "Please assist with the institution's account opening; the client has provided the original business license." At this point, based on this request, the teller can extract the sub-business process diagram structure related to the institution's account opening process from the pre-built intelligent agent-guided structured data, such as... Figure 3The diagram shown illustrates a sub-business process flow structure containing 12 business tasks, encompassing a total of 36 business action units. When executing each business action unit, it routes to the corresponding data description graph structure based on the mapping relationships and constraints defined in the structured data, such as... Figure 3 The diagram shows that the account opening process involves a total of 7 domains, such as counter service, business management, evidence storage, compliance, acceptance, business knowledge, and operator. These are associated with 37 data tables in the database. Under this structure, the business intelligence agent can perform subsequent task processing operations.

[0086] Furthermore, according to such Figure 3 The diagram shows that the business task sequence is {M1 (login to system) → M2 (institutional information entry) → M3 (collection of detailed institutional information) → M4 (collection of handler information) → M5 (collection of service information) → M6 (risk assessment) → M7 (collection of non-resident information) → M8 (account opening) → M9 (selection of custodian bank) → M10 (collection of files) → M11 (video collection) → M12 (acceptance result)}. Then, each business task can be extracted from the business task sequence and processed in this order. For example, if the current task processing cycle is the cycle corresponding to business task M1, then the business action units corresponding to business task M1 can be determined, namely submitLogin (system login) and confirmLoginStatus (confirm login status). At this time, the Agent can execute each business action in sequence. During the execution, the Agent can complete the operation according to the mapping relationship and constraint conditions specified in the structured data. After the execution of task M1 is completed, the subsequent tasks are executed in sequence until tasks M1 to M12 are all completed. Then the account opening result can be obtained, and the account opening result can be fed back to the teller.

[0087] In summary, driven by structured data guided by intelligent agents, business intelligent agents can strictly process different business processing requests according to the corresponding business processes. This allows the intelligent agents to completely eliminate model illusions and avoid repeatedly executing tasks, thereby effectively saving computing resources and outputting more accurate execution results in different business scenarios, effectively meeting the usage needs of different business scenarios.

[0088] Furthermore, when determining the target sub-business flowchart structure in the agent-guided structured data based on business processing requests, it is difficult to directly and accurately locate the business logic path that perfectly matches the current request from massive amounts of structured data. Therefore, in this embodiment, determining the target sub-business flowchart structure in the agent-guided structured data based on the business processing request submitted for the target business includes: The system receives a business processing request submitted for a target business, parses the business processing request to obtain business processing information, determines the target business process information corresponding to the business processing information, and determines the target sub-business process diagram structure in the intelligent agent's guided structured data according to the target business process information.

[0089] Specifically, business processing requests can be submitted in various forms. For example, they can receive structured request data from other systems via an application programming interface (API), or receive natural language text or preset options input by the user through a user interface. They can also receive asynchronous business event notifications through a message queue service. The submitted business processing requests can be converted into standardized information that the agent can understand and process. The parsing process can employ various techniques. For example, for natural language requests, Natural Language Processing (NLP) technology can be used for entity recognition, intent recognition, and semantic analysis to extract key business elements and operational intents. For structured requests, the required information can be extracted directly through data field mapping or pattern matching. Through parsing, business processing information, including business type, operation object, and key parameters, can be obtained, providing accurate input for subsequent business process matching. Target process information refers to the business process description information corresponding to the current business processing request. Based on the target business process information, the corresponding sub-business process diagram structure can be accurately located and extracted from the structured data. In practice, the agent can perform path traversal or node search within the structured data to quickly and accurately locate the specific execution path related to the current business request, i.e., the target sub-business process diagram structure. This avoids the agent blindly exploring complex business logic, ensuring the accuracy and efficiency of business processing.

[0090] Based on this, by introducing a sophisticated request parsing and process mapping mechanism, the agent can efficiently and accurately understand and execute complex business processing tasks. When the agent receives a business processing request for a specific target business, it first performs in-depth parsing of the request, transforming the raw, potentially unstructured request content into standardized business processing information. On this basis, the agent uses the obtained business processing information to determine the target business process information that best matches this information within a pre-defined business knowledge system. This allows it to plan the most suitable route based on the user's destination. It is precisely this precise matching that prevents the agent from getting lost in the vast business logic. Subsequently, based on the determined target business process information, the agent accurately locates and extracts the corresponding target sub-business process diagram structure from the pre-built agent-guided structured data. The agent-guided structured data, acting as the agent's "business map," contains the complete logic and data dependencies of business operations. Guided by the target business process information, the agent can directly route to the correct business path, thus avoiding the problems of divergent thinking and logical illusions that traditional agents often encounter when facing complex business scenarios. Compared to the basic solution where the agent directly determines the target sub-business process flow structure from the structured data guided by the agent, this mechanism adds a pre-process request parsing and process information determination step. This allows the agent to have a clearer understanding and more accurate positioning of the business intent before entering the process flow structure, which greatly improves the accuracy and efficiency of subsequent task execution and effectively solves the problem that the agent has difficulty in accurately locating the business logic path from massive amounts of structured data.

[0091] For example, a financial agent needs to process a business request for "a user to apply for a credit card." First, the agent receives the business processing request submitted by the user. This request might contain fields such as "Business Type: Credit Card Application," "Customer Name: A," and "ID Number: XXXX." Next, the agent parses the received request. If the request is a structured form, the agent can directly extract the values ​​of each field as business processing information. If the request is in natural language, the agent can use a Natural Language Understanding (NLU) model to identify "Apply for a credit card" as the business intent, "A" as the customer entity, and "ID number" as a key parameter, thus obtaining business processing information containing these elements. Then, based on the obtained business processing information, such as "Business Intent: Apply for a Credit Card," the agent searches for the corresponding target business process information in a pre-defined business rule base. Finally, according to the target business process information of "Credit Card Application Process," the agent determines the corresponding target sub-business process diagram structure in the agent-guided structured data and uses this to guide subsequent business processing tasks.

[0092] In summary, by receiving and parsing business processing requests, the intelligent agent can transform raw and diverse requests into unified and standardized business processing information, laying the foundation for subsequent accurate matching. Furthermore, by determining the target business process information corresponding to the business processing information, the intelligent agent can establish a clear association between business requests and predefined business logic, avoiding blind searching and decision-making biases in complex business scenarios.

[0093] Furthermore, the lack of effective control over invocation permissions when intelligent agents execute tasks can easily lead to unauthorized invocations or erroneous operations, resulting in business logic chaos or security risks. To address this, in this embodiment, the step of invoking at least one business action unit corresponding to the business task to be processed from the structured data guided by the intelligent agent includes: The action pointer corresponding to the business task to be processed is determined based on the structured data guiding the agent; at least one business action unit corresponding to the business task to be processed is called in the structured data guiding the agent according to the action pointer; wherein, the action pointer is used to constrain the calling permission of the business agent to call the business action unit.

[0094] Specifically, an action pointer is a reference or identifier in the structured data guiding an agent, used to indicate the business action unit associated with the business task to be processed. Its core function is that it not only points to a specific business action unit, but more importantly, it embeds or associates with the permission constraints that allow the business agent to invoke that business action unit. An action pointer can be a structure containing a business action unit ID and a list of permissions. When a business agent attempts to invoke a business action unit, the system first obtains its associated permission information through the action pointer and compares it with the current business agent's permissions. Only if the permissions match is the invocation allowed. Furthermore, an action pointer can also have a `hasAction` attribute array to ensure it has valid invocation permissions.

[0095] Based on this, when a business intelligence agent executes a task processing method, after determining the business task to be processed, it first searches for and identifies the action pointer associated with that task in the agent's guiding structured data. This action pointer is a key component of the agent's guiding structured data; it not only provides the location information of the business action unit but, more importantly, embeds or associates the permission constraints for the business intelligence agent to call that action unit. Before attempting to call a business action unit, the business intelligence agent must verify its permissions through this action pointer. Only when the business intelligence agent possesses the calling permissions specified by the action pointer will the system call and execute the corresponding business action unit in the agent's guiding structured data according to the action pointer's guidance. This mechanism ensures that the business intelligence agent's operational behavior is always strictly controlled by preset permissions when executing business tasks, effectively preventing unauthorized operations and erroneous operations. By tightly integrating permission constraints with action pointers, this solution further enhances the fine-grained control over the behavior of business intelligence agents based on the structured business processes and data dependencies provided by the agent's guiding structured data, enabling them to execute tasks more compliantly and securely in complex business scenarios.

[0096] Continuing with the previous example, after task M1 is completed, the business agent reads task M2 from the sequence and determines the eight business actions corresponding to task M2 through the hasAction pointer. These actions are: uploadBusinessLicense, ocrBusinessLicense (OCR recognition of business license), fillOrgInfo (filling in institution information with OCR results), confirmOrgInfo (user confirmation and supplementation), generateClientNo (generating client number), generateFundAccount (generating fund account), collectProfessionalInvestor (collecting professional investor information), and saveOrgInfo (submitting and saving institution information). Based on this, the agent will execute each business action sequentially; for example: Execute uploadBusinessLicense: The teller provides a picture of the business license, and the system reads the Action rule "When providing a license picture, OCR automatic extraction is given priority", which automatically triggers the subsequent OCRBusinessLicense.

[0097] Execute OCRBusinessLicense (Business License OCR Recognition): Call the "Business License OCR Interface" tool, and write the recognition result into the BusinessLicenseInfo (Business License Information) node of the data description graph structure. At this time, the Action rule "OCR results can only be written to the draft field first" takes effect, and the system marks the data status as "Pending Confirmation".

[0098] Execute fillOrgInfo (fill organization information) → confirmOrgInfo (user confirmation): Based on the prerequisite relationship of the dependsOn edge, the OCR result must be confirmed by the teller before it can be written into the official data. After confirmation, the system will persist the data to the business_license_infos (business license information table) of the DataSource.

[0099] Executing generateClientNo (generate customer ID) → generateFundAccount (generate fund account): The rule "customer ID must be generated first, then fund account" is enforced by the dependsOn edge, calling the "serial number generation interface" tool twice, and writing the output into the CustomerAccountInfo (customer account information) and FundAccountInfo (fund account information) nodes of the data description graph structure respectively. For example... Figure 4 As shown, when executing generateClientNo (generate customer number) → generateFundAccount (generate fund account), the corresponding sub-business task is to generate customer number. According to the mapping relationship between the graph structure and the database, the work unit {generate customer number} can be determined first. Its description information is specified in the structured data (call the serial number generation interface (serial_type=3) to generate customer number and write it into the acceptance form), input (operator field), output (fund account information), execution rules (customer number (3) must be generated first, and then fund account (4) is generated. The number is automatically incremented each time it is called. Do not call repeatedly to waste numbers. If operator_on / serial_type is missing, report 97973 / 120107). At the same time, determine the node {fund account information} in the associated data description graph structure. Its corresponding data is in the database {fund account information}. The corresponding task can be executed by using this link structure.

[0100] Execute saveOrgInfo (Institutional Information Submission and Saving): Call the "Acceptance Order Temporary Storage Interface" tool to submit and save all data in this module. The dependsOn side ensures that confirmOrgInfo (User Confirmation), generateFundAccount (Generate Fund Account), and collectProfessionalInvestor (Collect Professional Investor Information) are all completed before allowing the save to be executed.

[0101] In summary, the above-described procedures ensure that the operational behavior of the business intelligence agent is always strictly controlled by preset permissions, effectively preventing unauthorized calls or erroneous operations, thereby guaranteeing the compliance and security of business processes. Especially in highly compliant scenarios such as finance, this solution can significantly reduce the risks caused by agent decision-making biases or unauthorized operations, improving the reliability and controllability of business processing.

[0102] Furthermore, when handling complex business processes, intelligent agents may face situations involving concurrent multitasking or interruptions in the execution of long-running tasks. This can easily lead to the agent blindly starting the next task before the current task is completed, resulting in problems such as chaotic business logic, data processing conflicts, or inconsistent execution states. Therefore, in this embodiment, after the sequential execution of each business action unit step, the process further includes: If the pending business task is not completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle is refused to be executed, and task interception information is generated; if the pending business task is completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle is taken as the pending business task, and the step of calling at least one business action unit corresponding to the pending business task in the intelligent agent guiding structured data is executed.

[0103] Specifically, "pending business task not completed" means that the currently being processed business task has not yet reached its intended completion state. "Starting the next task processing cycle" refers to the system or agent attempting to begin processing the next logically or temporally consecutive business task. "Refusing to execute the next task processing cycle" means preventing the actual processing flow of the next task from starting. "Generating task interception information" indicates that the system records and may notify relevant parties that a task has been blocked from execution. This can be achieved by: recording the interception event, the identifier of the intercepted task, the reason for the interception, and the timestamp in the system log for subsequent auditing and troubleshooting; or sending an alert notification to the agent's control module or operations personnel, such as via message queues, email, or instant messaging tools, to indicate that the task has been intercepted.

[0104] Based on this, a state verification mechanism for task execution cycles is introduced to achieve fine-grained control over the business execution process of the intelligent agent. After the intelligent agent completes the business action units of the current pending business task in sequence, the system determines whether to allow the start of the next task processing cycle based on the completion status of the current task. Specifically, when it is detected that the current pending business task has not been completed, but the system or external event attempts to start the business task corresponding to the next task processing cycle, the system will actively refuse the start of the next task and generate corresponding task interception information. This mechanism ensures the atomicity and logical coherence of the business process, effectively preventing the intelligent agent from performing skipped operations or logical jumps due to concurrency or state judgment errors in complex business scenarios. For example, in a multi-step financial transaction process, if the fund transfer action has not been completed, the system will prevent the subsequent transaction confirmation or notification action from being initiated, thereby avoiding business risks caused by inconsistent states. Conversely, when it is detected that the current pending business task has been successfully completed, and the system or external event attempts to start the business task corresponding to the next task processing cycle, this solution allows the task to be started. At this point, the system will formally designate the business task corresponding to the next task processing cycle as a new business task to be processed, and then execute the step of calling at least one business action unit corresponding to the new business task to be processed in the intelligent agent's guided structured data.

[0105] Continuing with the previous example, once business task M2 is completed, each of the subsequent business tasks M3 through M12 will be executed sequentially. During this process, for example, when entering M6 (risk assessment), the rule for scoreRiskLevel (risk level scoring) – "score can only be calculated after all questions are answered" – is automatically verified by the system before execution. If this rule is not met, execution is blocked. The risk level result is written to the RiskResult node in the data description graph structure and linked to AcceptanceBill (acceptance order) via the evaluates edge, directly affecting the range of accounts that can be opened in the subsequent M8. Suppose that the Agent attempts to skip directly to configureTradingAccounts in M8 before the risk assessment in M6 is completed. In this case, it can be detected that the dependsOn precondition is not met and the RiskResult node has no valid data, so execution can be refused and an interception message can be returned. This mechanism can intercept arbitrary jumps at any stage. After M11 is completed and M12 is entered, `checkAcceptanceInfo` performs a final check on the entire process. Once all required data from M1 to M11 is confirmed to be complete, `submitAcceptance` calls the "Acceptance Form Submission Interface" to formally submit the application, returning the acceptance number, and the process ends. Throughout the entire process, the Agent can only see the Action corresponding to each task individually and cannot access the actions of other tasks. Tools are all attached to the leaf nodes of each Action; the task itself does not hold any tools, achieving dual-axis separation.

[0106] In summary, combining structured data with agent guidance further enhances the stability of agents when handling long-process business scenarios. The interception mechanism provides a clear feedback path for anomaly handling, significantly improving the reliability and security of agent task execution, enabling agents to complete business processing tasks more accurately and safely.

[0107] In practical implementation, when an agent is processing a predetermined business sequence, it may suddenly receive a new business processing request. Existing execution logic often lacks a mechanism to determine if the new task conflicts with the current business task sequence. This results in the agent being unable to flexibly handle sudden business requests, or interrupting the logical continuity of the original business when processing the new request, making it difficult to dynamically insert the new request while maintaining the integrity of the original task sequence. Therefore, in this embodiment, after the sequential execution of each business action unit step, the process further includes: If a business task in the business task sequence has not been completed and a new business processing request for the target business is received, it is determined whether the new business task corresponding to the new business processing request belongs to the business task sequence. If not, at least one new business action unit corresponding to the new business task is called in the intelligent agent guided structured data, and each new business action unit is executed sequentially as the task execution operation of the new business task. If the new business task is completed, the business tasks in the business task sequence continue to be executed.

[0108] Specifically, determining whether a newly received business processing request corresponds to a new business task belongs to a business task sequence aims to identify the correlation or conflict between the newly received business processing request and the currently executing business task sequence. This determination mechanism prevents the agent from blindly processing new tasks, thereby interrupting or confusing the original business logic. This determination can be made by comparing the identifier, business type, or associated business process of the new business task with existing task information in the current business task sequence. For example, it checks whether the business ID of the new task already exists in the current business task sequence, or whether its business process is consistent with the current sequence's business process. Furthermore, this determination can also be performed using preset business rules or strategies. For example, it can define that certain types of new tasks cannot interrupt the queue under any circumstances, or that certain types of new tasks are considered to "belong" to the current sequence if they belong to the same business module as a task in the current sequence.

[0109] Correspondingly, invoking at least one new business action unit corresponding to a new business task in the agent-guided structured data means that the business agent, based on the requirements of the new business task, searches for and activates the corresponding business action unit from the pre-built agent-guided structured data. The purpose of this invocation is to ensure that the execution of the new task remains within a controlled business framework, preventing the agent from exceeding its authority or performing unauthorized operations. The agent can retrieve predefined action unit mapping relationships in the agent-guided structured data based on the type or identifier of the new business task, thereby locating the specific new business action unit. Alternatively, the agent can match and invoke the action unit by parsing the semantic description of the new business task and combining it with the functional description of the action unit in the agent-guided structured data.

[0110] Correspondingly, executing each new business action unit sequentially means that the agent completes each business action unit included in the new business task one by one according to a preset order or logical flow. The purpose of this execution method is to ensure the integrity and correctness of the new business task and avoid business processing anomalies due to incorrect action execution order. The agent can strictly follow the order of the new business action units defined in the structured data guided by the agent to call and execute them. Alternatively, the agent can dynamically determine the execution order based on the dependencies between the new business action units. For example, if action unit B depends on the output of action unit A, the agent will execute B only after A has been executed.

[0111] Based on this, by introducing a dynamic task conflict detection and queue-jumping execution mechanism, the flexibility and robustness of the intelligent agent in complex business scenarios are significantly enhanced. When the business intelligent agent is executing a sequence of business tasks, if a new business processing request is received, the system first determines whether the new business task belongs to the currently executing sequence. This determination mechanism enables the intelligent agent to identify the correlation between the new task and the existing process, avoiding business logic confusion that may result from blind processing. If the determination result is that it does not belong to the current sequence, it indicates that the new task is an independent queue-jumping task that needs to be processed immediately. At this time, the intelligent agent will not interrupt or reset the original business process, but will use pre-built intelligent agent-guided structured data to accurately call and execute the business action units corresponding to the new business task in sequence. The intelligent agent-guided structured data provides the intelligent agent with clear business logic, action definitions, and execution constraints, ensuring that even queue-jumping tasks strictly follow business specifications, effectively preventing the intelligent agent from generating logical illusions or overstepping authority when handling sudden tasks. After the new business task is completed, the intelligent agent can automatically and seamlessly resume the original business task sequence and continue execution from the point of interruption. This mechanism ensures the continuity and consistency of business processing, enabling intelligent agents to handle both long-running business processes and unexpected, unforeseen tasks.

[0112] Using the previous example, during the execution of tasks M1 to M12, suppose the agent receives a request from the teller stating, "During the process, the bank's relevant information is incorrect and needs to be re-uploaded." At this point, the agent determines that the task does not belong to the original tasks M1 to M12. The agent can then retrieve the relevant information from the structured data guided by the intelligent agent and select the corresponding business action unit to directly execute the new task. For example, it can execute business action units such as "collect professional investor information," "collect detailed institutional information," "collect non-resident information," and "select custodian bank." After the new task is completed, the agent can continue to execute the tasks that have not yet been completed.

[0113] In summary, the above approach effectively addresses the lack of a dynamic handling mechanism for new business processing requests when intelligent agents are processing predetermined business sequences. This not only improves the response efficiency and adaptability of business processing but also ensures the standardization and security of business operations in highly compliant and heavily regulated scenarios, avoiding risks that may arise from decision-making biases or violations. Based on this, if a corresponding business process diagram structure exists within the structured data of the intelligent agent, the existing diagram structure can be directly selected to complete the corresponding task execution. Therefore, in this embodiment, the method further includes: The new sub-business flowchart structure corresponding to the new business processing request is determined from the structured data guided by the intelligent agent; the new business processing request is used as the business processing request, the new sub-business flowchart structure is used as the target sub-business flowchart structure, and the step of generating a business task sequence based on the target sub-business flowchart structure is executed; until the new business task corresponding to the new business processing request is completed, the business tasks contained in the business task sequence are continued to be executed.

[0114] Specifically, the structure of the new sub-business process diagram corresponding to the new business processing request is determined in the structured data guided by the intelligent agent, in order to identify and locate the specific business process path related to the newly introduced business processing request.

[0115] Based on this, a dynamic task insertion and nested execution mechanism is introduced, enabling the agent to accurately identify and process new business processes while ensuring the continuity of existing business processes. This mechanism ensures that the execution basis of new tasks remains consistent with the global business logic by accurately locating the new sub-business flowchart structure corresponding to the new business processing request in the agent's guiding structured data. Specifically, when the agent receives a new business processing request while executing a predetermined business task sequence, it does not simply interrupt or ignore the original tasks. Instead, it first identifies and loads the business process path associated with the new request—the new sub-business flowchart structure—in the agent's guiding structured data. Subsequently, the system temporarily uses the new business processing request and its corresponding new sub-business flowchart structure as the current business processing context and utilizes existing task sequence generation logic to generate an independent business task sequence for the new business. This process allows the new task to be incorporated into the agent's task scheduling framework in a standardized manner, ensuring the standardization and controllability of its execution. During the execution of the new business task sequence, the execution of the original business task sequence is paused. Only after all newly added business tasks corresponding to a new business processing request have been completed will the system resume and continue executing the remaining tasks in the original business task sequence. This design effectively solves the problem of logical discontinuity that may occur when inserting new tasks into long-flow business processes, enabling the agent to maintain the orderly progress of the original business process while handling sudden business events. In this way, the agent can flexibly respond to and handle dynamically changing business needs based on its guided structured data, while ensuring the continuity and integrity of business processing.

[0116] In summary, by determining the structure of the new sub-business process flowchart within the structured data guided by the intelligent agent, it is possible to ensure that the execution of new tasks remains consistent with the global business logic, avoiding divergent thinking and logical illusions. Transforming new business processing requests into target business processing requests and regenerating task sequences allows new tasks to be systematically incorporated into the agent's task scheduling framework, guaranteeing the standardization and controllability of execution. Furthermore, by setting the logic of returning to the original task sequence only after the new business task is completed, the atomicity and integrity of business processing are ensured, effectively resolving the potential logical discontinuity problem when inserting new tasks into long-flow business processes. This significantly enhances the resilience and response efficiency of the intelligent agent in complex and ever-changing financial scenarios, improving the flexibility of business processing and user experience. The following is in conjunction with the appendix Figure 5 Taking the application of the task processing method provided in this specification in an e-commerce intelligent customer service scenario as an example, the task processing method will be further explained. Figure 5 The present specification shows a flowchart of a task processing method according to an embodiment, which includes the following steps.

[0117] Step S502: Determine business logic information and data dependency information based on the global business information of the target business.

[0118] Step S504: The business logic information is hierarchically divided and structurally transformed according to the preset hierarchical division information and operation constraint information to obtain the local business process diagram structure and the local business action diagram structure. The local business process diagram structure includes at least one sub-business process diagram structure corresponding to a business process.

[0119] Step S506: Generate the business operation diagram structure corresponding to the target business based on the local business process diagram structure and the local business action diagram structure.

[0120] Step S508: Extract business entity information and business attribute information from the data dependency information according to the preset object description information, and construct the data description graph structure corresponding to the target business based on the business entity information and business attribute information.

[0121] Step S510: Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, and establish the access mapping relationship between the business data source of the target business and the data description graph structure. Generate the business topology structure based on the mapping relationship.

[0122] Step S512: Add cross-level restriction constraints and tool mounting constraints to the business operation graph structure in the business topology, and add execution direction constraints and execution permission constraints to the business topology. Generate the initial data structure based on the results of adding the constraints.

[0123] Step S514: Process the sub-data structures associated with the business operation graph structure in the initial data structure to obtain the business operation structure. The business operation structure includes an action sub-data structure and an orchestration sub-data structure. The action sub-data structure is associated with business actions, and the orchestration sub-data structure is associated with business scenarios.

[0124] Step S516: Process the sub-data structures of the associated data description graph structure in the initial data structure to obtain the data description structure, and process the sub-data structures of the associated business data sources in the initial data structure to obtain the data source registration structure.

[0125] Step S518: Based on the business operation structure, data description structure, and data source registration structure, construct the target data structure corresponding to the target business.

[0126] Step S520: Determine the database file corresponding to the business data source and the action executable file corresponding to the business operation graph structure. Based on the target data structure, database file and action executable file, construct the intelligent agent guidance structured data corresponding to the target business.

[0127] Step S522: Upon receiving a business processing request submitted for the target business, the business processing request is parsed to obtain business processing information.

[0128] Step S524: Determine the target business process information corresponding to the business processing information, and determine the target sub-business process diagram structure in the intelligent agent-guided structured data according to the target business process information.

[0129] Step S526: Generate a business task sequence based on the target sub-business process diagram structure, and determine the pending business tasks corresponding to the current task execution cycle in the business task sequence.

[0130] Step S528: Determine the action pointer corresponding to the business task to be processed based on the intelligent agent guiding structured data, and call at least one business action unit corresponding to the business task to be processed in the intelligent agent guiding structured data according to the action pointer.

[0131] Step S530: Execute each business action unit sequentially as a task execution operation for the business task to be processed.

[0132] Step S532: If all business tasks in the business task sequence have been completed, determine the task execution information as a response to the business processing request.

[0133] In summary, driven by structured data guided by intelligent agents, business intelligent agents can strictly process different business processing requests according to the corresponding business processes. This allows the intelligent agents to completely eliminate model illusions and avoid repeatedly executing tasks, thereby effectively saving computing resources and outputting more accurate execution results in different business scenarios, effectively meeting the usage needs of different business scenarios.

[0134] Corresponding to the above method embodiments, this specification also provides embodiments of a task processing system. Figure 6 A schematic diagram of the structure of a task processing system according to one embodiment of this specification is shown. Figure 6 As shown, the task processing system 600 includes an operation terminal 610 and a processing terminal 620, including: The operation terminal 610 is used to construct a business processing request in response to the business processing information submitted by the business user for the target business, and send the business processing request to the processing terminal; The processing terminal 620 is configured to: invoke a business agent to determine a target sub-business process flow diagram structure in the agent-guided structured data according to the business processing request, wherein the agent-guided structured data is constructed using the method described above; generate a business task sequence based on the target sub-business process flow diagram structure, and determine the pending business task corresponding to the current task execution cycle in the business task sequence; invoke at least one business action unit corresponding to the pending business task in the agent-guided structured data, and execute each business action unit sequentially as the task execution operation of the pending business task; and, if all business tasks included in the business task sequence are completed, determine the task execution information and feed it back to the operation terminal.

[0135] In an optional embodiment, determining the target sub-business flowchart structure in the agent-guided structured data based on the business processing request submitted for the target business includes: Receive a business processing request submitted for a target business, parse the business processing request, and obtain business processing information; Determine the target business process information corresponding to the business processing information, and determine the target sub-business process diagram structure in the intelligent agent-guided structured data according to the target business process information.

[0136] In an optional embodiment, the step of invoking at least one business action unit corresponding to the business task to be processed in the intelligent agent-guided structured data includes: The action pointer corresponding to the business task to be processed is determined based on the structured data guided by the intelligent agent. According to the action pointer, at least one business action unit corresponding to the business task to be processed is invoked in the intelligent agent's guided structured data; The action pointer is used to constrain the calling permissions of the business intelligence agent to invoke business action units.

[0137] In an optional embodiment, after the sequential execution of each business action unit step, the method further includes: If the pending business task is not completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle will be refused to be executed, and task interception information will be generated. When the pending business task is completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle is taken as the pending business task, and the step of calling at least one business action unit corresponding to the pending business task in the intelligent agent guiding structured data is executed.

[0138] In an optional embodiment, after the sequential execution of each business action unit step, the method further includes: If a business task in the business task sequence has not been completed and a new business processing request is received for the target business, it is determined whether the new business task corresponding to the new business processing request belongs to the business task sequence. If not, at least one new business action unit corresponding to the new business task is called in the intelligent agent-guided structured data, and each new business action unit is executed in sequence as the task execution operation of the new business task. If the newly added business task is completed, the business tasks included in the business task sequence will continue to be executed.

[0139] In an optional embodiment, the method further includes: The structure of the new sub-business flowchart corresponding to the new business processing request is determined from the structured data guided by the intelligent agent. The newly added business processing request is used as the business processing request, the newly added sub-business process flow diagram structure is used as the target sub-business process flow diagram structure, and the step of generating a business task sequence based on the target sub-business process flow diagram structure is executed. The business tasks contained in the business task sequence will continue to be executed until the new business task corresponding to the new business processing request is completed.

[0140] The above is an illustrative scheme of a task processing system according to this embodiment. It should be noted that the technical solution of this task processing system and the technical solution of the task processing method described above belong to the same concept. For details not described in detail in the technical solution of the task processing system, please refer to the description of the technical solution of the task processing method described above.

[0141] Corresponding to the above method embodiments, this specification also provides embodiments of a structured data construction apparatus. Figure 7 A schematic diagram of a structured data construction apparatus provided in one embodiment of this specification is shown. Figure 7 As shown, the device includes: The information determination module 702 is configured to determine business logic information and data dependency information based on the global business information of the target business. The graph structure construction module 704 is configured to construct a business operation graph structure based on the business logic information, and to construct a data description graph structure based on the data dependency information. The relationship establishment module 706 is configured to establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and to configure execution constraints for the business operation graph structure, the data description graph structure and the business data source to obtain the target data structure; The data construction module 708 is configured to construct intelligent agent-guided structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure.

[0142] In an optional embodiment, constructing the business operation graph structure based on the business logic information includes: The business logic information is hierarchically divided and structurally transformed according to the preset hierarchical division information and operation constraint information to obtain a local business process diagram structure and a local business action diagram structure. The local business process diagram structure includes at least one sub-business process diagram structure corresponding to a business process. Based on the local business process diagram structure and the local business action diagram structure, a business operation diagram structure corresponding to the target business is generated.

[0143] In an optional embodiment, constructing the data description graph structure based on the data dependency information includes: Based on the preset object description information, business entity information and business attribute information are extracted from the data dependency information; Based on the business entity information and the business attribute information, construct the data description graph structure corresponding to the target business.

[0144] In an optional embodiment, establishing the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configuring execution constraints for the business operation graph structure, the data description graph structure, and the business data source to obtain the target data structure includes: Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, and establish the access mapping relationship between the business data source of the target business and the data description graph structure. Generate a business topology structure based on the mapping relationship results. Add cross-level restriction constraints and tool mounting constraints to the business operation graph structure in the business topology, and add execution direction constraints and execution permission constraints to the business topology. Generate the target data structure based on the results of adding the constraints.

[0145] In an optional embodiment, the step of generating the target data structure by adding results based on constraints includes: Generate the initial data structure by adding results based on the constraints; The sub-data structures associated with the business operation graph structure in the initial data structure are processed to obtain the business operation structure, wherein the business operation structure includes an action sub-data structure and an orchestration sub-data structure, the action sub-data structure is associated with business actions, and the orchestration sub-data structure is associated with business scenarios; The sub-data structures associated with the data description graph structure in the initial data structure are processed to obtain the data description structure, and the sub-data structures associated with the business data source in the initial data structure are processed to obtain the data source registration structure. Based on the business operation structure, the data description structure, and the data source registration structure, construct the target data structure corresponding to the target business.

[0146] In an optional embodiment, the step of constructing the agent-guided structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure includes: Determine the database file corresponding to the business data source and the action executable file corresponding to the business operation graph structure; Based on the target data structure, the database file, and the action executable file, construct the intelligent agent guidance structured data corresponding to the target service; The step of constructing the intelligent agent-guided structured data execution step corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure further includes: Determine the file addition location corresponding to the business intelligence agent of the target business, and add the agent's guiding structured data to the file addition location; wherein, the agent's guiding structured data is used by the business intelligence agent to execute business processing tasks associated with the target business.

[0147] The above is an illustrative scheme of a structured data construction apparatus according to this embodiment. It should be noted that the technical solution of this structured data construction apparatus and the technical solution of the structured data construction method described above belong to the same concept. For details not described in detail in the technical solution of the structured data construction apparatus, please refer to the description of the technical solution of the structured data construction method described above.

[0148] Corresponding to the above method embodiments, this specification also provides embodiments of a task processing device. Figure 8 A schematic diagram of a task processing apparatus according to one embodiment of this specification is shown. Figure 8 As shown, this device is applied to a business intelligence agent and includes: The determination module 802 is configured to determine the target sub-business flowchart structure in the agent-guided structured data based on the business processing request submitted for the target business, wherein the agent-guided structured data is constructed using the method described above. The generation module 804 is configured to generate a business task sequence based on the target sub-business process diagram structure, and determine the pending business task corresponding to the current task execution cycle in the business task sequence; The execution module 806 is configured to call at least one business action unit corresponding to the business task to be processed in the intelligent agent-guided structured data, and execute each business action unit in sequence as a task execution operation of the business task to be processed. The determination module 808 is configured to determine task execution information as a response to the business processing request when all business tasks included in the business task sequence have been completed.

[0149] In an optional embodiment, determining the target sub-business flowchart structure in the agent-guided structured data based on the business processing request submitted for the target business includes: Receive a business processing request submitted for a target business, parse the business processing request, and obtain business processing information; Determine the target business process information corresponding to the business processing information, and determine the target sub-business process diagram structure in the intelligent agent-guided structured data according to the target business process information.

[0150] In an optional embodiment, the step of invoking at least one business action unit corresponding to the business task to be processed in the intelligent agent-guided structured data includes: The action pointer corresponding to the business task to be processed is determined based on the structured data guided by the intelligent agent. According to the action pointer, at least one business action unit corresponding to the business task to be processed is invoked in the intelligent agent's guided structured data; The action pointer is used to constrain the calling permissions of the business intelligence agent to invoke business action units.

[0151] In an optional embodiment, after the sequential execution of each business action unit step, the method further includes: If the pending business task is not completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle will be refused to be executed, and task interception information will be generated. When the pending business task is completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle is taken as the pending business task, and the step of calling at least one business action unit corresponding to the pending business task in the intelligent agent guiding structured data is executed.

[0152] In an optional embodiment, after the sequential execution of each business action unit step, the method further includes: If a business task in the business task sequence has not been completed and a new business processing request is received for the target business, it is determined whether the new business task corresponding to the new business processing request belongs to the business task sequence. If not, at least one new business action unit corresponding to the new business task is called in the intelligent agent-guided structured data, and each new business action unit is executed in sequence as the task execution operation of the new business task. If the newly added business task is completed, the business tasks included in the business task sequence will continue to be executed.

[0153] In an optional embodiment, the method further includes: The structure of the new sub-business flowchart corresponding to the new business processing request is determined from the structured data guided by the intelligent agent. The newly added business processing request is used as the business processing request, the newly added sub-business process flow diagram structure is used as the target sub-business process flow diagram structure, and the step of generating a business task sequence based on the target sub-business process flow diagram structure is executed. The business tasks contained in the business task sequence will continue to be executed until the new business task corresponding to the new business processing request is completed.

[0154] The above is an illustrative scheme of a task processing device according to this embodiment. It should be noted that the technical solution of this task processing device and the technical solution of the task processing method described above belong to the same concept. For details not described in detail in the technical solution of the task processing device, please refer to the description of the technical solution of the task processing method described above.

[0155] Figure 9 A structural block diagram of a computing device 900 according to one embodiment of this specification is shown. The components of the computing device 900 include, but are not limited to, a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0156] The computing device 900 also includes an access device 940, which enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 940 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0157] In one embodiment of this specification, the above-described components of the computing device 900 and Figure 9 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 9 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0158] The computing device 900 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 900 can also be a mobile or stationary server.

[0159] The processor 920 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described structured data construction method or task processing method.

[0160] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device belongs to the same concept as the technical solution of the structured data construction method or task processing method described above. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the structured data construction method or task processing method described above.

[0161] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described structured data construction method or task processing method.

[0162] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the structured data construction method or task processing method described above. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the structured data construction method or task processing method described above.

[0163] An embodiment of this specification also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described structured data construction method or task processing method.

[0164] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product belongs to the same concept as the technical solution of the structured data construction method or task processing method described above. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the structured data construction method or task processing method described above.

[0165] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0166] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0167] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0168] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0169] The preferred embodiments disclosed above are merely illustrative of this specification. Optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described in this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification.

Claims

1. A method for constructing structured data, characterized in that, include: Based on the global business information of the target business, determine the business logic information and data dependency information; A business operation graph structure is constructed based on the business logic information, and a data description graph structure is constructed based on the data dependency information; Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configure execution constraints for the business operation graph structure, the data description graph structure and the business data source to obtain an initial data structure. Process the sub-data structures associated with the business operation graph structure, the data description graph structure and the business data source in the initial data structure to obtain the target data structure. The execution constraints include cross-level restriction constraints, tool mounting constraints, execution direction constraints and execution permission constraints. Based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure, construct the intelligent agent guidance structured data corresponding to the target business.

2. The structured data construction method according to claim 1, characterized in that, The step of constructing the business operation graph structure based on the business logic information includes: The business logic information is hierarchically divided and structurally transformed according to the preset hierarchical division information and operation constraint information to obtain a local business process diagram structure and a local business action diagram structure. The local business process diagram structure includes at least one sub-business process diagram structure corresponding to a business process. Based on the local business process diagram structure and the local business action diagram structure, a business operation diagram structure corresponding to the target business is generated.

3. The structured data construction method according to claim 1, characterized in that, The step of constructing a data description graph structure based on the data dependency information includes: Based on the preset object description information, business entity information and business attribute information are extracted from the data dependency information; Based on the business entity information and the business attribute information, construct the data description graph structure corresponding to the target business.

4. The structured data construction method according to claim 1, characterized in that, The steps of establishing the execution mapping relationship between the business operation graph structure and the data description graph structure, the access mapping relationship between the business data source of the target business and the data description graph structure, and configuring execution constraints for the business operation graph structure, the data description graph structure, and the business data source to obtain the initial data structure include: Establish the execution mapping relationship between the business operation graph structure and the data description graph structure, and establish the access mapping relationship between the business data source of the target business and the data description graph structure. Generate a business topology structure based on the mapping relationship results. Add cross-level restriction constraints and tool mounting constraints to the business operation graph structure in the business topology, and add execution direction constraints and execution permission constraints to the business topology. Generate the initial data structure based on the results of adding the constraints.

5. The structured data construction method according to claim 4, characterized in that, The process of processing the sub-data structures associated with the business operation graph structure, the data description graph structure, and the business data source in the initial data structure to obtain the target data structure includes: The sub-data structures associated with the business operation graph structure in the initial data structure are processed to obtain the business operation structure, wherein the business operation structure includes an action sub-data structure and an orchestration sub-data structure, the action sub-data structure is associated with business actions, and the orchestration sub-data structure is associated with business scenarios; The sub-data structures associated with the data description graph structure in the initial data structure are processed to obtain the data description structure, and the sub-data structures associated with the business data source in the initial data structure are processed to obtain the data source registration structure. Based on the business operation structure, the data description structure, and the data source registration structure, construct the target data structure corresponding to the target business.

6. The structured data construction method according to any one of claims 1 to 5, characterized in that, The step of constructing intelligent agent-guided structured data corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure includes: Determine the database file corresponding to the business data source and the action executable file corresponding to the business operation graph structure; Based on the target data structure, the database file, and the action executable file, construct the intelligent agent guidance structured data corresponding to the target service; The step of constructing the intelligent agent-guided structured data execution step corresponding to the target business based on the target data structure, the business data source, and the action executable file corresponding to the business operation graph structure further includes: Determine the file addition location corresponding to the business intelligence agent of the target business, and add the agent's guiding structured data to the file addition location; wherein, the agent's guiding structured data is used by the business intelligence agent to execute business processing tasks associated with the target business.

7. A task processing method, characterized in that, Applied to business intelligence agents, including: Based on the business processing request submitted for the target business, the target sub-business flowchart structure is determined in the agent-guided structured data, wherein the agent-guided structured data is constructed by the method described in any one of claims 1 to 6; A business task sequence is generated based on the target sub-business process diagram structure, and the pending business task corresponding to the current task execution cycle is determined in the business task sequence. In the structured data guided by the intelligent agent, at least one business action unit corresponding to the business task to be processed is invoked, and each business action unit is executed in sequence as the task execution operation of the business task to be processed. If all business tasks included in the business task sequence have been completed, task execution information is determined as a response to the business processing request.

8. The task processing method according to claim 7, characterized in that, The process of determining the target sub-business flowchart structure in the intelligent agent-guided structured data based on the business processing request submitted for the target business includes: Receive a business processing request submitted for a target business, parse the business processing request, and obtain business processing information; Determine the target business process information corresponding to the business processing information, and determine the target sub-business process diagram structure in the intelligent agent-guided structured data according to the target business process information.

9. The task processing method according to claim 7, characterized in that, The step of invoking at least one business action unit corresponding to the business task to be processed in the structured data guided by the intelligent agent includes: The action pointer corresponding to the business task to be processed is determined based on the structured data guided by the intelligent agent. According to the action pointer, at least one business action unit corresponding to the business task to be processed is invoked in the intelligent agent's guided structured data; The action pointer is used to constrain the calling permissions of the business intelligence agent to invoke business action units.

10. The task processing method according to claim 7, characterized in that, After the sequential execution of each business action unit step, the process further includes: If the pending business task is not completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle will be refused to be executed, and task interception information will be generated. When the pending business task is completed and the business task corresponding to the next task processing cycle is started, the business task corresponding to the next task processing cycle is taken as the pending business task, and the step of calling at least one business action unit corresponding to the pending business task in the intelligent agent guiding structured data is executed.

11. The task processing method according to claim 7, characterized in that, After the sequential execution of each business action unit step, the process further includes: If a business task in the business task sequence has not been completed and a new business processing request is received for the target business, it is determined whether the new business task corresponding to the new business processing request belongs to the business task sequence. If not, at least one new business action unit corresponding to the new business task is called in the intelligent agent-guided structured data, and each new business action unit is executed in sequence as the task execution operation of the new business task. If the newly added business task is completed, the business tasks included in the business task sequence will continue to be executed.

12. The task processing method according to claim 11, characterized in that, The method further includes: The structure of the new sub-business flowchart corresponding to the new business processing request is determined from the structured data guided by the intelligent agent. The newly added business processing request is used as the business processing request, the newly added sub-business process flow diagram structure is used as the target sub-business process flow diagram structure, and the step of generating a business task sequence based on the target sub-business process flow diagram structure is executed. The business tasks contained in the business task sequence will continue to be executed until the new business task corresponding to the new business processing request is completed.

13. A task processing system, characterized in that, Includes the operation end and the processing end, including: The operation terminal is used to construct a business processing request in response to the business processing information submitted by the business user for the target business, and send the business processing request to the processing terminal; The processing terminal is configured to invoke a business agent to determine a target sub-business process flow diagram structure in the agent-guided structured data according to the business processing request, wherein the agent-guided structured data is constructed using the method described in any one of claims 1 to 6; generate a business task sequence based on the target sub-business process flow diagram structure, and determine the pending business task corresponding to the current task execution cycle in the business task sequence; invoke at least one business action unit corresponding to the pending business task in the agent-guided structured data, and execute each business action unit sequentially as the task execution operation of the pending business task; and, if all business tasks included in the business task sequence are completed, determine task execution information and feed the task execution information back to the operation terminal.

14. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 6 or 7 to 12.

15. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 6 or 7 to 12.

16. A computer program product, characterized in that, Includes a computer program or instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 6 or 7 to 12.

Images