Method, apparatus and electronic device for service execution

Abstract

The application discloses a method, device and electronic equipment for business execution. The method comprises: obtaining a target instruction file from a target path, wherein the target instruction file comprises a business operation instruction generated by an intelligent agent according to a preset business rule, and the target path is an interaction path between the intelligent agent and a business system; analyzing the target instruction file to obtain multi-dimensional information, wherein the multi-dimensional information is used to reflect from multiple dimensions whether the business operation instruction conforms to a preset compliance strategy of the business system; and executing the business operation instruction in a case where the multi-dimensional information indicates that the business operation instruction conforms to the preset compliance strategy. The application solves the technical problem of poor system stability in the related art that a body intelligent agent driven by a large language model executes external operations by directly calling an enterprise business system API, which has a risk of executing unauthorized system operations.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, apparatus, and electronic device for performing business operations. Background Technology

[0002] For the interaction between embodied agents based on large language models and enterprise business systems, the relevant technologies adopt a strong coupling mode of directly calling APIs or static hooks, which gives the agent high privileges to actively initiate network requests. Once the agent generates malicious instructions due to illusions or prompt word injection, it can directly trigger illegal operations, which poses a risk of executing unauthorized system operations.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method, apparatus, and electronic device for business execution, which at least solves the technical problem in the related art where embodied intelligent agents driven by large language models execute external operations by directly calling enterprise business system APIs, which poses a risk of performing unauthorized system operations and leads to poor system stability.

[0005] According to one aspect of the embodiments of this application, a method for business execution is provided, comprising: obtaining a target instruction file from a target path, wherein the target instruction file includes business operation instructions generated by an intelligent agent according to preset business rules, and the target path is an interaction path between the intelligent agent and a business system; parsing the target instruction file to obtain multi-dimensional information, wherein the multi-dimensional information is used to reflect whether the business operation instructions conform to the preset compliance policy of the business system from multiple dimensions; and executing the business operation instructions when the multi-dimensional information indicates that the business operation instructions conform to the preset compliance policy.

[0006] In some embodiments of this application, the target path includes a first path determined based on the object storage service; the first path is a virtual storage path pre-configured in the object storage service; obtaining the target instruction file from the target path includes: calling the file query interface of the object storage service and periodically detecting the instruction files in the first path to obtain a first detection result; if the first detection result indicates that a first instruction file exists in the first path, the first instruction file is used as the target instruction file, wherein the first instruction file includes files that have not been read in the first path; or, the target path includes a second path determined based on the file transfer protocol service; the second path is a local file system path on the server where the file transfer protocol service is deployed; obtaining the target instruction file from the target path includes: establishing a connection with the server corresponding to the file transfer protocol service and periodically detecting the instruction files in the second path to obtain a second detection result; if the second detection result indicates that a second instruction file exists in the second path, the second instruction file is used as the target instruction file, wherein the second instruction file includes files that have not been read in the second path.

[0007] In some embodiments of this application, the target instruction file includes an identifier for an intelligent agent and business operation parameters. Parsing the target instruction file yields multi-dimensional information, including: obtaining the operation permissions corresponding to the identifier of the intelligent agent, and comparing the target business type corresponding to the business operation instruction with the operation permissions to obtain a first comparison result, wherein the operation permissions reflect the range of business types controlled by the intelligent agent; comparing the business operation parameters with preset parameter thresholds to obtain a second comparison result; obtaining a first business context state corresponding to the business system and a second business context state on which the business operation instruction depends, and comparing the first business context state with the second business context state to obtain a third comparison result; and using the first comparison result, the second comparison result, and the third comparison result together as multi-dimensional information.

[0008] In some embodiments of this application, the method further includes: upon completion of a business operation instruction, obtaining the execution result corresponding to the business operation instruction and converting the execution result into an execution result file, wherein the execution result file has the same format as the target instruction file; if multi-dimensional information indicates that the business operation instruction does not meet the preset compliance policy, intercepting the business operation instruction and generating an interception result file, wherein the interception result file is used to instruct the intelligent agent to initiate a manual review process; and writing the execution result file or the interception result back to the target path.

[0009] According to another aspect of the embodiments of this application, another method for business execution is also provided, including: obtaining business requirements of a business system, and generating an instruction file based on the business requirements and preset business rules, wherein the instruction file includes business operation instructions; writing the instruction file into a target path, wherein the target path is the interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform business operations corresponding to the business operation instructions.

[0010] In some embodiments of this application, the target path includes one of the following: a first path determined based on the object storage service and a second path determined based on the file transfer protocol service, wherein the first path is a virtual storage path pre-configured in the object storage service, and the second path is a local file system path on the server where the file transfer protocol service is deployed; writing the instruction file to the target path includes: calling the file upload interface of the object storage service to store the instruction file in the form of an object in the first path; or, establishing a connection with the server corresponding to the file transfer protocol service to upload the instruction file to the second path in the form of a file transfer.

[0011] In some embodiments of this application, the method further includes: receiving an execution result file from a target path, wherein the execution result file includes the result obtained by the business system when the business operation instruction meets the preset compliance policy; and generating a decision instruction based on the execution result file.

[0012] According to another aspect of the embodiments of this application, a business execution apparatus is also provided, comprising: an acquisition module, configured to acquire a target instruction file from a target path, wherein the target instruction file includes business operation instructions generated by an intelligent agent according to preset business rules, and the target path is an interaction path between the intelligent agent and a business system; a parsing module, configured to parse the target instruction file to obtain multi-dimensional information, wherein the multi-dimensional information is used to reflect whether the business operation instructions conform to the preset compliance policy of the business system from multiple dimensions; and an execution module, configured to execute the business operation instructions when the multi-dimensional information indicates that the business operation instructions conform to the preset compliance policy.

[0013] According to another aspect of the embodiments of this application, another business execution apparatus is also provided, including: a generation module, used to obtain business requirements of a business system, and generate an instruction file based on the business requirements and preset business rules, wherein the instruction file includes business operation instructions; and a writing module, used to write the instruction file to a target path, wherein the target path is an interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform business operations corresponding to the business operation instructions.

[0014] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method for implementing the above-mentioned business execution.

[0015] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned service execution method by running the computer program.

[0016] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the above-described method for performing business operations.

[0017] In this embodiment, an asynchronous instruction transmission method based on an intermediate file is adopted. By obtaining the business operation instruction file generated by the agent from the preset interaction path, parsing the file to obtain multi-dimensional information reflecting the compliance of the instruction, and executing the instruction after confirming that the multi-dimensional information meets the preset compliance policy, the purpose of achieving permission isolation and pre-verification between the agent and the business system is achieved. This realizes the technical effect of securely triggering business operations without the agent directly calling the enterprise business system API, thereby solving the technical problem in related technologies where embodied agents driven by large language models execute external operations by directly calling the enterprise business system API, which poses a risk of executing unauthorized system operations and leads to poor system stability. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 This is a hardware structure block diagram of a computer terminal for a business execution method according to an embodiment of this application;

[0020] Figure 2 This is a flowchart of a business execution method according to an embodiment of this application;

[0021] Figure 3 This is a flowchart of another method for performing a business operation according to an embodiment of this application;

[0022] Figure 4 This is a system architecture diagram of a business execution method according to an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of a business execution apparatus according to an embodiment of this application;

[0024] Figure 6 This is a schematic diagram of the structure of an apparatus for performing another service according to an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0028] Intelligent Agent: In the field of artificial intelligence, an intelligent agent refers to a software entity that has the ability to perceive the environment, make autonomous decisions, and execute actions. It can understand task objectives and generate operation sequences based on a large language model. In the embodiments of this application, an intelligent agent refers to an embodied intelligent software entity driven by a large language model and used to automate enterprise business tasks. It does not directly call the business system API, but participates in system interaction by generating structured instruction files.

[0029] Business System: A software system within an enterprise that supports core business processes, such as ERP, CRM, financial systems, or databases. It has data processing, transaction execution, and access control capabilities. In this embodiment, the business system is responsible for actively pulling instruction files generated by the intelligent agent, performing compliance verification, and executing the corresponding business operations after passing the verification. It is the trusted execution environment for the final execution action.

[0030] Target Path: A shared storage space used to transmit instructions between the agent and the business system. It is usually an object storage directory or FTP path and has file-level read and write permission control capabilities. In the embodiments of this application, the target path serves as a physical isolation layer, allowing only the agent to write to and the business system to read the instruction file, thereby decoupling the network and permissions between the two.

[0031] Instruction File: A structured data file generated by the agent and following a preset format, used to describe the business operation instructions to be executed. It typically includes fields such as operation type, parameters, timestamp, and signature. In this embodiment, the instruction file is stored in the target path in JSON format, serving as the sole communication medium between the agent and the business system, carrying all operational intents, and avoiding security exposure caused by real-time API calls.

[0032] Multi-dimensional information: A data set of multiple independent judgment dimensions parsed from the instruction file for comprehensively evaluating the compliance of operations, including but not limited to operation permissions, parameter rationality, and business context consistency. In this application embodiment, multi-dimensional information is the input basis for the business system to perform pre-audit to determine whether the instruction meets the enterprise's preset security and business rules.

[0033] Pre-defined compliance policy: A set of business rules and security policies pre-configured by an enterprise to constrain the operational behavior of intelligent agents, including constraints such as permission whitelists, amount limits, and state dependencies. In this embodiment, the pre-defined compliance policy is built into the decision engine of the business system to automatically compare the multi-dimensional information carried by the instruction file and decide whether to allow the execution of the operation. It is the core control basis for realizing risk interception and audit closed loop.

[0034] Asynchronous interaction refers to an interaction mode in which the two parties do not rely on real-time connection and complete the request and response in a time-sharing manner through an intermediate medium. In the embodiments of this application, the intelligent agent and the business system realize asynchronous interaction through the instruction file in the target path: after the intelligent agent writes the instruction, it does not need to wait for the response. After the business system detects the file, it actively pulls it and processes it, thereby avoiding the coupling and attack surface expansion caused by real-time calls. It is a key mechanism for building a secure isolation architecture.

[0035] With the rapid development of Large Language Model (LLM) technology, the Embodied Agent framework has been widely applied. The agent is no longer limited to natural language interaction, but can understand complex business goals and actively operate external systems to complete end-to-end tasks.

[0036] To achieve a closed-loop business process, intelligent agents need to frequently interact with core enterprise business systems (such as ERP, CRM, and databases). However, related technologies generally employ direct Skill calls or API hook mechanisms, granting intelligent agents the authority to directly initiate network requests and execute external operations, resulting in a tightly coupled architecture between the intelligent agent and the business system. In this model, if an intelligent agent generates malicious or erroneous commands due to model illusion or a prompt injection attack, it can easily trigger unauthorized system operations (such as accidental data deletion or illegal transfers), severely blurring security boundaries.

[0037] Furthermore, the synchronous invocation method of related technologies usually adopts the process of "execute first, record later", which lacks the ability to pre-audit the operation instructions, making it impossible for the audit system to perform in-depth verification of compliance before the action is executed. In addition, since the instructions generated by the large language model have high uncertainty and unpredictability, their concurrent invocation of business APIs can easily cause the target system to experience service interruption or system crash due to abnormal traffic surges or illegal parameter injection, which significantly affects business continuity and system stability.

[0038] To address the aforementioned technical problems, this application provides corresponding solutions, which are detailed below.

[0039] The service execution method embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a method of business execution is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0040] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the business execution method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned business execution method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0042] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0043] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0044] It should be noted here that, in some optional embodiments, the above... Figure 1The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0045] In the above operating environment, this application provides a method embodiment for business execution. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.

[0046] Figure 2 This is a flowchart of a business execution method according to an embodiment of this application, wherein the execution subject is a business system, such as... Figure 2 As shown, the method includes the following steps:

[0047] Step S202: Obtain the target instruction file from the target path, wherein the target instruction file includes business operation instructions generated by the agent according to preset business rules, and the target path is the interaction path between the agent and the business system.

[0048] In step S202 above, the target path refers to a shared storage area with access control used by the agent and the business system to transmit operation instructions. Essentially, it is an isolated, non-real-time communication intermediary, such as a directory space accessible to both parties (e.g., a specific path in object storage or a restricted directory in an FTP server) that can exist in the form of a file system, achieving physical decoupling between the agent and the business system. It should be noted that by storing the instructions as files in this path, the agent does not need network call permissions, and the business system does not need to expose external interfaces, thus completely blocking the potential path for the agent to directly attack the business system.

[0049] The target instruction file is an instruction carrier that follows a standardized structure and is generated by the intelligent agent according to the task objectives and preset business rules. It includes, but is not limited to, structured information such as operation type, parameter value, timestamp, intelligent agent identity and digital signature. This file is the only legal information transmission medium between the intelligent agent and the business system in this application, and is used to transform unpredictable model output into structured instruction units that can be audited, verified and traced.

[0050] An intelligent agent refers to an autonomous software entity built on a large language model that has the ability to understand tasks and generate instructions. It does not directly control external systems, but expresses operational intentions by generating target instruction files. For example, an intelligent agent can be designed as a restricted environment component with "no network access permissions", and all its external interaction behaviors are restricted to writing operations to the target path.

[0051] In some embodiments of this application, the business system deploys a file listener to continuously monitor file change events in the target path. When the agent writes the target instruction file to the path, the file system triggers a "new file" event. The listener captures this event and immediately starts the file reading process, loading the file content into memory for parsing. Alternatively, the business system starts a polling task at fixed time intervals to actively scan all newly added or modified files in the target path. By comparing the file creation timestamps with the previous polling record, it identifies target instruction files that have not yet been processed.

[0052] In high-concurrency scenarios, after the agent generates the target instruction file, the file may fail to be written to the target path due to file system write delays, permission errors, disk overload, or path mount failures. Consequently, the business system cannot execute the corresponding operations because it has not received the instruction, resulting in an "instruction loss" phenomenon. To resolve this issue, the following steps can be performed:

[0053] After generating the target instruction file, the agent writes the file to the target path and simultaneously generates a corresponding verification identifier file. The verification identifier file contains the hash value of the target instruction file, the generation time, and the agent's identity identifier, and its filename has a unique mapping relationship with the target instruction file. Before reading the target instruction file from the target path, the business system first reads the corresponding verification identifier file to verify its integrity and consistency. If the verification identifier file is missing, the hash value does not match, or the timestamp is abnormal, the business system determines that the target instruction file is in an abnormal state, returns an error response to the agent, and triggers a retry mechanism, requiring the agent to regenerate and write the target instruction file and the verification identifier file. If the verification identifier file is valid, the business system then reads the target instruction file and executes the subsequent compliance verification and operation execution process. After successfully executing the operation, the business system deletes the target instruction file and its corresponding verification identifier file to ensure that the instruction is executed only once and to avoid duplicate processing.

[0054] The above embodiments, by introducing a dual confirmation mechanism of file and verification identifier, construct an atomicity guarantee for instruction delivery, effectively solve the problem of instruction loss caused by abnormal file writing, and significantly improve the system's fault tolerance and execution reliability in complex environments.

[0055] In some embodiments of this application, object storage services and file transfer protocol services can be used as intermediaries, specifically:

[0056] (1) The target path includes a first path determined based on the object storage service, which is a virtual storage path pre-configured in the object storage service. Based on this, the target instruction file can be obtained from the target path in the following way: call the file query interface of the object storage service and periodically check the instruction files in the first path to obtain a first check result; if the first check result indicates that a first instruction file exists in the first path, the first instruction file is used as the target instruction file, wherein the first instruction file includes files that have not been read in the first path.

[0057] It should be noted that the first path refers to the virtual directory space provided by the object storage platform, which has high availability and cross-regional access capabilities. Logically, this path is represented by a URL path prefix, and its underlying structure consists of distributed file blocks, without being bound to a specific physical server.

[0058] Specifically, the business system deploys a timed polling task that queries the metadata of all files under the first path at fixed intervals (e.g., 5 seconds) through the standard file list interface provided by the object storage service. It compares the creation timestamp of each file with the last processing time recorded locally, selecting unprocessed new files as the first instruction file. Alternatively, the business system subscribes to the object storage service's event notification mechanism. When a new file is written to the first path, the object storage platform proactively sends a file creation event to the callback address registered by the business system. Upon receiving the event, the business system immediately triggers the instruction file download process, eliminating the need for periodic polling.

[0059] (2) The target path includes a second path determined based on the File Transfer Protocol (LTP) service, which is the local file system path on the server where the LTP service is deployed. Based on this, the target instruction file can be obtained from the target path in the following way: establish a connection with the server corresponding to the LTP service, and periodically check the instruction file in the second path to obtain a second check result; if the second check result indicates that a second instruction file exists in the second path, use the second instruction file as the target instruction file, wherein the second instruction file includes files that have not been read in the second path.

[0060] It should be noted that the second path refers to the local file system directory deployed on the enterprise's intranet server and providing file read and write services to the outside world through file transfer protocols (such as FTP or SFTP). This path relies on the traditional network file service architecture and has the characteristics of low dependency, easy deployment, and strong firewall penetration.

[0061] Specifically, the business system establishes a secure connection with the file transfer protocol server via SFTP at a fixed frequency (e.g., every 3 seconds), lists all files under the second path, determines whether there are any unprocessed new files based on filename hash values ​​or modification timestamps, and loads files that meet the requirements as second instruction files into the local cache. Alternatively, the business system deploys a lightweight file monitoring agent on the file transfer protocol server. When a new file is written to the second path, the agent program triggers a local script to send a "instruction arrived" notification to the business system via a preset HTTP callback or message queue. Upon receiving the notification, the business system immediately initiates an SFTP connection to retrieve the file.

[0062] Step S204: Parse the target instruction file to obtain multi-dimensional information, which is used to reflect whether the business operation instructions comply with the preset compliance policy of the business system from multiple dimensions.

[0063] In step S204 above, multi-dimensional information refers to a set of multiple independent data dimensions extracted after deep analysis of the target instruction file, which constitute the basis for compliance judgment. Each dimension corresponds to a preset business or security rule, including but not limited to: legality of operation permissions, reasonableness of parameter value range, consistency of business context state, compliance of operation frequency threshold, validity of signature, etc.

[0064] In some embodiments of this application, after receiving the target instruction file, the business system first loads a pre-set structured parsing template and extracts key fields such as operation type, parameter value, agent number, and timestamp item by item according to the file format (such as JSON field mapping). Then, these fields are input into the rule engine module built into the business system. The rule engine performs parallel comparisons based on multiple pre-configured business rules (such as permission list, amount limit, and state dependency graph) and outputs Boolean judgment results for each dimension, which are combined to form a complete multi-dimensional information set.

[0065] In some embodiments of this application, the target instruction file includes an identifier for an intelligent agent and business operation parameters. Based on this, the target instruction file can be parsed to obtain multi-dimensional information in the following ways: obtaining the operation permissions corresponding to the intelligent agent's identifier, and comparing the target business type corresponding to the business operation instruction with the operation permissions to obtain a first comparison result, wherein the operation permissions reflect the range of business types controlled by the intelligent agent; comparing the business operation parameters with preset parameter thresholds to obtain a second comparison result; obtaining a first business context state corresponding to the business system and a second business context state on which the business operation instruction depends, and comparing the first business context state with the second business context state to obtain a third comparison result; and using the first comparison result, the second comparison result, and the third comparison result together as multi-dimensional information.

[0066] It should be noted that business operation parameters refer to the structured data set explicitly specified by the intelligent agent in the target instruction file, which describes the specific business operation content. These parameters include, but are not limited to, the operation object, numerical value, time range, and associated entity number. For example, the instruction "Purchase 500 units of material A at a unit price of 120 yuan" includes three key parameters: "material code," "quantity," and "unit price." The system uses these parameters to determine whether the total purchase quantity exceeds the budget and whether the unit price deviates from the historical range.

[0067] Operation permissions refer to the set of business types that are pre-configured for a specific intelligent agent and that it is allowed to execute. Operation permissions are the basis for the first dimension of compliance judgment. Their role is to implement the "principle of least privilege" and ensure that the intelligent agent can only execute the business actions that are explicitly authorized. For example, if an intelligent agent is granted the permission to "approve a purchase application" but does not have the permission to "initiate a payment", even if it generates a payment instruction, the system will intercept it based on the permission comparison result.

[0068] The first business context state refers to the global operational state of the business system at the current moment, including real-time data strongly related to business logic such as account balance, inventory quantity, order status, financial cycle, and approval flow. The second business context state refers to the preconditions upon which business operation instructions depend, usually implicitly or explicitly stated in the instruction content, such as "refunds can only be executed when the order status is 'pending shipment'" or "purchases can only be initiated when the budget balance is greater than 0." This state is the logical premise for the legality of the operation, and its function is to implement "state-driven operation control" to prevent destructive behavior from being triggered in incorrect contexts.

[0069] Specifically, after receiving the target instruction file, the business system queries its internal permission management database using the agent identifier as the key to retrieve the whitelist of authorized business types, such as "purchase application, inventory query, supplier maintenance". At the same time, it extracts the operation type field (such as "initiate payment") from the instruction content and compares it with the whitelist item by item. If the operation type is not in the whitelist, the first comparison result is "not approved".

[0070] Furthermore, the business system defines fixed parameter thresholds for each type of business operation based on a preset rule configuration file. For example, the maximum amount for a single purchase is 5,000 yuan, and the quantity of a single inventory transfer cannot exceed 30% of the total inventory. The system extracts the parameter values ​​from the instructions and directly compares them with these thresholds. If the values ​​are exceeded, the system marks the operation as "non-compliant".

[0071] Furthermore, the business system obtains the current real status (e.g., "Contract C20260315 status = approved") by directly querying the core database, and extracts the prerequisite status (e.g., "required contract status = accepted") from the target instruction file. By comparing the strings or enumerated values, it determines whether the two are consistent. If they are inconsistent, the third comparison result is "no".

[0072] During the multi-dimensional information generation process, if the validation of a certain dimension is not completed due to data retrieval failure, missing fields, or system delays, but the system continues to perform subsequent judgments, it may lead to a dangerous situation where "some dimensions are missing but are passed by default." To solve this problem, the following steps can be performed:

[0073] After parsing the target instruction file, independent verification status flags are set for the two key verification dimensions: operation permission comparison and business context state comparison. The initial status is "not executed". Before executing the verification of each key dimension, the availability of the corresponding data source is checked. If the permission database is inaccessible or the core business system status is not synchronized, the verification status of the corresponding dimension is marked as "data missing". After all verifications are completed, if the verification status of any key dimension is "data missing", the process is immediately terminated, no valid multi-dimensional information is generated, and a compliance assessment interruption alarm is triggered. Only when the verification status of all key dimensions is "completed" and the verification results are all "compliant" is the overall compliance determined and the execution of business operation instructions is allowed. Otherwise, it is determined as "not in compliance with the preset compliance policy".

[0074] In some embodiments of this application, the second business context state can be determined by: parsing the business operation instruction text contained in the target instruction file, identifying the explicitly mentioned business entity identifiers and operation semantics, wherein the business entity identifiers include, but are not limited to, contract number, order number, account identifier, material code, or process instance ID, and the operation semantics include, but are not limited to, "cancel," "release," "transfer," "pay," "close," "update," or "confirm," etc.; and mapping the identified operation semantics to a set of standardized pre-state constraints according to a pre-set business semantic rule library, wherein the business semantic rule library is defined by enterprise business experts and corresponds one-to-one with various operation types. The logical relationship set includes an operation type, a set of required prerequisite states, and logical relationships between states for each rule. If the required prerequisite states are explicitly declared in the business operation instruction, the explicit declaration is extracted as a candidate value for the second business context state. If not explicitly declared, the standard value of the second business context state is automatically generated based on the standard prerequisite state constraints obtained from the above mapping. In addition, the generated second business context state can be structurally standardized and converted into a unified expression format consistent with the internal state representation format of the business system, including field naming conventions, enumeration value encoding, timestamp format, or state machine identifier code, to ensure that it can be directly used by the business system for state comparison.

[0075] Step S206: If the multi-dimensional information indicates that the business operation instruction complies with the preset compliance policy, execute the business operation instruction.

[0076] In step S206 above, the preset compliance policy refers to a predefined set of rules used to regulate the behavioral boundaries of an intelligent agent when performing operations on an external system. These policies are usually solidified in the system in the form of a business rule engine, including but not limited to permission lists, amount limits, state dependencies, operation frequency limits, etc.

[0077] In some embodiments of this application, after the business system obtains multi-dimensional information that has passed all dimension verifications, it calls the internally pre-built "security execution controller". The controller generates a standardized business operation request based on the instruction type, parameter value and operation object in the multi-dimensional information, and sends it to the target business module (such as the financial system or ERP) through the internal service bus.

[0078] In some embodiments of this application, the following steps may also be performed: when a business operation instruction is completed, obtain the execution result corresponding to the business operation instruction and convert the execution result into an execution result file, wherein the execution result file has the same format as the target instruction file; when multi-dimensional information indicates that the business operation instruction does not meet the preset compliance policy, intercept the business operation instruction and generate an interception result file, wherein the interception result file is used to instruct the intelligent agent to initiate a manual review process; and write the execution result file or the interception result back to the target path.

[0079] It should be noted that the execution result file refers to a structured feedback file generated by the business system after the business operation instruction is successfully executed. This file is identical in format to the target instruction file and includes fields such as operation status (success / failure), execution time, returned data, operation result summary, and transaction identifier. The interception result file refers to a structured notification file automatically generated by the business system when multi-dimensional information determines that a business operation instruction does not comply with the preset compliance policy. This file clearly indicates the reason for the interception (e.g., "amount exceeded," "insufficient permissions," "state conflict") and guides the intelligent agent to initiate a manual review process.

[0080] Write-back to the target path refers to writing the execution result file or interception result file into a shared intermediate storage area in the same storage method (object storage or file transfer protocol path) as the original target instruction file, so that the agent can read it. It should be noted that the intermediate storage area here can also be a shared storage path different from the path for reading the instruction file, which is not limited here.

[0081] Specifically, after the business system completes the call to the underlying business module, it extracts key fields (such as status code, business number, and transaction amount) from the returned response data, and fills in the corresponding fields according to the JSON structure template of the target instruction file to generate an execution result file with the same format; or, if any dimension verification fails (such as exceeding the amount limit or missing permissions), the system directly fills in the interception reason field (such as "single purchase amount exceeds the authorized limit of 5,000 yuan") and the guiding action (such as "please submit manual approval") according to the preset interception template, generates an interception result file with the same structure as the target instruction file, and writes it to the shared storage path.

[0082] If, during the process of writing the result file or intercepted result file back to the target path, the result file fails to be written successfully or is incorrectly overwritten due to storage service exceptions, filename conflicts, incorrect write permissions, or concurrent write contention, the agent will be unable to obtain feedback. To resolve this issue, the following steps can be performed:

[0083] After generating the execution result file or interception result file, the business system first calculates the unique hash value of the file content and generates a unique filename strongly associated with the original instruction file ID. Before writing to the target path, the business system checks if a file with the same name exists. If it does, it compares the file hash values. If the hash values ​​are the same, it is determined to be a duplicate write and the write is abandoned. If the hash values ​​are different, the original file is marked as "expired" and the log is retained before the new file is written. After the file is written, the business system actively verifies the physical existence and content integrity of the file in the target path by reading and recalculating the hash value and comparing it with the value generated to ensure no write loss or truncation. If file writing or verification fails, the system stores the feedback task in the "Feedback Queue to be Retried" and initiates an exponential backoff retry mechanism, with no more than three retries (as an example), and the retry interval increasing with each retry. If it still fails after three retries, the system packages the original content of the instruction, the processing log, and the reason for failure into a "Feedback Loss Alarm Record," pushes it to the operation and maintenance audit platform, and triggers the intelligent agent's state freeze to prevent it from continuing to generate similar instructions until manual intervention is required. All successfully written result files must have a "Command-Feedback" two-way binding index established in the business system to ensure that any subsequent audit or anomaly investigation can quickly locate the corresponding feedback file through the instruction ID.

[0084] Through steps S202 to S206 above, an asynchronous instruction transmission method based on intermediate files is adopted. By obtaining the business operation instruction file generated by the agent from the preset interaction path, parsing the file to obtain multi-dimensional information reflecting the compliance of the instruction, and executing the instruction after confirming that the multi-dimensional information meets the preset compliance policy, the purpose of achieving permission isolation and pre-verification between the agent and the business system is achieved. This realizes the technical effect of securely triggering business operations without the agent directly calling the enterprise business system API. In turn, it solves the technical problem in related technologies where embodied agents driven by large language models execute external operations by directly calling the enterprise business system API, which may lead to the risk of executing unauthorized system operations and poor system stability.

[0085] Figure 3 This is a flowchart of another business execution method according to an embodiment of this application, where the execution subject is an intelligent agent, such as... Figure 3 As shown, the method includes:

[0086] Step S302: Obtain the business requirements of the business system, and generate an instruction file based on the business requirements and preset business rules. The instruction file includes business operation instructions.

[0087] In step S302 above, business requirements refer to the original intentions or task objectives triggered by enterprise operation scenarios, external user interactions, or automated processes, which require the participation of business systems. They usually exist in the form of natural language descriptions, event signals, or system triggering conditions. They are used to transform abstract business objectives (such as "automatic replenishment when inventory is below the threshold") into input signals that can be understood and processed by intelligent agents. For example, if the ERP system detects that the inventory of a certain material has dropped to 10 units, it sends a "trigger replenishment decision" event to the intelligent agent. This event constitutes a business requirement.

[0088] Preset business rules refer to a set of structured logical constraints used to guide intelligent agents in generating legal, safe, and compliant operation instructions. These rules cover dimensions such as operation prerequisites, permission boundaries, parameter restrictions, and state dependencies. For example, the rules may explicitly state that "the amount of a single purchase shall not exceed 5,000 yuan" and "payment is only allowed before 18:00 on working days." Intelligent agents must strictly follow these rules when generating instructions.

[0089] In some embodiments of this application, the intelligent agent receives business demand signals triggered by the business system (e.g., "the inventory monitoring module detected that the inventory of material A001 is lower than the safety threshold") by listening to preset event notification channels (such as file system monitoring, local message queue or system log polling); then, the intelligent agent calls the built-in rule engine to perform semantic parsing of the demand according to preset business rules, extract key elements (such as operation type "procurement" and object "material A001"), and combines them with the parameter range set in the rules (such as quantity limit and price range) to generate a structured instruction file that meets the format requirements and write it to a restricted output directory.

[0090] In other embodiments of this application, the intelligent agent does not rely on external systems to actively push business requirements. Instead, it periodically and proactively queries the status interfaces of multiple business systems (such as inventory systems, order systems, and contract systems) to comprehensively analyze multiple business statuses and autonomously identify potential business opportunities. For example, the intelligent agent scans the order system every hour and finds that "a customer has placed orders for the same material three times in a row but has not replenished it." Combined with the information in the inventory system that "the current stock of this material is lower than the consumption of the procurement cycle," and based on implicit business rules such as "high customer loyalty → priority replenishment," it automatically derives the business requirement of "suggesting to initiate replenishment" and generates an instruction file accordingly.

[0091] Step S304: Write the instruction file to the target path, where the target path is the interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform the business operation corresponding to the business operation instruction.

[0092] In some embodiments of this application, the target path includes one of the following: a first path determined based on the object storage service and a second path determined based on the file transfer protocol service, wherein the first path is a virtual storage path pre-configured in the object storage service, and the second path is a local file system path on the server where the file transfer protocol service is deployed; writing the instruction file to the target path includes: calling the file upload interface of the object storage service to store the instruction file in the form of an object in the first path; or, connecting to the server corresponding to the file transfer protocol service to upload the instruction file to the second path in the form of a file transfer.

[0093] In some embodiments of this application, the following steps may also be performed: receiving an execution result file from the target path, wherein the execution result file includes the result obtained by the business system when the business operation instruction meets the preset compliance policy; generating a decision instruction based on the execution result file.

[0094] It should be noted that the decision instruction refers to the next round of operation intention instructions that the intelligent agent autonomously generates based on the content, business objectives and preset logical rules after receiving the execution result file. It is used to drive subsequent business behaviors, such as initiating manual review, adjusting parameters for retry, notifying the person in charge or terminating the process.

[0095] Specifically, the agent continuously monitors the shared storage paths it is authorized to access through periodic polling or file monitoring mechanisms. When a new file is detected and its filename matches a preset format, it automatically reads the file content and parses it. Alternatively, the agent does not actively poll the file system but instead receives "result-ready notifications" from the business system via operating system signals or inter-process communication mechanisms through a local lightweight event monitoring agent. For example, after the business system completes a write operation, it sends a predefined Unix signal (such as SIGUSR1) to the container where the agent resides or writes a flag bit through shared memory. Upon receiving the notification, the agent immediately loads the latest file from the specified path.

[0096] Furthermore, the agent triggers pre-defined conditional branching logic based on the "Status" field value in the execution result file. For example, if the status is "successful," it logs the results and proceeds to the next task; if the status is "failed" and the reason field is "amount exceeded," it generates a decision instruction: "submit manual approval application," encapsulates it into a new instruction file, and writes it into the approval process entry directory.

[0097] It should be noted that, Figure 3 Preferred embodiments of the shown examples can be found in [reference needed]. Figure 2 The corresponding solutions in the illustrated embodiments will not be described in detail here.

[0098] Figure 4This is a system architecture diagram of a business execution method according to an embodiment of this application, such as... Figure 4 As shown, in some embodiments of this application, the system can perform the following steps:

[0099] (1) Instruction solidification under restricted environment.

[0100] To achieve physical isolation and minimize permissions between the agent and the core business system, this embodiment implements a strict permission constraint mechanism in the agent's runtime environment (such as an OpenClaw framework instance). Upon startup, the agent is placed in a sandbox-like restricted runtime environment. The system completely disables its active access to external networks through containerized isolation or operating system-level permission control, including but not limited to disabling TCP / UDP connections, disabling DNS resolution, and blocking HTTP / HTTPS request interfaces, ensuring that the agent cannot directly call any business system APIs or remote services.

[0101] Based on this, only a pre-configured, write-only restricted storage path is mounted to the agent. This path is an object storage directory independent of the business system's network domain (such as a private cloud object storage compatible with the S3 protocol) or a write-only FTP / SFTP entry directory, and this directory does not support read, delete, or rename operations. During task execution, the agent can only generate a compliant instruction file through the local file system interface, based on a preset structured template, and write the file to the specified path atomically. The instruction file adopts the standard JSON format and contains at least the following fields: a unique operation identifier (ActionID), the type of operation to be executed and the parameter set, a digital signature generated by the agent's private key (for authentication), an operation initiation timestamp (UTC format), and a unique agent identifier (AgentID).

[0102] This mechanism cuts off the direct control channel of intelligent agents over business systems at the source, completely separating "operation intent generation" from "operation permission execution" in physical space and permission dimension, fundamentally eliminating the risk of illegal direct connection of instructions caused by model illusion or prompt word injection.

[0103] (2) Asynchronous triggering and retrieval.

[0104] To avoid establishing any real-time communication link between the agent and the business system, this embodiment adopts an asynchronous interaction mode of "active retrieval, non-real-time response". On the business system side, an independent file monitoring service or scheduled task is deployed. This service runs in an internal secure area completely isolated from the agent environment, without exposing any ports or interfaces. This service continuously monitors the aforementioned mounted restricted storage directory, using file system event listening or periodic polling mechanisms (default scan interval not exceeding 10 seconds) to detect whether there are any newly created files that conform to the command file naming conventions. Once a new file is detected, the system immediately performs a local file read operation, obtains the file content and loads it into the memory buffer. The entire process does not rely on network requests, does not trigger remote service calls, and does not establish bidirectional connections, ensuring that the interaction is entirely triggered unidirectionally by the business system.

[0105] This asynchronous mechanism effectively shields the system from the influence of external environments such as network latency, service jitter, and firewall policies, giving the system strong fault tolerance. At the same time, it avoids business system overload or denial-of-service attacks caused by abnormally active agents, realizing a secure interaction paradigm where "instructions are issued by the agent and execution rights are controlled by the business system".

[0106] (3) Multidimensional compliance decision engine.

[0107] After the business system receives the instruction file generated by the intelligent agent, it does not immediately execute the operation. Instead, it initiates a structured, multi-layered compliance decision engine to deeply validate the instruction content, ensuring that every operation complies with the enterprise's security policy, business logic, and data consistency requirements. This engine has three core validation dimensions, which are executed sequentially. If any dimension fails, the process terminates and execution is prohibited.

[0108] Authentication check: The system parses the AgentID and operation type fields in the instruction file, queries the enterprise's unified permission management platform for the minimum set of permissions granted to the agent, and determines whether it has explicit authorization to execute the current operation type (such as "initiate payment", "delete contract", "change inventory"). If there is no corresponding permission, or the permission has been temporarily frozen, it is determined as "insufficient permissions", the process is immediately terminated, and an audit log is recorded.

[0109] Logical Judgment: The system performs compliance analysis on the parameters in the instructions based on a pre-defined business rule base. This rule base, defined by business experts, covers logical rules such as amount thresholds (e.g., a single transfer cannot exceed 5,000 yuan), operation frequency limits (e.g., only three inventory adjustments are allowed per day), numerical ranges (e.g., material codes must exist in the master data list), and object status constraints (e.g., purchase requests are limited to contracts in the "approved" status). The engine compares the instruction parameters with the rules item by item. If any violation occurs, such as exceeding the amount limit, a negative quantity, or a non-existent object, it is judged as a "logical violation," and a specific reason for the violation is generated.

[0110] State consistency verification: The system queries the current state snapshot of the core business system in real time (such as account balance, inventory quantity, order status, and contract lifecycle nodes) and compares it with the preceding states that are implicitly or explicitly depended upon in the instruction. For example, if the instruction requires "closing contract number C20260315", the system must confirm that the current state of the contract is "accepted" rather than "in execution" or "under dispute". If there is a state conflict, it is determined as "context inconsistency", and execution is refused to prevent damage to the integrity of the business process.

[0111] (4) Closed-loop feedback of instructions.

[0112] After the multi-dimensional compliance decision engine determines the legality of the instruction, the business system executes the actual business operation. Upon completion of the operation, regardless of success or failure, a feedback file with the exact same structure as the original instruction file must be generated as the final response to the agent. This feedback file also uses standard JSON format, containing the same ActionID as the original instruction to achieve bidirectional association, and adding fields including: execution status ("success" or "failure"), execution timestamp, operation return data (such as the generated order number, updated inventory quantity, and system-returned error code), and an optional explanation of the failure reason. This feedback file is atomically written by the business system to the same restricted storage path as the original instruction (it should be noted that if the first target path for the agent to write the instruction file is set to a unidirectional path, a second target path can be set for writing back; under this second target path, the business system only has write permissions, and the agent only has read permissions), for the agent to read during the next poll or event trigger. By parsing this feedback file, the agent can clearly know the final result of the previous action and decide whether to trigger subsequent operations, such as initiating manual approval, adjusting parameters for retry, notifying the responsible party, or terminating the task flow.

[0113] Figure 5 This is a structural diagram of a service execution apparatus according to an embodiment of this application, such as... Figure 5 As shown, the device includes:

[0114] The acquisition module 502 is used to acquire a target instruction file from the target path, wherein the target instruction file includes business operation instructions generated by the agent according to preset business rules, and the target path is the interaction path between the agent and the business system;

[0115] The parsing module 504 is used to parse the target instruction file to obtain multi-dimensional information, which is used to reflect whether the business operation instructions comply with the preset compliance policy of the business system from multiple dimensions.

[0116] The execution module 506 is used to execute business operation instructions when multi-dimensional information indicates that the business operation instructions comply with the preset compliance policy.

[0117] It should be noted that, Figure 5 The device shown is used to execute business operations. Figure 2 The method of business execution shown, therefore Figure 2 The relevant explanations and instructions in the business execution methods also apply to Figure 5 The apparatus for executing the business shown will not be described in detail here.

[0118] Figure 6 This is a structural diagram of an apparatus for performing another service according to an embodiment of this application, such as... Figure 6 As shown, the device includes:

[0119] The generation module 602 is used to obtain the business requirements of the business system and generate an instruction file based on the business requirements and preset business rules. The instruction file includes business operation instructions.

[0120] The writing module 604 is used to write the instruction file to the target path, where the target path is the interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform the business operation corresponding to the business operation instruction.

[0121] It should be noted that, Figure 6 Another business execution device shown is used to execute Figure 3 The method of business execution shown, therefore Figure 3 The relevant explanations and instructions in the business execution methods also apply to Figure 6 The devices that execute business processes are not described in detail here.

[0122] This application also provides an electronic device, which includes a memory and a processor. The memory is used to store program instructions, and the processor is connected to the memory and is used to execute the steps of the methods for implementing the business execution methods in the various embodiments of this application.

[0123] This application also provides a non-volatile storage medium including a stored computer program, wherein the device containing the non-volatile storage medium executes the steps of the service execution methods in various embodiments of this application by running the computer program.

[0124] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the business execution methods in various embodiments of this application.

[0125] This application also provides a computer program that, when executed by a processor, implements the steps of the business execution methods in various embodiments of this application.

[0126] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0127] In the above embodiments of this application, 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.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0129] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0130] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0131] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0132] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for executing business operations, characterized in that, include: Obtain the target instruction file from the target path, wherein the target instruction file includes business operation instructions generated by the intelligent agent according to preset business rules, and the target path is the interaction path between the intelligent agent and the business system; The target instruction file is parsed to obtain multi-dimensional information, which is used to reflect whether the business operation instruction conforms to the preset compliance policy of the business system from multiple dimensions. If the multi-dimensional information indicates that the business operation instruction conforms to the preset compliance policy, the business operation instruction shall be executed.

2. The method according to claim 1, characterized in that, The target path includes a first path determined based on the object storage service; the first path is a virtual storage path pre-configured in the object storage service. Obtaining a target instruction file from a target path includes: calling the file query interface of the object storage service and periodically checking the instruction files in the first path to obtain a first detection result; if the first detection result indicates that a first instruction file exists in the first path, using the first instruction file as the target instruction file, wherein the first instruction file includes files that have not been read in the first path; or... The target path includes a second path determined based on the File Transfer Protocol (LTP) service; the second path is a local file system path on the server where the LTP service is deployed; obtaining the target instruction file from the target path includes: establishing a connection with the server corresponding to the LTP service, and periodically detecting the instruction files in the second path to obtain a second detection result; if the second detection result indicates that a second instruction file exists in the second path, the second instruction file is used as the target instruction file, wherein the second instruction file includes files that have not been read in the second path.

3. The method according to claim 1, characterized in that, The target instruction file includes the agent's identifier and business operation parameters; parsing the target instruction file yields multi-dimensional information, including: Obtain the operation permission corresponding to the identifier of the intelligent agent, and compare the target business type corresponding to the business operation instruction with the operation permission to obtain a first comparison result, wherein the operation permission is used to reflect the range of business types controlled by the intelligent agent; The business operation parameters are compared with preset parameter thresholds to obtain a second comparison result; Obtain the first business context state corresponding to the business system and the second business context state on which the business operation instruction depends, and compare the first business context state with the second business context state to obtain a third comparison result; The first comparison result, the second comparison result, and the third comparison result are collectively used as the multi-dimensional information.

4. The method according to claim 1, characterized in that, The method further includes: Upon completion of the business operation instruction, the execution result corresponding to the business operation instruction is obtained, and the execution result is converted into an execution result file, wherein the execution result file has the same format as the target instruction file; If the multi-dimensional information indicates that the business operation instruction does not meet the preset compliance policy, the business operation instruction is intercepted and an interception result file is generated, wherein the interception result file is used to instruct the intelligent agent to initiate a manual review process; Write the execution result file or the interception result back to the target path.

5. A method for executing business operations, characterized in that, include: Obtain the business requirements of the business system, and generate an instruction file based on the business requirements and preset business rules, wherein the instruction file includes business operation instructions; The instruction file is written to the target path, wherein the target path is the interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform the business operation corresponding to the business operation instruction.

6. The method according to claim 5, characterized in that, The target path includes one of the following: a first path determined based on the object storage service and a second path determined based on the file transfer protocol service, wherein the first path is a pre-configured virtual storage path in the object storage service, and the second path is a local file system path on the server where the file transfer protocol service is deployed; writing the instruction file to the target path includes: Call the file upload interface of the object storage service to store the instruction file as an object in the first path; or, A connection is established with the server corresponding to the file transfer protocol service, and the instruction file is uploaded to the second path via file transfer.

7. The method according to claim 6, characterized in that, The method further includes: Receive an execution result file from the target path, wherein the execution result file includes the result obtained by the business system when the business operation instruction meets the preset compliance policy; Decision instructions are generated based on the execution result file.

8. An apparatus for executing business operations, characterized in that, include: The acquisition module is used to acquire a target instruction file from the target path, wherein the target instruction file includes business operation instructions generated by the intelligent agent according to preset business rules, and the target path is the interaction path between the intelligent agent and the business system; The parsing module is used to parse the target instruction file to obtain multi-dimensional information, wherein the multi-dimensional information is used to reflect whether the business operation instruction conforms to the preset compliance policy of the business system from multiple dimensions; The execution module is used to execute the business operation instruction when the multi-dimensional information indicates that the business operation instruction conforms to the preset compliance policy.

9. An apparatus for executing business operations, characterized in that, include: The generation module is used to obtain the business requirements of the business system and generate an instruction file based on the business requirements and preset business rules, wherein the instruction file includes business operation instructions; The writing module is used to write the instruction file to a target path, wherein the target path is the interaction path between the business system and the intelligent agent, and the instruction file is used to instruct the business system to perform a business operation corresponding to the business operation instruction.

10. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method of implementing the business execution according to any one of claims 1 to 4, or to execute the method of implementing the business execution according to any one of claims 5 to 7.

Images