A device, method, and electronic device for workflow execution

Abstract

This application discloses a device, method, and electronic device for workflow execution. It employs a process execution intelligent agent to determine the intent category corresponding to the user-input text information, and to determine a first identifier of the engine tool corresponding to that intent category, as well as a first parameter value of the first attribute parameter required to execute the intent category. The process engine module determines a second identifier of the target process node based on the first identifier, the first attribute parameter, and the first parameter value, and sends information about hook events containing the second identifier to the process execution intelligent agent. The process execution intelligent agent determines and executes a second target operation based on the node type of the target process node and the preset operations corresponding to the node type. This allows the process engine module to focus solely on receiving standardized tool identifiers and executing corresponding target operations, while leaving easily changeable business rules to the process execution intelligent agent. This enables flexible control of the process nodes to be executed during the entire workflow execution process.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more particularly to an apparatus, method and electronic device for workflow execution. Background Technology

[0002] In existing workflow engine technologies, users typically use process design tools to draw workflow (BPMN) diagrams. This involves designing the workflow using the process design tool, exporting the workflow's XML configuration file in XML format, and deploying this XML configuration file to the workflow engine to complete the workflow configuration. When the workflow is completed based on the configuration, the workflow engine activates the corresponding process nodes in the workflow in sequence according to the workflow's XML configuration file and executes the operations of each process node to complete the workflow.

[0003] However, in existing technologies, workflow engines can only accept fixed instructions when executing workflow nodes. For example, users can reject or approve by clicking fixed buttons, which results in a lack of flexibility in controlling the workflow nodes that need to be executed during the entire workflow process. Therefore, there is an urgent need for a more interactive and flexible workflow execution device. Summary of the Invention

[0004] This application provides a device, method, and electronic device for workflow execution, designed to flexibly control the process nodes that need to be executed.

[0005] In a first aspect, embodiments of this application provide a workflow execution device, the device comprising: A process execution intelligent agent is used to determine the intent category corresponding to the text information input by the user interface; determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; and send the first identifier, the first attribute parameter and the first parameter value to the process engine module. The process engine module is configured to receive the first identifier, the first attribute parameter, and the first parameter value; determine the operation type of the engine tool of the first identifier; determine the first target operation according to the operation type and the operation corresponding to the preset operation type; execute the first target operation according to the first parameter value to determine the second identifier of the target process node; and send information of the hook event containing the second identifier to the process execution agent. The process execution agent is also configured to receive information about the hook event, obtain the node type of the target process node of the second identifier from a pre-saved process configuration file according to the second identifier, determine the second target operation according to the node type and the preset operation corresponding to the node type, and execute the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

[0006] In one possible implementation, the process execution agent is specifically used to determine the similarity between each example text and the text information for at least one example text corresponding to each pre-saved intent category; and to take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0007] In one possible implementation, the process execution agent is specifically used to add the text information and the first attribute parameter to a preset template to construct a prompt word, input the prompt word and the text information into a trained large model, and extract the first parameter value corresponding to the first attribute parameter from the text information based on the large model; The process execution agent is also used to determine whether all the first attribute parameters required for the intent category have corresponding first parameter values. If not, it determines the second attribute parameters that are missing first parameter values, and for each second attribute parameter, it fills the second attribute parameter into a preset general template to generate corresponding follow-up text and displays it to prompt the second attribute parameter that is missing first parameter value; and receives the first parameter value of the input second attribute parameter.

[0008] In one possible implementation, the process engine module is specifically configured to: if the operation type is "new creation", determine that the first target operation is to determine, based on the first parameter value, a target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in a preset database, and modify the node state of the first process node in the target workflow to "new creation", thus determining the first process node as the target process node; if the operation type is not "new creation", determine that the first target operation is to modify the node state of the currently executing process node to the target state corresponding to the operation type based on the state corresponding to the operation type and a preset operation type, and determine the processing process node corresponding to the target state based on the process order in the process configuration file and the target state, thus determining the processing process node as the target process node.

[0009] In one possible implementation, the process engine module is specifically configured to: if the target status is "passed" or "queried", determine the next process node of the currently executed process node as the processing process node according to the process order in the process configuration file; if the target status is "rejected", determine the previous process node of the currently executed process node as the processing process node according to the process order in the process configuration file; if the target status is "transferred", "pushed", or "expedited", determine the process node corresponding to the specified user who transferred, pushed, or expedited the process as the processing process node according to the process order in the process configuration file.

[0010] In one possible implementation, the process execution agent is specifically configured to: if the node type of the target process node is a non-human process node, determine the second target operation as follows: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process pre-configured in the process configuration file for the target process node; execute the specific operation process to obtain result data; if the node type of the target process node is a human process node, determine the second target operation as follows: using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the preset result data of the target process node corresponding to the second identifier, match the target configuration data field that matches the identifier field, and determine the second parameter value corresponding to the target configuration data field as the result data.

[0011] In one possible implementation, the process engine module is further configured to: modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter using the first parameter value of the first attribute parameter; determine the configuration data field that matches the event payload field and the corresponding target parameter value based on the configuration data field of the target process node and the modified parameter value of the configuration data field of the target process node, according to the event payload field that needs to be filled in corresponding to the preset hook event; write the target parameter value into the corresponding event payload field that needs to be filled in for the hook event; and determine the event payload field and the target parameter value as the information of the hook event. The process execution agent is specifically used to, when the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

[0012] In one possible implementation, the device further includes: a page rendering engine module; The process execution agent is further configured to obtain the page configuration of the target process node of the second identifier from a pre-saved process configuration file based on the second identifier; and send the result data and the page configuration to the page rendering engine module. The page rendering engine module is used to receive the result data and the page configuration; generate a rendered page based on the result data and the page configuration; and push the rendered page to the user interface.

[0013] Secondly, embodiments of this application provide a workflow execution method applied to a process execution intelligent agent, the method comprising: Determine the intent category corresponding to the text input in the user interface; Determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; send the first identifier, the first attribute parameter, and the first parameter value to the process engine module; Receive information about hook events sent by the process engine module, wherein the information about the hook events includes a second identifier of the target process node; Based on the second identifier, the node type of the target process node of the second identifier is obtained from the pre-saved process configuration file; wherein, the second identifier is obtained by the process engine module determining the operation type of the engine tool of the first identifier based on receiving the first identifier, the first attribute parameter and the first parameter value; determining the first target operation based on the operation type and the operation corresponding to the preset operation type, and executing the first target operation according to the first parameter value to determine the second identifier of the target process node; Based on the node type and the preset operation corresponding to the node type, a second target operation is determined and executed; wherein, the process configuration file includes the process nodes for executing the workflow and the node type and identifier of each process node.

[0014] In one possible implementation, determining the intent category corresponding to the text information input into the user interface includes: For each pre-saved intent category corresponding to at least one example text, determine the similarity between each example text and the text information; and take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0015] In one possible implementation, the first parameter value for determining the first attribute parameter required to execute the intent category includes: The text information and the first attribute parameter are added to a preset template to construct a prompt word. The prompt word and the text information are input into a trained large model. Based on the large model, the first parameter value corresponding to the first attribute parameter is extracted from the text information.

[0016] Before sending the first identifier, the first attribute parameter, and the first parameter value to the process engine module, the method further includes: Determine whether all the first attribute parameters required for the intent category have corresponding first parameter values. If not, determine the second attribute parameters that are missing first parameter values. For each second attribute parameter, fill the second attribute parameter into a preset general template to generate corresponding follow-up text and display it to indicate the second attribute parameter with missing first parameter values. Receive the first parameter value of the input second attribute parameter.

[0017] In one possible implementation, determining the second target operation based on the node type and a preset operation corresponding to the node type, and then executing the second target operation, includes: If the node type of the target process node is a non-manual process node, then the second target operation is determined to be: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process of the target process node pre-configured in the process configuration file; execute the specific operation process to obtain result data; If the node type of the target process node is a manual process node, then the second target operation is determined to be to use the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the target process node's preset result data corresponding to the second identifier, match the target configuration data field that is consistent with the identifier field, and determine the second parameter value corresponding to the target configuration data field as the result data.

[0018] In one possible implementation, determining the second target operation based on the node type and a preset operation corresponding to the node type, and then executing the second target operation, includes: When the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data. Specifically, the hook event information is obtained by the process engine module using the first parameter value of the first attribute parameter to modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the preset event payload field corresponding to the hook event, and based on the configuration data field of the target process node and the modified parameter value of the configuration data field of the target process node, determining the configuration data field that matches the event payload field and the corresponding target parameter value; writing the target parameter value into the corresponding event payload field that needs to be filled in the hook event; and determining the event payload field and the target parameter value as the hook event information.

[0019] In one possible implementation, after performing the second target operation, the method further includes: Based on the second identifier, the page configuration of the target process node of the second identifier is obtained from the pre-saved process configuration file; the result data and the page configuration are sent to the page rendering engine module so that the page rendering engine module receives the result data and the page configuration; based on the result data and the page configuration, a rendered page is generated and the rendered page is pushed to the user interface.

[0020] Thirdly, embodiments of this application provide a workflow execution method applied to a process engine module, the method comprising: The process execution agent receives a first identifier, a first attribute parameter, and a first parameter value sent by the process execution agent. The first identifier, the first attribute parameter, and the first parameter value are obtained by the process execution agent in determining the intent category corresponding to the text information input by the user interface; determining the first identifier of the engine tool corresponding to the intent category; and determining the first parameter value of the first attribute parameter required to execute the intent category. Determine the operation type of the engine tool identified by the first identifier; Based on the operation type and the operation corresponding to the preset operation type, a first target operation is determined, and the first target operation is executed according to the first parameter value to determine the second identifier of the target process node; The hook event information containing the second identifier is sent to the process execution agent, so that the process execution agent receives the hook event information, obtains the node type of the target process node of the second identifier from the pre-saved process configuration file according to the second identifier, determines the second target operation according to the node type and the preset operation corresponding to the node type, and executes the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

[0021] In one possible implementation, determining a first target operation based on the operation type and the operation corresponding to the preset operation type, and determining the target process node by executing the first target operation according to the first parameter value includes: If the operation type is new creation, then the first target operation is determined to be to determine the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database according to the first parameter value, and to modify the node status of the first process node in the target workflow to new creation, and to determine the first process node as the target process node; If the operation type is not newly created, then the first target operation is determined to be to modify the node state of the currently executed process node to the target state corresponding to the operation type according to the state corresponding to the preset operation type, and to determine the processing process node corresponding to the target state according to the process sequence in the process configuration file and the target state, and to determine the processing process node as the target process node.

[0022] In one possible implementation, determining the processing flow node corresponding to the target state based on the flow sequence in the flow configuration file and the target state includes: If the target status is passed or queried, then the next process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is rejected, then the previous process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is transferred, pushed, or expedited, then the process node corresponding to the specified user who transferred, pushed, or expedited the task will be determined as the processing process node according to the process sequence in the process configuration file.

[0023] In one possible implementation, before sending information about the hook event containing the second identifier to the process execution agent, the method further includes: Using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; Based on the event payload field that needs to be filled in corresponding to the preset hook event, and based on the configuration data field of the target process node and the parameter value of the modified configuration data field of the target process node, determine the configuration data field that matches the event payload field and the corresponding target parameter value. Write the target parameter value into the corresponding event payload field that needs to be filled in for the hook event; The event payload field and the target parameter value are determined as the information of the hook event, so that when the node type of the target process node is a manual process node, the second target operation of the process execution agent is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the hook event information, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

[0024] Fourthly, this application provides an electronic device, which includes at least a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the steps of the methods described in the second and third aspects above.

[0025] Fifthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the methods described in the second and third aspects.

[0026] In a sixth aspect, this application provides a computer program product that, when invoked by a computer, causes the computer to perform the methods described in the second and third aspects.

[0027] In this embodiment, the process execution agent determines the intent category corresponding to the user-input text information, and determines the first identifier of the engine tool corresponding to the intent category, as well as the first parameter value of the first attribute parameter required to execute the intent category. The process engine module determines the second identifier of the target process node based on the first identifier, the first attribute parameter, and the first parameter value, and sends the hook event information containing the second identifier to the process execution agent. The process execution agent determines the second target operation and executes it based on the node type of the target process node and the preset operation corresponding to the node type. This allows the process engine module to focus only on receiving standardized tool identifiers and executing the corresponding target operations, while leaving the business rules that users can easily change to the process execution agent. This allows the entire workflow to flexibly control the process nodes that need to be executed during the execution process. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A schematic diagram of the structure of a workflow execution device provided in an embodiment of this application; Figure 2 This application provides an embodiment of a low-code flow canvas that may have pre-configured flow nodes; Figure 3 An overall architecture diagram of a workflow execution device provided in this application embodiment; Figure 4 An example diagram illustrating the interaction timing of various modules in a device, provided for an embodiment of this application; Figure 5 A flowchart illustrating the execution process of an active monitoring mechanism provided in this application; Figure 6 This application provides a schematic diagram of workflow execution in a natural gas dispatching scenario. Figure 7 This application provides a schematic diagram of a workflow execution method according to an embodiment of the present application. Figure 8 This is a schematic diagram illustrating another workflow execution method provided in an embodiment of this application; Figure 9 This is a schematic diagram of an electronic device structure provided in an embodiment of this application. Detailed Implementation

[0030] To make the objectives and implementation methods of this application clearer, the exemplary implementation methods of this application will be clearly and completely described below with reference to the accompanying drawings of the exemplary embodiments of this application. Obviously, the exemplary embodiments described are only some embodiments of this application, and not all embodiments.

[0031] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0032] The terms "first," "second," "third," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar or related objects or entities, and do not necessarily imply a specific order or sequence, unless otherwise specified. It should be understood that such terms are interchangeable where appropriate.

[0033] The terms “comprising” and “having”, and any variations thereof, are intended to cover but not exclude inclusion, for example, a product or device that includes a range of components is not necessarily limited to all of the components that are clearly listed, but may include other components that are not clearly listed or that are inherent to such product or device.

[0034] The term "module" refers to any known or subsequently developed hardware, software, firmware, artificial intelligence, fuzzy logic, or combination of hardware and / or software code that is capable of performing the functions associated with that element.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0036] For ease of explanation, the above description has been provided in conjunction with specific embodiments. However, the above exemplary discussion is not intended to be exhaustive or to limit the embodiments to the specific forms disclosed above. Various modifications and variations can be obtained based on the above teachings. The selection and description of the above embodiments are for the purpose of better explaining the principles and practical applications, thereby enabling those skilled in the art to better utilize the described embodiments and various different variations of embodiments suitable for specific use considerations.

[0037] Before introducing the workflow execution device provided in the embodiments of this application, some terms used in the embodiments of this application will be explained to facilitate understanding by those skilled in the art.

[0038] Workflow, also known as workflow, refers to a process designed for a specific business goal or task, which includes various process nodes. The workflow includes the sequential relationship between these process nodes, i.e., the process sequence.

[0039] A process execution agent refers to a software or automated program capable of determining and executing preset operations or triggering other modules. In the embodiments of this application, it acts as the core interaction and coordination role between the user and the process engine, responsible for intent recognition, parameter extraction, process node type determination, and node operation execution.

[0040] The process engine module is a core component specifically responsible for managing, scheduling, and executing workflows. It can operate specific engine tools and determine the next process node to be executed according to the preset process sequence.

[0041] A user interface (UI) is a graphical or command-line interface through which a user interacts with a device. Users input text information through this interface.

[0042] Example 1: Figure 1 This is a schematic diagram of the structure of a workflow execution device provided in an embodiment of this application, such as... Figure 1 As shown, the device 100 includes a process execution agent 101 and a process engine module 102.

[0043] In this embodiment, all workflows in device 100 are pre-configured. When generating the pre-configured workflows, a trained process design agent can be used to automatically generate a JSON-formatted process configuration file based on the user's actual needs. For example, the user's workflow requirement described in natural language might be "I need a daily natural gas scheduling process, including five steps: data acquisition, scheme generation, approval, execution, and monitoring." The process design agent automatically generates a JSON-formatted process configuration file based on this requirement. This configuration file includes the workflow name, each process node, the process order between each process node, and the node identifier, node type, designated user, page configuration, constraints, etc., for each process node. The workflow name is a string used to identify and distinguish the workflow. How the agent automatically generates the workflow configuration file based on the user's actual needs is existing technology and will not be elaborated here. In other words, the process design agent assists the user in completing low-code configuration of the workflow using natural language, generating a JSON-formatted process configuration file.

[0044] In one possible implementation, after generating a workflow, the process design agent can send the workflow's configuration file to a pre-defined page rendering engine module. This module then renders the page based on the configuration file, generates a flowchart, and sends the flowchart to a low-code workflow canvas. How the page rendering engine module generates the flowchart from the configuration file is existing technology and will not be elaborated here. The low-code workflow canvas displays the received workflow's corresponding flowchart on the user interface. If the user reviews the generated workflow and determines that it needs modification, they can use the pre-configured process nodes in the low-code workflow canvas to modify the flowchart, thereby generating a flowchart that meets the requirements. Subsequently, the low-code workflow canvas sends this compliant workflow flowchart to the process design agent, which then generates a corresponding JSON-formatted process configuration file based on the modified workflow flowchart.

[0045] The pre-configured process nodes include manual process nodes and non-manual process nodes. Manual process nodes are those that require human intervention and need to display the task on the user interface of a specified user. Non-manual process nodes are those that do not require human intervention and can be executed autonomously by the process execution agent. Figure 2 This application provides an embodiment of a low-code flow canvas that may have pre-configured flow nodes. For example... Figure 2 As shown, non-manual process nodes include process control nodes, such as start nodes, end nodes, sub-processes, exclusive gateways, API calls, data collection, script execution, variable aggregation, data display, etc.; manual process nodes include approval nodes, filling nodes, countersigning nodes, copy nodes, etc.

[0046] During the actual execution phase of the workflow, the process execution agent 101 is used to determine the intent category corresponding to the input text information; determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; and send the first identifier, the first attribute parameter and the first parameter value to the process engine module.

[0047] To enable the workflow engine to possess semantic understanding capabilities during workflow execution, allowing it to process user's natural language commands and improve user experience, the workflow execution intelligent agent 101 receives text information input by the user on the user interface and determines the intent category corresponding to the text information. This text information is the user's natural language command, such as "Generate tomorrow's natural gas scheduling plan" or "There's a problem, return it." In this embodiment, the text information can be input by the user via voice or text on the user interface, without limitation. The user interface for inputting text information is deployed on the front-end display interface of device 100.

[0048] Intent categories refer to category labels obtained by classifying the user's purpose or need expressed in the text information entered by the user. For example, intent categories can be new creation, rejection, approval, transfer, query, and expediting.

[0049] In one possible implementation, when determining the intent category corresponding to text information, it can be determined by precisely matching pre-saved specific words using regular expressions. For example, the pre-saved specific words might be "rejection," "return," or "generate." The regular expression is used to determine whether the user-input text information contains these pre-saved specific words. If it does, the contained target specific words are extracted, and the intent category corresponding to the target specific word is determined based on the preset relationship between the specific words and intent categories. The intent category corresponding to the target specific word is then determined as the intent category corresponding to the text information. It should be noted that this application does not limit how the intent category corresponding to the text information is determined; any method that can determine the intent category is acceptable.

[0050] After determining the intent category corresponding to the text information, the first identifier of the engine tool corresponding to that intent category is determined based on that intent category and the pre-saved correspondence between intent categories and engine tool identifiers. An engine tool refers to an independent functional unit or plug-in within the process engine module capable of performing specific business operations, such as a query engine tool or a reminder engine tool. The engine tool identifier is an identifier used to uniquely identify and locate a specific engine tool within the process engine module.

[0051] In this embodiment, the process execution agent 101 pre-stores the identifiers of each intent category and the corresponding engine tool. The process execution agent 101 completes its interaction with the process engine module 102 by sending the first identifier of the engine tool to the process engine module 102. Table 1 shows the intent categories, engine tool identifiers, and descriptions of the functions corresponding to each engine tool that the process execution agent provided in this application may store.

[0052]

[0053] Table 1

[0054] As shown in Table 1, the process execution agent pre-stores the identifiers and descriptions of the deployment engine tools corresponding to the intent category "Deployment," the new creation engine tools corresponding to the intent category "New Creation," the query engine tools corresponding to the intent category "Query," the pass engine tools corresponding to the intent category "Pass," the rejection engine tools corresponding to the intent category "Rejection," the transfer engine tools corresponding to the intent category "Transfer," the push engine tools corresponding to the intent category "Push," and the expediting engine tools corresponding to the intent category "Extend," etc. It is worth noting that this application only lists some common engine tools and does not limit the inclusion of the listed engine tools or prohibit the use of other engine tools. In practical applications, technicians can configure these tools according to their actual needs.

[0055] When determining the first identifier of the engine tool corresponding to an intent category, the first identifier of the engine tool corresponding to the intent category is determined based on the intent category and the pre-saved correspondence between intent categories and engine tool identifiers. For example, if the intent category corresponding to the text information input by the user interface is determined to be "new creation", then the first identifier of the newly created engine tool corresponding to the intent category is determined based on the intent category and the pre-saved correspondence between intent categories and engine tool identifiers. For example, the first identifier of the newly created engine tool is workflow.start_instance or newly created engine tool y.

[0056] To accurately execute the workflow, it is also necessary to define the first attribute parameters required for executing the intent category. These first attribute parameters refer to the input parameters of the engine tool needed to execute the intent corresponding to each intent category. For example, for a newly created engine tool, its first attribute parameters include the scenario; for a rejection engine tool, its first attribute parameters include the reason, etc. The first attribute parameters required for each intent category can be designed by technical personnel according to actual needs, and are not limited thereto. Table 2 shows the possible required first attribute parameters for each intent category provided in this application.

[0057]

[0058] Table 2

[0059] As shown in Table 2, the process execution agent pre-stores the first attribute parameter corresponding to each intent category.

[0060] The process execution agent 101 is specifically used to add text information and the first attribute parameter to a preset template to construct prompt words, input the prompt words and text information into the trained large model, and extract the first parameter value corresponding to the first attribute parameter from the text information based on the large model.

[0061] The first parameter value refers to the specific value or content corresponding to the first attribute parameter. For example, if the first attribute parameter is a scenario, the specific content corresponding to the scenario may be natural gas dispatch; if the first attribute parameter is a cause, the specific content corresponding to the cause may be incomplete materials.

[0062] In this embodiment, the process execution agent 101 determines the first attribute parameter corresponding to the intent category based on the determined intent category and the pre-saved first attribute parameter corresponding to each intent category; it then adds the text information and the first attribute parameter to a preset template to construct a prompt word, and inputs the prompt word and text information into a trained large model. Based on the large model, it extracts the first parameter value corresponding to the first attribute parameter from the text information. The preset template is designed according to actual needs.

[0063] For example, if the preset template is "Extract the first parameter value corresponding to the first attribute parameter from [text information]", and the text information is "Rejected, materials incomplete, return to Zhang San", then based on the determined intent category of "Rejection" and the pre-saved first attribute parameters corresponding to each intent category, the first attribute parameters corresponding to the "Rejection" intent are determined to be [Reason] and [Specified User]. Then, a prompt word 'a' is generated: "Extract the first parameter value corresponding to [Reason] and [Specified User] from [Rejection, materials incomplete, return to Zhang San]". This prompt word 'a' and "Rejection, materials incomplete, return to Zhang San" are input into the trained large model. Based on this large model, the first parameter value corresponding to [Reason] is extracted as "Materials incomplete", and the first parameter value corresponding to [Specified User] is extracted as Zhang San. The output of the large model is: the first parameter value corresponding to [Reason] is "Materials incomplete", and the first parameter value corresponding to [Specified User] is "Zhang San". The large model was trained using a large number of training samples. Each training sample included text information, the corresponding first attribute parameter and the first parameter value, and the standard output result.

[0064] The process execution intelligent agent 101 sends the determined first identifier, first attribute parameter and first parameter value to the process engine module 102.

[0065] The process engine module 102 is used to receive the first identifier, the first attribute parameter, and the first parameter value; determine the operation type of the engine tool of the first identifier; determine the first target operation according to the operation type and the operation corresponding to the preset operation type; execute the first target operation according to the first parameter value to determine the second identifier of the target process node; and send the hook event information containing the second identifier to the process execution agent.

[0066] The process engine module 102 receives the first identifier, the first attribute parameter, and the first parameter value sent by the process execution agent 101.

[0067] In this embodiment, the operation type includes new creation and non-new creation. Operations of non-new creation type refer to operations with intent categories such as approval or rejection. Operations of new creation type refer to operations with intent categories such as creating a new workflow. In this embodiment, creating a new workflow essentially refers to the workflow engine module starting a pre-deployed workflow that has not yet run. The workflow engine module 102 determines the engine tool corresponding to the received first identifier and determines the operation type of that engine tool. For example, if the first identifier is "workflow.start_instance", then the engine tool corresponding to the first identifier is determined to be a new creation engine tool, and the operation type of that new creation engine tool is determined to be new creation.

[0068] The process engine module 102 pre-stores operations corresponding to different operation types. Based on the determined operation type and the preset operation corresponding to the operation type, the first target operation corresponding to the operation type is determined; the first target operation is executed according to the received first parameter value, and the second identifier of the target process node is determined.

[0069] For example, if the determined operation type is "new creation," and the preset operation type is also "new creation," the corresponding first target operation is as follows: Based on the received first parameter value, determine the target workflow stored in the preset database that has the highest semantic similarity to the first parameter value, modify the node status of the first process node in the target workflow to "new creation," determine the first process node as the target process node, and obtain the second identifier of the target process node. Thus, by executing this first target operation, the second identifier of the target process node is determined.

[0070] In this embodiment, the first target operation is a specific operation pre-set according to different operation types. The first target operation corresponding to each operation type will be described in detail in subsequent embodiments, and will not be repeated here.

[0071] A hook event is a mechanism that triggers before, after, or when a result is returned from a specific operation, and is used to return data to the caller. The timing of the hook event triggering is preset based on the actual situation. In this embodiment, the hook event is used to return completion information to the process execution agent after the first target operation is executed.

[0072] The process engine module 102 constructs information for a hook event, which includes at least a definite second identifier. This information is then sent back to the process execution agent 101.

[0073] The process execution agent 101 is also used to receive information about the hook event, obtain the node type of the target process node of the second identifier from the pre-saved process configuration file according to the second identifier, determine the second target operation according to the node type and the preset operation corresponding to the node type, and execute the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

[0074] A process configuration file is a structured file that pre-defines the complete workflow logic and is persistently stored for each scenario. This file can be in JSON, XML, or YAML format. To facilitate the execution and invocation of the process execution agent, this embodiment uses a JSON format process configuration file. Furthermore, the process configuration file includes at least the workflow nodes, the identifier of each workflow node, and the node type of each workflow node. The identifier of each workflow node refers to its unique ID.

[0075] The process execution agent 101 receives information about hook events sent by the process engine module 102. Based on the second identifier in the hook event information, it searches for the target process node with the second identifier in the pre-saved process configuration file and obtains the node type of the target process node.

[0076] The process execution agent 101 pre-stores operations corresponding to different node types. Based on the obtained node type of the target process node and the preset operations corresponding to the node type, it determines the second target operation corresponding to that node type and executes the second target operation, thereby completing the recognition and mapping of the user's text information intent into the execution instructions of the process engine module 102. The second target operation corresponding to each node type will be described in detail in subsequent embodiments, and will not be repeated here.

[0077] In this embodiment, the process execution agent determines the intent category corresponding to the user-input text information, and determines the first identifier of the engine tool corresponding to the intent category, as well as the first parameter value of the first attribute parameter required to execute the intent category. The process engine module determines the second identifier of the target process node based on the first identifier, the first attribute parameter, and the first parameter value, and sends the hook event information containing the second identifier to the process execution agent. The process execution agent determines the second target operation and executes it based on the node type of the target process node and the preset operation corresponding to the node type. This allows the process engine module to focus only on receiving standardized tool identifiers and executing the corresponding target operations, while leaving the business rules that users can easily change to the process execution agent. This allows the entire workflow to flexibly control the process nodes that need to be executed during the execution process.

[0078] Example 2: In order to accurately determine the intent category corresponding to the text information input by the user interface, in this embodiment of the application, the process execution agent 101 is specifically used to determine the similarity between each example text and the text information for at least one example text corresponding to each pre-saved intent category; and to take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0079] In this embodiment of the application, in order to accurately determine the intent category corresponding to the text information, at least one intent category is pre-stored, and for each intent category, at least one example text is pre-stored. As shown in Table 3, this embodiment of the application provides possible intent categories and corresponding example texts.

[0080]

[0081] Table 3

[0082] After receiving the text information input from the user interface, the process execution agent 101 calculates the similarity between the text information and each example text for each pre-saved intent category corresponding to at least one example text, determines the similarity between each example text and the text information, and then selects the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0083] For example, if the user interface inputs text information A, and example texts a1 and a2 corresponding to intent a are pre-saved, and example texts b1 and b2 corresponding to intent b are pre-saved, then the similarity of text information A with example texts a1, a2, b1, and b2 is calculated respectively. The similarity between text information A and example text a1 is 0.9, the similarity between text information A and example text a2 is 0.95, the similarity between text information A and example text b1 is 0.6, and the similarity between text information A and example text b1 is 0.8. Since 0.95 is the maximum value, the intent category a to which example text a2 corresponding to 0.95 belongs is selected as the intent category corresponding to text information A.

[0084] In this embodiment of the application, the intent category of the text information is accurately determined by calculating the similarity between the text information input by the user and at least one example text corresponding to each pre-saved intent category, and selecting the intent category to which the example text corresponding to the highest similarity belongs. This avoids the problem of misjudging intent category caused by the diverse expressions of language.

[0085] Example 3: To more accurately determine the intent category corresponding to the text information input by the user interface, based on the above embodiment 2, the process execution agent 101 is further used to determine whether the maximum similarity is greater than a preset threshold; if so, the intent category to which the example text corresponding to the maximum similarity value belongs is determined to be the intent category corresponding to the text information; otherwise, it is determined that it is not recognized, and a prompt to re-enter the text information is output.

[0086] In this embodiment, to more accurately determine the intent category corresponding to the text information input from the user interface, a corresponding preset threshold is saved for each intent category. This preset threshold can also be called a confidence threshold. Table 4 shows the possible intent categories, corresponding example texts, and saved preset thresholds provided in this embodiment.

[0087]

[0088] Table 4

[0089] The preset thresholds for each intent category can be the same or different. These preset thresholds are set by technical personnel based on actual needs.

[0090] The process execution agent 101, based on the intent category of the example text corresponding to the highest similarity selected in Embodiment 2 and the corresponding preset threshold, determines whether the highest similarity is greater than the preset threshold. If so, it proves that the intent category of the example text corresponding to the highest similarity is highly reliable and confirms that the intent category of the example text corresponding to the highest similarity is indeed the intent category of the text information. Otherwise, it proves that the intent category of the example text corresponding to the highest similarity is low reliable and confirms that the text information is not recognized, outputting a prompt to re-enter the text information. This prompt can be in the form of voice or text, for example, outputting the text "Re-enter text information" in the text box of the user input interface.

[0091] In this embodiment of the application, by determining whether the maximum similarity is greater than a preset threshold, if so, it is proven that the intent category of the example text corresponding to the maximum similarity is highly credible, thereby further ensuring the accuracy of the intent category corresponding to the determined text information.

[0092] Example 4: To ensure the accuracy of device execution in workflow execution, based on the above embodiments, the process execution intelligent agent 101 is further configured to determine whether all the first attribute parameters required for the intent category correspond to first parameter values. If not, it determines the second attribute parameters that lack first parameter values, and for each second attribute parameter, it fills the second attribute parameter into a preset general template to generate corresponding follow-up text and displays it to indicate the second attribute parameter that lacks a first parameter value; and receives the first parameter value of the input second attribute parameter.

[0093] In this embodiment, after determining the first parameter value of the first attribute parameter required for the execution intent category, the process execution intelligent agent 101 also determines whether all the determined required first attribute parameters have corresponding first parameter values. If not, it proves that the first parameter values ​​required to execute the intent corresponding to this intent category are incomplete and need to be supplemented by the user. Therefore, the process execution intelligent agent 101 determines the second attribute parameter that is missing a first parameter value. For example, if the text information input by the user interface is "formulate tomorrow's scheduling plan", then the intent category is determined to be "new creation". The first attribute parameters required to execute the intent corresponding to "new creation" include "[scenario]. After extracting the first parameter value corresponding to the first attribute parameter from the text information based on the large model, it is determined that the first parameter value corresponding to the first attribute parameter "[scenario]" does not exist. Therefore, the second attribute parameter that is missing a first parameter value is determined to be "[scenario]. For example, if the text information input in the user interface is "Reject, return to Zhang San", then the intent category is determined to be rejection. The first attribute parameters required to execute the intent corresponding to rejection include [reason] and [specified user]. After extracting the first parameter value corresponding to the first attribute parameter from the text information based on the large model, it is determined that the first parameter value corresponding to the first attribute parameter [reason] does not exist. Therefore, the second attribute parameter that is missing the first parameter value is determined to be [reason].

[0094] For each second attribute parameter, the system fills it into a preset general template to generate corresponding follow-up text, which is then displayed on the user interface to indicate to the user the missing second attribute parameter and to receive the input first parameter value. The preset general template is used to generate the corresponding follow-up text, for example, "Missing content for [second attribute parameter]"; or, "You want [intent category], what is [second attribute parameter]?" Technicians can customize this preset general template according to their needs.

[0095] In one possible implementation, a preset general template can contain multiple second attribute parameters simultaneously. For example, it can contain parameters such as "[second attribute parameter] is missing," "...," "[second attribute parameter] corresponding content is missing," "You want [intent category]," "What is [second attribute parameter]," "...," "[second attribute parameter]?" When multiple second attribute parameters exist, each second attribute parameter is simultaneously filled into the preset general template, generating corresponding follow-up text and displaying it on the user interface. Upon receiving new text information input from the user interface, the new text information and the second attribute parameters are added to the preset template to construct prompts. These prompts and the new text information are then input into the trained large model. Based on the large model, the first parameter value corresponding to the second attribute parameter is extracted from the new text information. The preset template is the same as the preset template where text information and the first attribute parameter are added. The process of extracting the first parameter value corresponding to the second attribute parameter is the same as the process of extracting the first parameter value corresponding to the first attribute parameter described above, and will not be repeated here.

[0096] In this embodiment of the application, the above method ensures that all required first attribute parameters have corresponding first parameter values, thereby further ensuring the accuracy of the device executing the workflow.

[0097] Example 5: This application embodiment specifically defines the first target operation of the process engine module for different operation types, thereby determining the target process node.

[0098] The process engine module 102 is specifically configured to, if the operation type is "new creation", determine the first target operation as follows: based on the first parameter value, determine the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database, modify the node state of the first process node in the target workflow to "new creation", and determine the first process node as the target process node; if the operation type is not "new creation", determine the first target operation as follows: based on the operation type and the state corresponding to the preset operation type, modify the node state of the currently executing process node to the target state corresponding to the operation type, and determine the processing process node corresponding to the target state based on the process order in the process configuration file and the target state, and determine the processing process node as the target process node.

[0099] In this embodiment, the operation type includes new creation and non-new creation. A pre-deployed database stores multiple pre-deployed workflows and their corresponding workflow names. For example, workflow M and its corresponding workflow name are "Natural Gas Dispatch," workflow N and its corresponding workflow name are "Amount Approval," etc. Here, the workflow name refers to the name or unique identifier of a workflow specific to a particular business or scenario.

[0100] If the operation type of the engine tool is determined to be "new creation", then the first target operation is determined to be: based on the received first parameter value, determine the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database, modify the node status of the first process node in the target workflow to "new creation", and determine the first process node as the target process node.

[0101] For example, if the first identifier is "workflow.start_instance" and the first parameter value is "natural gas scheduling", then the engine tool corresponding to the first identifier is determined to be a newly created engine tool, and the operation type of the newly created engine tool is "new creation". The first target operation corresponding to the newly created operation type is determined as follows: based on the workflow name and the first parameter value stored in the preset database, the semantic similarity between each workflow name and the first parameter value is calculated, and the workflow name with the highest semantic similarity and the target workflow corresponding to the workflow name with the highest semantic similarity are determined as workflow M. The node status of the first process node in the target workflow M is modified to "new creation", and the first process node is determined as the target process node.

[0102] In one possible implementation, if multiple first parameter values ​​are extracted for the intent category of "new creation" after the large model is extracted, then for each first parameter value, the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database is determined based on the first parameter value; then, from the multiple target workflows corresponding to the workflow names with the highest semantic similarity, the target workflow corresponding to the workflow name with the highest semantic similarity is selected, thereby modifying the node state of the first process node in the target workflow to "new creation", and determining the first process node as the target process node.

[0103] If the operation type of the engine tool is determined to be "not newly created," then the first target operation is determined to be modifying the node state of the currently executing process node to the target state corresponding to the operation type, based on the operation type and the preset state corresponding to the operation type. The "not newly created" operation types include query, approve, reject, transfer, push, and expedit. For each operation type, the corresponding state is pre-saved; for example, if the operation type is "approved," the corresponding state is "approved"; if the operation type is "rejected," the corresponding state is "rejected."

[0104] Based on the process order and target status in the pre-saved process configuration file, the system determines the processing process node corresponding to the target status and designates it as the target process node. Specifically, the process engine module determines the next process node after the currently executing process node as the processing process node if the target status is "passed" or "queried," according to the process order in the process configuration file. If the target status is "rejected," it determines the previous process node after the currently executing process node as the processing process node, according to the process order in the process configuration file. If the target status is "transferred," "pushed," or "expedited," it determines the first attribute parameter as the target attribute parameter for the specified user, and determines the first parameter value corresponding to the target attribute parameter based on the target attribute parameter and its corresponding first parameter value. Then, according to the process order in the process configuration file, the process node corresponding to the user who transferred, pushed, or expedited the process is designated as the processing process node. The process configuration file pre-configures associated handlers for each process node; for example, each process node is bound to a specified user or a list of specified users.

[0105] Specifically, the process engine module determines the first attribute parameter as the target attribute parameter for the specified user based on the first attribute parameter. Then, based on the target attribute parameter and the corresponding first parameter value, it determines the first parameter value for the target attribute parameter. In other words, the process engine module obtains the specified user from the first parameter value. Then, in the subsequent process nodes of this process node in the process configuration file, it searches for subsequent process nodes where the associated handler matches the specified user, and identifies these subsequent process nodes as the processing process node. For example, suppose an approval process follows this sequence: Process node A (applicant) → Process node B (supervisor approval) → Process node C (financial approval) → Process node D (end). Currently, process node B is being executed. The user performs a "transfer" operation and sets the "specified user" to "Li Si" (Li Si is bound to the handler of process node C in the process configuration file). At this point, the process engine module searches in the subsequent process nodes (C and D) of process node B according to the process sequence. If it finds that the associated handler of process node C matches "Li Si," then process node C is identified as the processing process node. In other words, the C process node should have been after the B process node in the process sequence, but the handover operation still followed the process sequence and accurately matched the node corresponding to the specified user.

[0106] In one possible implementation, when multiple subsequent process nodes are matched, the process engine module can first calculate the minimum number of hops required to reach each successfully matched subsequent process node from the currently executing process node along the process sequence defined in the process configuration file. Then, the process node corresponding to the minimum hop count is determined as the processing process node. For example, if the current process node is A, the subsequent process sequence is A→B→C, and both process nodes B and C match the specified user "Li Si", then process node B has a hop count of 1, and process node C has a hop count of 2. Therefore, process node B is selected as the processing process node. Optionally, when multiple nodes have the same minimum hop count, information indicating that a processing process node cannot be determined can be returned to the process execution agent, allowing the agent to indicate an abnormal situation for the current user.

[0107] In the embodiments of this application, different first target operations are defined for different operation types, thereby accurately determining the target process node and ensuring the accuracy of workflow execution.

[0108] Example 6: This application embodiment specifically defines the process execution agent's second target operation for different node types, thereby obtaining the result data after execution.

[0109] The process execution agent 101 is specifically configured to, if the node type of the target process node is a non-manual process node, determine the second target operation as follows: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process pre-configured in the process configuration file for the target process node; execute the specific operation process to obtain result data; if the node type of the target process node is a manual process node, determine the second target operation as follows: using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the preset result data of the target process node corresponding to the second identifier, match the target configuration data field that matches the identifier field, and determine the second parameter value corresponding to the target configuration data field as the result data.

[0110] In the embodiments of this application, the node types of process nodes include manual process nodes and non-manual process nodes.

[0111] The process execution agent 101 determines the node type corresponding to the target process node based on the node type pre-set for each process node in the configuration process file. If the node type of the target process node is a non-manual process node, the second target operation is determined by identifying the target process node corresponding to the second identifier in the received hook event information and the pre-saved process configuration file, as well as the specific operation process of the target process node in the process configuration file. The specific operation process is then executed to obtain the result data. Non-manual process nodes include, but are not limited to, data acquisition process nodes and scheme generation process nodes. The workflow configuration file pre-configures specific operation procedures for each non-manual process node. For example, the specific operation procedure for the data acquisition process node includes: calling each preset interface and collecting data from the database corresponding to that preset interface, etc. The specific operation procedure for the solution generation process node includes: determining whether the previous process node of the solution generation process node in the workflow is a data acquisition process node. If so, it obtains the target data collected by the data acquisition process node, and according to the preset vector similarity retrieval rules, obtains historical data with a semantic vector similarity greater than a threshold with the target data from the preset historical domain knowledge base corresponding to the workflow, as well as obtains the hard safety rules stored in the historical domain database. Based on the historical data, hard safety rules, and target data collected by the data acquisition process node, a solution is generated by comprehensively using a preset large model, etc. For specific industries, such as natural gas dispatching, a domain expert knowledge base, i.e., a historical domain knowledge base, is injected to improve the accuracy of the solution generation by the process execution agent. It is worth noting that the specific operation procedures pre-configured for each non-manual process node in the workflow configuration file can be set by those skilled in the art according to the actual situation, and this application does not limit them to the specific operation procedures mentioned above.

[0112] If the node type corresponding to the target process node is a manual process node, then the second target operation is determined to be the first parameter value of the first attribute parameter, and the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter is modified; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the target process node's preset result data corresponding to the second identifier, the target configuration data field that matches the identifier field is matched, and the second parameter value corresponding to the target configuration data field is determined as the result data.

[0113] Specifically, based on the first attribute parameter corresponding to the determined intent category, the field of the first attribute parameter that matches the configuration data field of the target process node is determined; for ease of description later, this can be referred to as the matching field. Then, based on the first parameter value of the first attribute parameter, the default parameter value of the matching field is modified. For each manual process node, the configuration data field and its default parameter value are pre-configured in the process configuration file. For example, if a manual process node W is for review by a specified user w, the configuration data fields of the manual process node W pre-saved in the process configuration file include [ID of specified user w], and optionally [Field 1], [Field 2], etc. In this embodiment, for configuration data fields that do not match, the original default parameter values ​​are retained.

[0114] The process configuration file also includes an identifier field for the result data to be displayed for each manual process node. After modifying the default parameter values ​​of the configuration fields, based on the modified configuration data fields and their corresponding second parameter values, and the identifier field for the result data of the target process node corresponding to the second identifier, the target configuration data field that matches the identifier field is matched, and the second parameter value corresponding to the target configuration data field is determined as the result data.

[0115] In this embodiment of the application, the manual process nodes and non-manual process nodes are distinguished by the above method, so that the process execution intelligent agent can autonomously execute non-manual process nodes, while ensuring that the manual process nodes that need to be strictly configured according to the process are handled manually, thereby improving the execution efficiency of the workflow.

[0116] Example 7: To improve the execution efficiency of the process execution agent and avoid inconsistencies between the first parameter value used by the process execution agent and the first parameter value used by the process engine module, the process engine module 102 of this application is further configured to: modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter using the first parameter value of the first attribute parameter; determine the configuration data field matching the event payload field and the corresponding target parameter value based on the configuration data field of the target process node and the modified parameter value of the configuration data field of the target process node, according to the event payload field to be filled in corresponding to the preset hook event; write the target parameter value into the corresponding event payload field to be filled in for the hook event; and determine the event payload field and the target parameter value as the information of the hook event. The process execution agent 101 is specifically used to, when the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the preset result data of the target process node corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

[0117] In this embodiment, the process engine module 102 uses the first parameter value of the first attribute parameter to modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter. This process is consistent with the process of modifying the default parameter value of the configuration data field of the target process node described in Embodiment 6 above, and will not be repeated here.

[0118] The process engine module 102 pre-stores the event payload fields that need to be filled in for each hook event. The event payload fields are the fields required for the process execution agent to generate result data. For example, when the intent category is transfer, the process engine module determines that the event payload field that needs to be filled in for the hook event is the specified user. Table 5 shows the hook events that the process engine module provided in this application may store, the triggering time of the hook events, and the event payload fields that may correspond to each hook event.

[0119]

[0120] Table 5

[0121] As shown in Table 5, the table includes possible hook events, the triggering time of the hook events, and the preset event payload fields corresponding to the hook events. Among them, the preset event payload fields corresponding to the hook events include the identifier field of the preset result data of the corresponding target process node.

[0122] Based on the preset event payload fields required for the hook event, and the parameter values ​​of the configuration data fields of the target process node and the modified configuration data fields of the target process node, a configuration data field matching the event payload field is determined, and the target parameter value corresponding to this matching configuration data field is determined. The target parameter value is written into the corresponding event payload field required for the hook event; the event payload field and the target parameter value are then used to define the hook event information. For example, if the preset event payload fields required for the hook event include "[Specified User]", and the parameter value of the configuration data field "[Specified User]" of the target process node and the modified configuration data field "[Specified User]" is "Zhang San", then the configuration data field matching the event payload field is determined to be "[Specified User]", and the target parameter value corresponding to "[Specified User]" is determined to be "Zhang San". "Zhang San" is written into the corresponding event payload field "[Specified User]" required for the hook event; the event payload field "[Specified User]" and the target parameter value "Zhang San" are then used to define the hook event information.

[0123] When the node type of the target process node is a manual process node, the second target operation of the process execution intelligent agent 101 is to directly match the target event payload field that matches the target event payload field with the target parameter value corresponding to the event payload field and the identifier field of the target process node preset result data corresponding to the second identifier in the information of the hook event, and determine the target parameter value corresponding to the target event payload field as the result data.

[0124] In this embodiment, the specific parameter values ​​used by the process engine module are sent to the process execution agent, so that the process execution agent can directly use the specific data of the process engine module to determine the result data to be displayed by the manual process node, thereby improving the execution efficiency of the process execution agent and avoiding the occurrence of inconsistencies between the first parameter values ​​used by the process execution agent and the first parameter values ​​used by the process engine module.

[0125] Example 8: To enhance the flexibility of user interaction during workflow execution and enable users to perceive the execution results of each process node in the workflow in real time, the workflow execution device of this application also includes: a page rendering engine module; The process execution agent is further configured to obtain the page configuration of the target process node of the second identifier from a pre-saved process configuration file based on the second identifier; and send the result data and the page configuration to the page rendering engine module. The page rendering engine module is used to receive the result data and the page configuration; generate a rendered page based on the result data and the page configuration; and push the rendered page to the user interface.

[0126] The page rendering engine module 103 connects to the process execution agent 101 via bidirectional communication (WebSocket). In one possible implementation, the communication method could also be one-way push technology (Server-SentEvents, SSE) or subscription message queue (RabbitMQ / Kafka).

[0127] In this embodiment, when the process engine module 102 triggers a hook event, the process execution agent 101 obtains the page configuration of the target process node of the second identifier from the pre-saved process configuration file according to the received second identifier, and sends the result data obtained from executing the second target operation and the page configuration to the page rendering engine module 103 through bidirectional communication.

[0128] The page rendering engine module 103 generates a rendered page based on the received result data and the page configuration, and pushes the rendered page to the user interface. Specifically, it determines the page layout and display components based on the page configuration; for each display component, it matches the field name of the display component with the field name corresponding to the field in the result data, obtains the value corresponding to the field name, and fills the corresponding value into the display component, rendering the display component; based on all rendered display components and the layout, it generates a rendered page and pushes the rendered page to the user interface. This rendered page can also be called the interactive interface. The process execution agent saves the result data in the form of the result data and its corresponding fields.

[0129] In one possible implementation, the execution behavior corresponding to each hook event and the user-perceived content template are pre-stored in the process execution agent. While the page rendering engine module generates the rendered page based on the result data and page configuration, and pushes the rendered page to the user interface, the process execution agent can also notify the user of the current hook event based on the pre-stored execution behavior and user-perceived content template, allowing the user to perceive the current hook event through text information. Specifically: the process execution agent 101 determines the target execution behavior and the corresponding user-perceived content template based on the received hook event and the pre-stored execution behavior corresponding to each hook event; then, the process execution agent writes the current workflow status into the user-perceived content template to obtain the user-perceived content, where the current status can be completed, new task, task assigned, task overdue, execution failed, etc.; and executes the target execution behavior, notifying the user of the user-perceived content. For example, the pre-saved execution behavior is "call the notification tool to send a pending notification to the user interface," and the corresponding user-aware content template is "You have xxx to process." When the process execution agent receives a hook event, it determines the target execution behavior as "call the notification tool to send a pending notification to the user interface" and the corresponding user-aware content template as "You have xxx to process" based on the pre-saved execution behavior corresponding to each hook event. The process execution agent then writes the current workflow state "New Task" into the user-aware content template, obtaining the user-aware content "You have a new task to process," and executes the target execution behavior to notify the user of this user-aware content, i.e., calling the notification tool to send "You have a new task to process" to the user interface. At this point, the user interface receives the notification "You have a new task to process."

[0130] Table 6 shows the hook events that the process execution agent provided in this application may save, the execution behavior of the process execution agent, and the user-perceived content that each hook event may generate through the corresponding template.

[0131]

[0132] Table 6

[0133] As shown in Table 6, the table includes possible hook events, the actions of the process execution agent when each hook event occurs, and the user-perceived content generated using the corresponding template. The possible hook events, the actions of the process execution agent when each hook event occurs, and the user-perceived content generated using the corresponding template can be designed by technical personnel according to actual needs.

[0134] In one possible implementation, after the user interface receives the text information input by the user for the rendered page, it sends the text information to the process execution agent through a two-way communication method, so that the process execution agent can advance the workflow based on the text information.

[0135] In this embodiment of the application, the addition of a page rendering engine module provides feedback on the results of each process node in the workflow to the user interface, which can significantly improve the flexibility of user interaction during workflow execution and enable users to perceive the execution results of each process node in the workflow in real time.

[0136] Example 9: To ensure the accuracy of workflow execution and avoid data writing errors or malicious tampering, the process execution intelligent agent 101 described in this application is further configured to determine the target user based on the target parameter value corresponding to the specified user field in the field corresponding to the result data; verify whether the target user has the permission to operate the target process node according to the target user and the pre-saved user permission list; if so, send the result data and the page configuration to the page rendering engine module; if not, display a prompt on the user interface that the target user does not have the permission to process the target process node.

[0137] The process execution agent 101 pre-stores a list of user permissions, which includes each process node and the identifier of the user with operation permissions for each process node.

[0138] Before sending the result data and page configuration to the page rendering engine module, the process execution intelligent agent 101 also determines the user identifier of the target user based on the target parameter value corresponding to the specified user field in the field corresponding to the result data. For example, if the specified user field in the field corresponding to the result data corresponds to Zhang San, then the target user is determined to be Zhang San.

[0139] Based on the target user and a pre-saved list of user permissions, the system verifies whether the target user has permission to operate the target process node. If so, meaning the target user's user ID is in the user permission list, then the target user has the necessary permissions, and this result data, along with the page configuration, is sent to the page rendering engine module. If not, meaning the target user's user ID is not in the user permission list, then the target user does not have the necessary permissions, and a message indicating that the target user is not authorized to process the target process node is displayed on the user interface.

[0140] In one possible implementation, a process node capable of dynamically assigning target user approval is pre-defined in the configured process configuration file. For this process node, dynamic task allocation rules are pre-defined, whereby these rules are preset based on specific business scenarios. That is, when this process node is activated or executed, the process execution agent automatically determines the target user's identifier using the preset rules governing dynamic task allocation. At this point, user permission verification is no longer required, and the result data and page configuration can be directly sent to the page rendering engine module and displayed on the target user's user interface. Table 7 shows some business scenarios that may exist in this field and their corresponding allocation rules provided by this application.

[0141]

[0142] Table 7

[0143] As shown in Table 7, when setting up process nodes that can be dynamically assigned to target users for review, the design can be based on the allocation rules in Table 7. For example, in a workflow scenario for monetary approval, when configuring data fields for manual process nodes requiring manual review, a pre-set rule for dynamic task allocation can be established: for amounts less than or equal to 10,000, the node can be assigned to the user ID corresponding to the department manager; otherwise, it can be assigned to the user ID corresponding to the finance director. The process configuration file pre-stores different user IDs corresponding to different responsibilities, such as the user IDs corresponding to the department manager and the finance director. When the manual process node is subsequently activated or executed, the process execution agent automatically determines the user ID assigned to the target user based on the current amount, sends the result data and page configuration to the page rendering engine module, and displays it on the target user's user interface.

[0144] In this embodiment of the application, by adding verification of the target user, it is ensured that the workflow is not arbitrarily manipulated during execution, and only users with operation permissions can process the target process nodes.

[0145] Example 10: Figure 3 This is a general architecture diagram of a workflow execution device provided in an embodiment of this application. Figure 3As shown, (1) in the stage of generating a pre-configured workflow: the process designer uses the process design agent to edit the actual requirements through the command line interface (CLI), that is, to perform a general initial construction of the pre-deployed workflow and obtain the process / page configuration JSON. In other words, the process design agent assists the user in completing the low-code configuration of the workflow through natural language and generates the JSON process configuration file corresponding to the workflow. The specific content described in the above embodiment 1 will not be repeated here. At the same time, for the obtained process / page configuration JSON, if the process designer determines that the workflow needs to be changed after reviewing the generated workflow, the low-code execution canvas can be used to modify it through the graphical user interface (GUI), that is, graphical modification, thereby generating a flowchart of the workflow that meets the requirements. Subsequently, the low-code process canvas sends the flowchart of the workflow that meets the requirements to the process design agent. The process design agent generates the corresponding JSON format process configuration file according to the modified workflow flowchart, thereby obtaining the process configuration file corresponding to the workflow. And the process configuration file corresponding to the workflow is pre-configured in the process execution agent. (2) During the actual execution phase of the workflow: The process execution agent receives the completion message (text information) input by the user in the user interface. Based on this text information, it calls the process engine to execute the process, that is, it calls the process engine module to execute the workflow. The process execution agent listens for hook events of the process engine module, that is, the process execution agent receives information about hook events of the process engine module and executes the second target operation corresponding to the hook event, which is the specific content described in Embodiment 1 above, and will not be repeated here. Obtain the result data; and render the result data through the page rendering engine module and push the page to the user interface, that is, translate the technical events of the process engine module into user-perceptible behaviors. This part is the specific content described in Embodiments 1 to 8, and will not be repeated here. In addition, when the user completes the processing of the pushed page in the user interface, the process execution agent triggers a callback to the process engine module based on the text information of the completion message input again, thereby completing the recognition and mapping of the user's natural language operation intention into the technical instructions of the process engine module. Furthermore, in this embodiment, when the process execution agent generates a solution process node in a non-manual process node, a domain knowledge base, namely a historical domain knowledge base, is introduced to inject expert knowledge for a specific industry. Specifically, the scheduling rules, security constraints, historical cases, optimization goals, etc., are determined through unstructured (RGA) retrieval. The domain knowledge base can support online annotation by experts and incremental knowledge updates, which is the specific content described in Embodiment 6 above, and will not be repeated here.

[0146] Figure 4 This is an example diagram illustrating the interaction timing of various modules in a device provided in an embodiment of this application. For example... Figure 4As shown, in the submission scenario, the user enters the text "I'm done, submit" in the user interface. The user interface sends this text to the process execution agent. The process execution agent performs intent recognition and extracts the first parameter value based on the text. It then sends the first identifier of the engine tool corresponding to the intent and the first parameter value to the process engine module. The process engine module determines the first target operation based on the operation type of the engine tool with the first identifier and the operation corresponding to the preset operation type, and determines the second identifier of the target process node. It then sends a hook event containing the second identifier corresponding to the target process node to the process execution agent. The process execution agent performs the second target operation based on the received second identifier and obtains the result data. It then sends the result data and page configuration to the page rendering engine module. The page rendering engine module renders the page based on the result data and page configuration, and sends the rendered page to the user interface. In a rejection scenario, if the user enters the text "Rejected, please have him fill it out again" in the user interface, the user interface sends this text to the process execution agent. The process execution agent, based on this text, performs intent recognition and extracts the first parameter value corresponding to the first attribute parameter. If it determines that the first parameter value corresponding to the first attribute parameter [reason] does not exist, it generates a corresponding follow-up question text based on a preset general template, such as "You want to reject the task, please confirm the reason for rejection." It also receives text input from the user, such as "Incomplete materials." The user interface sends the user-input text to the process execution agent. The process execution agent sends the first identifier of the engine tool corresponding to the intent and the first parameter value to the process engine module. The process engine module, based on the operation type of the engine tool with the first identifier and the operation corresponding to the preset operation type, determines the first target operation and the second identifier of the target process node. It then sends a hook event containing the second identifier corresponding to the target process node to the process execution agent. The process execution agent performs the second target operation based on the received second identifier, obtains the result data, and displays the text prompt "Rejected, applicant notified to fill out again" in the user interface. The specific details described in Examples 1 to 9 above will not be repeated here.

[0147] In one possible implementation, the process execution agent has a pre-set proactive monitoring mechanism, which specifically includes: Timeout reminder: For each process node requiring manual processing, the task creation time is saved after the processing task is sent. The status of each process node waiting for manual processing is periodically scanned. For each process node waiting for manual processing, if the time interval between the current time and the task creation time of the process node is greater than the node's preset timeout threshold, a reminder notification is automatically triggered. In one possible implementation, if the task is still not processed after the first reminder and a preset waiting time has elapsed, the corresponding pre-saved superior is notified.

[0148] Anomaly localization: When any process node fails, the process execution agent automatically analyzes the error type, such as network timeout, data verification failure, insufficient permissions, or unavailable dependent services, and generates a structured fault report, including: the faulty node, the cause of the error, the scope of impact, remediation suggestions, and a one-click retry entry. The automatic error type analysis and structured fault report generation by the process execution agent is existing technology and will not be elaborated upon here.

[0149] Process query: Any relevant person can query the progress of the process using natural language, and the process execution agent will reply after obtaining the real-time status from the process engine module.

[0150] Figure 5 This is a flowchart illustrating the execution process of an active monitoring mechanism provided in this application. Figure 5 As shown, the process execution intelligence agent proactively monitors each process node in the workflow. This primarily involves: for running process nodes, proactively monitoring for execution failures; if failures occur, automatically analyzing the error type and generating a structured fault report; for process nodes awaiting manual processing, if the time interval between the current time and the task creation time of that node exceeds the node's pre-set timeout threshold, automatically triggering a reminder notification; if the task remains unprocessed after the first reminder and a preset waiting time, notifying the superior according to escalation rules for handling; and ending the monitoring task upon rejection or approval.

[0151] In this embodiment of the application, the traditional passive waiting operation is upgraded to active perception and intervention through the active monitoring mechanism of the process execution intelligent agent. Specifically, the timed-out task is automatically triggered to expedite the process without the need for manual monitoring by maintenance personnel; when a node fails to execute, the error cause is automatically analyzed and an actionable repair suggestion is given, reducing the problem location time from about 30 minutes to about 5 minutes.

[0152] In addition, all process operations, including natural language operation records, can be written to the audit log to meet corporate compliance requirements.

[0153] Example 11: To facilitate understanding, a typical application scenario will be used to briefly introduce this application.

[0154] Assume the business scenario involves a city gas company that needs to formulate a natural gas dispatch plan every day, which involves multiple gas sources (pipeline gas, liquefied natural gas LNG, gas storage facilities), multi-level pipeline networks (high pressure, medium pressure, low pressure), and multiple types of users (residential, industrial, commercial, CNG refueling stations).

[0155] Figure 6 The workflow execution diagram provided in this application for a natural gas dispatch scenario is as follows: S601: The user inputs the text message "Start developing tomorrow's natural gas dispatch plan".

[0156] S602: The process execution agent determines that the intent category is newly created; and determines the first identifier of the engine tool corresponding to the newly created intent category, as well as the first attribute parameter "natural gas scheduling" required to execute the newly created intent category, and sends the first identifier, the first attribute parameter and the first parameter value to the process engine module so that the process engine module determines the target process node "data acquisition".

[0157] In other words, the process execution intelligent agent 101 calls the engine tool workflow.start_instance("gas-dispatch-daily") in the process engine module through the first identifier to create a new corresponding workflow.

[0158] S603: After the process engine module determines the target process node as data acquisition, the process execution intelligent agent performs data acquisition.

[0159] Upon receiving a hook event containing a second identifier from the process engine module, the second target operation is performed. This second identifier corresponds to a non-manual process node, also known as a dynamic node, within the process (data acquisition) node. Specifically: the collect-data node (dynamic node) is activated; the agent goal and data sources configuration of the target process node are read; the following calls are made sequentially: the SCADA API of the Supervisory Control and Data Acquisition (SCADA) system is used to obtain real-time network data; the meteorological bureau API is used to obtain tomorrow's weather forecast, such as a minimum temperature of -12°C and a cold wave warning; the user system database is used to query historical demand for the same period, such as a predicted 40% increase in demand tomorrow; and the acquisition results are output.

[0160] In one possible implementation, the process execution agent can also notify the user: "Data collection complete. Tomorrow's weather: low of -12°C, cold wave warning issued. Demand is predicted to increase by 40% compared to today; it is recommended to activate the contingency plan in advance."

[0161] S604: After the process engine module determines the next target process node as the generation scheduling scheme, the process execution intelligent agent generates a scheduling scheme based on the collected data.

[0162] In this step, the process execution agent activates and generates a scheduling (optimize-plan) process node, which is a dynamic node. It then loads the domain knowledge base "natural_gas_dispatch," where relevant rules are retrieved: rule-001: Prioritize civilian use during peak periods; rule-004: Low-temperature protection for LNG regasification stations; rule-005: Peak-shaving strategy for gas storage facilities. Similar historical cases are also retrieved, resulting in case-001: Insufficient pipeline pressure during the cold wave in January 2024 (92% similarity). Based on these rules, similar historical cases, and the preset optimization objectives for the process nodes, a scheduling plan is generated.

[0163] In one possible implementation, the process execution agent determines the risk level based on the collected data. The agent then notifies the user: "The scheduling plan has been generated. Due to a cold wave warning for tomorrow, the risk level is medium, requiring approval from the scheduling supervisor."

[0164] S605: After the process engine module determines that the target process node is the risk assessment gateway, it selects a branch, determines the target process node, and triggers a hook event according to the specific process configuration of the preset process node.

[0165] In this process, the branch is selected based on the specific process configuration of the preset process node. This means that the processing process node corresponding to the target state is determined according to the process sequence and target state in the process configuration file, and the processing process node is determined as the target process node.

[0166] For example, when the risk assessment is low, the target process node is a non-manual process node that "automatically issues instructions"; when the risk assessment is medium or high, the target process node is a manual process node, where the node type of the target process node is a manual process node, which can also be called a deterministic node.

[0167] S606: The process execution agent determines the page configuration of the target process node based on the information of the hook event; it sends the result data of the generated scheduling scheme and the page configuration to the page rendering engine module. The page rendering engine module generates the rendered page based on the result data and the page configuration, and pushes the rendered page to the user interface.

[0168] In one possible implementation, the process execution agent can also generate business language notifications, such as: Tomorrow's scheduling plan requires your approval; Plan summary: 60% pipeline gas + 30% LNG + 10% gas storage; Risk warning: Cold wave warning, low temperature may affect LNG gasification efficiency; Recommendation: Start LNG preheating 2 hours in advance; Deadline: 18:00 today (2 hours remaining).

[0169] S607: The process execution agent receives the user's text information, determines the intent category of the text information, the first identifier of the corresponding engine tool for the intent category, and the first attribute parameter required to execute the intent category. The agent sends the first identifier, the first attribute parameter, and the first parameter value to the process engine module so that the process engine module can determine the target process node.

[0170] For example, a dispatch manager (operating via natural language) might say: "The plan looks fine, but could the LNG percentage be increased by another 5%? The gas storage facility should be reserved as a final emergency reserve."

[0171] The process execution agent performs intent recognition: rejection (modify_plan); and determines the first attribute parameter and the first parameter value corresponding to the first attribute parameter, for example, "adjustment": "+5%", "reason": "gas storage to be reserved for emergency use".

[0172] The process engine module calls workflow.reject_task(reason="LNG ratio needs to be adjusted"); and the process engine module determines the target process node to fall back to the previous process node, that is, generates the scheduling plan (optimize-plan) node.

[0173] S608: After the process engine module determines that the next target process node is the generated scheduling scheme, the process execution intelligent agent regenerates the scheduling scheme based on the first parameter value and the generated scheduling scheme; and pushes the approval page again.

[0174] S609: The process execution agent receives the user's text information, determines the intent category of the text information, the first identifier of the corresponding engine tool for the intent category, and the first attribute parameter required to execute the intent category. The agent sends the first identifier, the first attribute parameter, and the first parameter value to the process engine module so that the process engine module can determine the target process node.

[0175] For example, a scheduling supervisor might say, "This plan is acceptable; approve execution." The process execution agent determines that the intent category of this text message is "pass," identifies the first identifier of the corresponding engine tool for this intent category (e.g., workflow.complete_task(task_id, form_data), and determines the first attribute parameters required to execute this intent category (e.g., "opinion": "approved or passed," "signature": "signed"). The agent then sends the first identifier, first attribute parameters, and first parameter values ​​to the process engine module so that the process engine module can determine the target process node.

[0176] S610: After the process engine module determines the next target process node as the instruction to be issued, it will issue the instruction of the approved scheduling scheme and determine the next target process node as the monitoring and execution process node.

[0177] In other words, the process engine executes the approved process node by calling a pre-defined API interface to send scheduling instructions to the SCADA system. This then propels the process to the monitor-execution process node, which is a non-manual process node.

[0178] S611: The process execution intelligent agent performs real-time monitoring based on the preset process configuration data of the monitored execution process node, and automatically adjusts according to the preset operation.

[0179] In other words, the process execution intelligent agent activates and monitors the process execution nodes, collecting pipeline network data in real time. When an anomaly is detected, rule-002: pipeline network pressure safety constraint is triggered, and preset operations are automatically executed. For example, if the pressure in a certain area drops to 0.28 MPa, which is below the safety lower limit of 0.3 MPa, the process execution intelligent agent automatically adjusts the opening of valve V-103 by 20%, triggers an alarm, and notifies the dispatcher: "[Alarm] Pipeline network pressure in XX area is 0.28 MPa. Valve V-103 has been automatically adjusted. Please pay attention to subsequent changes."

[0180] In one possible implementation, once the pressure returns to 0.35 MPa, the process execution agent will notify the scheduler: "The pressure has returned to normal, and the fault has been automatically repaired."

[0181] Compared with traditional workflow execution methods, this application has the following advantages.

[0182]

[0183] In this application embodiment, industry expert knowledge is structured for the first time, i.e., a historical domain knowledge base is injected into the process execution intelligent agent, enabling the process execution intelligent agent to possess expert-level decision-making capabilities in a specific domain, such as natural gas scheduling. Furthermore, hard constraint enforcement verification reduces the violation rate of safety rules, improves decision accuracy, and reduces the workload of experts. Moreover, experts annotate cases online, and the historical domain knowledge base continues to grow, making the device increasingly intelligent with use. In other words, compared to traditional expert systems, this application combines the determinism of the rule-process engine module with the generalization capability of a large language model, enabling it to handle both known and unknown scenarios.

[0184] Example 12: Based on the same technical concept, this application provides a method for workflow execution, such as... Figure 7 As shown, Figure 7 This application provides a schematic diagram of a workflow execution method, which includes the following steps: S701: Determine the intent category corresponding to the text information input into the user interface; S702: Determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; send the first identifier, the first attribute parameter, and the first parameter value to the process engine module; S703: Receive information about a hook event sent by the process engine module, wherein the information about the hook event includes a second identifier of the target process node; S704: Based on the second identifier, obtain the node type of the target process node of the second identifier from the pre-saved process configuration file; wherein, the second identifier is obtained by the process engine module determining the operation type of the engine tool of the first identifier based on receiving the first identifier, the first attribute parameter and the first parameter value; determining the first target operation based on the operation type and the operation corresponding to the preset operation type, and determining the second identifier of the target process node by executing the first target operation according to the first parameter value; S705: Determine the second target operation based on the node type and the preset operation corresponding to the node type, and execute the second target operation; wherein, the process configuration file includes the process nodes for executing the workflow and the node type and identifier of each process node.

[0185] In one possible implementation, determining the intent category corresponding to the text information input into the user interface includes: For each pre-saved intent category corresponding to at least one example text, determine the similarity between each example text and the text information; and take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0186] In one possible implementation, the first parameter value for determining the first attribute parameter required to execute the intent category includes: The text information and the first attribute parameter are added to a preset template to construct a prompt word. The prompt word and the text information are input into a trained large model. Based on the large model, the first parameter value corresponding to the first attribute parameter is extracted from the text information.

[0187] Before sending the first identifier, the first attribute parameter, and the first parameter value to the process engine module, the method further includes: Determine whether all the first attribute parameters required for the intent category have corresponding first parameter values. If not, determine the second attribute parameters that are missing first parameter values. For each second attribute parameter, fill the second attribute parameter into a preset general template to generate corresponding follow-up text and display it to indicate the second attribute parameter with missing first parameter values. Receive the first parameter value of the input second attribute parameter.

[0188] In one possible implementation, determining the second target operation based on the node type and a preset operation corresponding to the node type, and then executing the second target operation, includes: If the node type of the target process node is a non-manual process node, then the second target operation is determined to be: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process of the target process node pre-configured in the process configuration file; execute the specific operation process to obtain result data; If the node type of the target process node is a manual process node, then the second target operation is determined to be to use the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the target process node's preset result data corresponding to the second identifier, match the target configuration data field that is consistent with the identifier field, and determine the second parameter value corresponding to the target configuration data field as the result data.

[0189] In one possible implementation, determining the second target operation based on the node type and a preset operation corresponding to the node type, and then executing the second target operation, includes: When the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data. Specifically, the hook event information is obtained by the process engine module using the first parameter value of the first attribute parameter to modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the preset event payload field corresponding to the hook event, and based on the configuration data field of the target process node and the modified parameter value of the configuration data field of the target process node, determining the configuration data field that matches the event payload field and the corresponding target parameter value; writing the target parameter value into the corresponding event payload field that needs to be filled in the hook event; and determining the event payload field and the target parameter value as the hook event information.

[0190] In one possible implementation, after performing the second target operation, the method further includes: Based on the second identifier, the page configuration of the target process node of the second identifier is obtained from the pre-saved process configuration file; the result data and the page configuration are sent to the page rendering engine module so that the page rendering engine module receives the result data and the page configuration; based on the result data and the page configuration, a rendered page is generated and the rendered page is pushed to the user interface.

[0191] Example 13: Based on the same technical concept, this application provides another method for workflow execution, such as... Figure 8 As shown, Figure 8 This application provides a schematic diagram of another workflow execution method, which includes the following steps: S801: Receive a first identifier, a first attribute parameter, and a first parameter value sent by the process execution agent, wherein the first identifier, the first attribute parameter, and the first parameter value are obtained by the process execution agent in determining the intent category corresponding to the text information input by the user interface; determining the first identifier of the engine tool corresponding to the intent category; and the first parameter value of the first attribute parameter required to execute the intent category. S802: Determine the operation type of the engine tool identified by the first identifier; S803: Based on the operation type and the operation corresponding to the preset operation type, determine the first target operation, and execute the first target operation according to the first parameter value to determine the second identifier of the target process node; S804: Send the hook event information containing the second identifier to the process execution agent, so that the process execution agent receives the hook event information, obtains the node type of the target process node of the second identifier from the pre-saved process configuration file according to the second identifier, determines the second target operation according to the node type and the preset operation corresponding to the node type, and executes the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

[0192] In one possible implementation, determining a first target operation based on the operation type and the operation corresponding to the preset operation type, and determining the target process node by executing the first target operation according to the first parameter value includes: If the operation type is new creation, then the first target operation is determined to be to determine the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database according to the first parameter value, and to modify the node status of the first process node in the target workflow to new creation, and to determine the first process node as the target process node; If the operation type is not newly created, then the first target operation is determined to be to modify the node state of the currently executed process node to the target state corresponding to the operation type according to the state corresponding to the preset operation type, and to determine the processing process node corresponding to the target state according to the process sequence in the process configuration file and the target state, and to determine the processing process node as the target process node.

[0193] In one possible implementation, determining the processing flow node corresponding to the target state based on the flow sequence in the flow configuration file and the target state includes: If the target status is passed or queried, then the next process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is rejected, then the previous process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is transferred, pushed, or expedited, then the process node corresponding to the specified user who transferred, pushed, or expedited the task will be determined as the processing process node according to the process sequence in the process configuration file.

[0194] In one possible implementation, before sending information about the hook event containing the second identifier to the process execution agent, the method further includes: Using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; Based on the event payload field that needs to be filled in corresponding to the preset hook event, and based on the configuration data field of the target process node and the parameter value of the modified configuration data field of the target process node, determine the configuration data field that matches the event payload field and the corresponding target parameter value. Write the target parameter value into the corresponding event payload field that needs to be filled in for the hook event; The event payload field and the target parameter value are determined as the information of the hook event, so that when the node type of the target process node is a manual process node, the second target operation of the process execution agent is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the hook event information, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

[0195] Example 14: Based on the same technical concept, this application also provides an electronic device. Figure 9 This application provides a schematic diagram of an electronic device structure, such as... Figure 9 As shown, it includes: processor 901, communication interface 902, memory 903 and communication bus 904, wherein processor 901, communication interface 902 and memory 903 communicate with each other through communication bus 904. The memory 903 stores a computer program. When the program is executed by the processor 901, the processor 901 performs the following steps: Determine the intent category corresponding to the text input in the user interface; Determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; send the first identifier, the first attribute parameter, and the first parameter value to the process engine module; Receive information about hook events sent by the process engine module, wherein the information about the hook events includes a second identifier of the target process node; Based on the second identifier, the node type of the target process node of the second identifier is obtained from the pre-saved process configuration file; wherein, the second identifier is obtained by the process engine module determining the operation type of the engine tool of the first identifier based on receiving the first identifier, the first attribute parameter and the first parameter value; determining the first target operation based on the operation type and the operation corresponding to the preset operation type, and executing the first target operation according to the first parameter value to determine the second identifier of the target process node; Based on the node type and the preset operation corresponding to the node type, a second target operation is determined and executed; wherein, the process configuration file includes the process nodes for executing the workflow and the node type and identifier of each process node.

[0196] In one possible implementation, the processor 901 is specifically used for: For each pre-saved intent category corresponding to at least one example text, determine the similarity between each example text and the text information; and take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

[0197] In one possible implementation, the processor 901 is specifically used for: The text information and the first attribute parameter are added to a preset template to construct a prompt word. The prompt word and the text information are input into a trained large model. Based on the large model, the first parameter value corresponding to the first attribute parameter is extracted from the text information.

[0198] The processor 901 is also used for: Determine whether all the first attribute parameters required for the intent category have corresponding first parameter values. If not, determine the second attribute parameters that are missing first parameter values. For each second attribute parameter, fill the second attribute parameter into a preset general template to generate corresponding follow-up text and display it to indicate the second attribute parameter with missing first parameter values. Receive the first parameter value of the input second attribute parameter.

[0199] In one possible implementation, the processor 901 is specifically used for: If the node type of the target process node is a non-manual process node, then the second target operation is determined to be: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process of the target process node pre-configured in the process configuration file; execute the specific operation process to obtain result data; If the node type of the target process node is a manual process node, then the second target operation is determined to be to use the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the modified configuration data field and the corresponding second parameter value, and the identifier field of the target process node's preset result data corresponding to the second identifier, match the target configuration data field that is consistent with the identifier field, and determine the second parameter value corresponding to the target configuration data field as the result data.

[0200] In one possible implementation, the processor 901 is specifically used for: When the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data. Specifically, the hook event information is obtained by the process engine module using the first parameter value of the first attribute parameter to modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; based on the preset event payload field corresponding to the hook event, and based on the configuration data field of the target process node and the modified parameter value of the configuration data field of the target process node, determining the configuration data field that matches the event payload field and the corresponding target parameter value; writing the target parameter value into the corresponding event payload field that needs to be filled in the hook event; and determining the event payload field and the target parameter value as the hook event information.

[0201] In one possible implementation, the processor 901 is further configured to: Based on the second identifier, the page configuration of the target process node of the second identifier is obtained from the pre-saved process configuration file; the result data and the page configuration are sent to the page rendering engine module so that the page rendering engine module receives the result data and the page configuration; based on the result data and the page configuration, a rendered page is generated and the rendered page is pushed to the user interface.

[0202] The processor 901 also executes the steps of a workflow execution method on the process engine module side, specifically: The process execution agent receives a first identifier, a first attribute parameter, and a first parameter value sent by the process execution agent. The first identifier, the first attribute parameter, and the first parameter value are obtained by the process execution agent in determining the intent category corresponding to the text information input by the user interface; determining the first identifier of the engine tool corresponding to the intent category; and determining the first parameter value of the first attribute parameter required to execute the intent category. Determine the operation type of the engine tool identified by the first identifier; Based on the operation type and the operation corresponding to the preset operation type, a first target operation is determined, and the first target operation is executed according to the first parameter value to determine the second identifier of the target process node; The hook event information containing the second identifier is sent to the process execution agent, so that the process execution agent receives the hook event information, obtains the node type of the target process node of the second identifier from the pre-saved process configuration file according to the second identifier, determines the second target operation according to the node type and the preset operation corresponding to the node type, and executes the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

[0203] In one possible implementation, the processor 901 is specifically used for: If the operation type is new creation, then the first target operation is determined to be to determine the target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database according to the first parameter value, and to modify the node status of the first process node in the target workflow to new creation, and to determine the first process node as the target process node; If the operation type is not newly created, then the first target operation is determined to be to modify the node state of the currently executed process node to the target state corresponding to the operation type according to the state corresponding to the preset operation type, and to determine the processing process node corresponding to the target state according to the process sequence in the process configuration file and the target state, and to determine the processing process node as the target process node.

[0204] In one possible implementation, the processor 901 is specifically used for: If the target status is passed or queried, then the next process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is rejected, then the previous process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is transferred, pushed, or expedited, then the process node corresponding to the specified user who transferred, pushed, or expedited the task will be determined as the processing process node according to the process sequence in the process configuration file.

[0205] In one possible implementation, the processor 901 is further configured to: Using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter; Based on the event payload field that needs to be filled in corresponding to the preset hook event, and based on the configuration data field of the target process node and the parameter value of the modified configuration data field of the target process node, determine the configuration data field that matches the event payload field and the corresponding target parameter value. Write the target parameter value into the corresponding event payload field that needs to be filled in for the hook event; The event payload field and the target parameter value are determined as the information of the hook event, so that when the node type of the target process node is a manual process node, the second target operation of the process execution agent is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the hook event information, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

[0206] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0207] The communication interface 902 is used for communication between the above-mentioned electronic device and other devices.

[0208] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0209] The processors mentioned above can be general-purpose processors, including central processing units, network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits, field-programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0210] Example 15: Based on the same technical concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a processor. When the program runs on the processor, it causes the processor to execute any of the workflow execution methods described above. Since the principle by which the above-described computer-readable storage medium solves the problem is similar to that of a workflow execution method, the implementation of the above-described computer-readable storage medium can be referred to the implementation of the method, and repeated details will not be elaborated further.

[0211] The aforementioned computer-readable storage medium can be any available medium or data storage device that can be accessed by the processor in an electronic device, including but not limited to magnetic storage such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), optical storage such as CDs, DVDs, BDs, HVDs, etc., and semiconductor storage such as ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs), etc.

[0212] Based on the same technical concept, this application also provides a computer program product, which includes computer program code. When the computer program code is run on a computer, it causes the computer to execute any of the above embodiments. Since the principle of solving the problem by the above computer program product is similar to the method of workflow execution, the implementation of the above computer program product can refer to the implementation of the method, and repeated details will not be described again.

[0213] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0214] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0215] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0216] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0217] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A device for executing workflows, characterized in that, The device includes: A process execution intelligent agent is used to determine the intent category corresponding to the text information input by the user interface; determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; and send the first identifier, the first attribute parameter and the first parameter value to the process engine module. The process engine module is configured to receive the first identifier, the first attribute parameter, and the first parameter value; determine the operation type of the engine tool of the first identifier; determine the first target operation according to the operation type and the operation corresponding to the preset operation type; execute the first target operation according to the first parameter value to determine the second identifier of the target process node; and send information of the hook event containing the second identifier to the process execution agent. The process execution agent is also configured to receive information about the hook event, obtain the node type of the target process node of the second identifier from a pre-saved process configuration file according to the second identifier, determine the second target operation according to the node type and the preset operation corresponding to the node type, and execute the second target operation; wherein, the process configuration file includes the process nodes that execute the workflow and the node type and identifier of each process node.

2. The device according to claim 1, characterized in that, The process execution agent is specifically used to determine the similarity between each example text and the text information for at least one example text corresponding to each pre-saved intent category; and to take the intent category to which the example text with the highest similarity belongs as the intent category corresponding to the text information.

3. The device according to claim 1, characterized in that, The process execution agent is specifically used to add the text information and the first attribute parameter to a preset template to construct a prompt word, input the prompt word and the text information into a trained large model, and extract the first parameter value corresponding to the first attribute parameter from the text information based on the large model; The process execution agent is also used to determine whether all the first attribute parameters required for the intent category have corresponding first parameter values. If not, it determines the second attribute parameters that are missing first parameter values, and for each second attribute parameter, it fills the second attribute parameter into a preset general template to generate corresponding follow-up text and displays it to prompt the second attribute parameter that is missing first parameter value; and receives the first parameter value of the input second attribute parameter.

4. The device according to claim 1, characterized in that, The process engine module is specifically used to determine the first target operation as a target workflow corresponding to the workflow name with the highest semantic similarity to the first parameter value stored in the preset database, based on the first parameter value, if the operation type is new creation, and to modify the node status of the first process node in the target workflow to new creation, and to determine the first process node as the target process node; If the operation type is not newly created, then the first target operation is determined to be to modify the node state of the currently executed process node to the target state corresponding to the operation type according to the state corresponding to the preset operation type, and to determine the processing process node corresponding to the target state according to the process sequence in the process configuration file and the target state, and to determine the processing process node as the target process node.

5. The device according to claim 4, characterized in that, The process engine module is specifically used to determine the next process node of the currently executed process node as the processing process node according to the process order in the process configuration file if the target status is passed or queried. If the target status is rejected, then the previous process node of the currently executed process node is determined as the processing process node according to the process order in the process configuration file. If the target status is transferred, pushed, or expedited, then the process node corresponding to the specified user who transferred, pushed, or expedited the task will be determined as the processing process node according to the process sequence in the process configuration file.

6. The device according to claim 1, characterized in that, The process execution agent is specifically configured to, if the node type of the target process node is a non-human process node, determine the second target operation as follows: based on the second identifier in the hook event information and the process configuration file, determine the target process node corresponding to the second identifier and the specific operation process pre-configured in the process configuration file for the target process node; execute the specific operation process to obtain result data; if the node type of the target process node is a human process node, determine the second target operation as follows: using the first parameter value of the first attribute parameter, modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter. Based on the modified configuration data fields and their corresponding second parameter values, and the identifier field of the target process node preset result data corresponding to the second identifier, the target configuration data field that matches the identifier field is matched, and the second parameter value corresponding to the target configuration data field is determined as the result data.

7. The device according to claim 1, characterized in that, The process engine module is further configured to use the first parameter value of the first attribute parameter to modify the default parameter value in the configuration data field of the target process node that matches the field of the first attribute parameter. Based on the event payload field that needs to be filled in corresponding to the preset hook event, and based on the configuration data field of the target process node and the parameter value of the modified configuration data field of the target process node, determine the configuration data field that matches the event payload field and the corresponding target parameter value. Write the target parameter value into the corresponding event payload field that needs to be filled in for the hook event; determine the event payload field and the target parameter value as the information of the hook event; The process execution agent is specifically used to, when the node type of the target process node is a manual process node, the second target operation is to directly match the target event payload field that is consistent with the identifier field based on the event payload field in the information of the hook event, the target parameter value corresponding to the event payload field, and the identifier field of the target process node preset result data corresponding to the second identifier, and determine the target parameter value corresponding to the target event payload field as the result data.

8. The device according to any one of claims 6 and 7, characterized in that, The device also includes: a page rendering engine module; The process execution agent is further configured to obtain the page configuration of the target process node of the second identifier from a pre-saved process configuration file based on the second identifier; and send the result data and the page configuration to the page rendering engine module. The page rendering engine module is used to receive the result data and the page configuration; generate a rendered page based on the result data and the page configuration; and push the rendered page to the user interface.

9. A workflow execution method, applied to a process execution intelligent agent, characterized in that, The method includes: Determine the intent category corresponding to the text input in the user interface; Determine the first identifier of the engine tool corresponding to the intent category, and the first parameter value of the first attribute parameter required to execute the intent category; send the first identifier, the first attribute parameter, and the first parameter value to the process engine module; Receive information about hook events sent by the process engine module, wherein the information about the hook events includes a second identifier of the target process node; Based on the second identifier, the node type of the target process node of the second identifier is obtained from the pre-saved process configuration file; wherein, the second identifier is obtained by the process engine module based on receiving the first identifier, the first attribute parameter and the first parameter value; determining the operation type of the engine tool of the first identifier; determining the first target operation based on the operation type and the operation corresponding to the preset operation type; and executing the first target operation according to the first parameter value to determine the second identifier of the target process node; Based on the node type and the preset operation corresponding to the node type, a second target operation is determined and executed; wherein, the process configuration file includes the process nodes for executing the workflow and the node type and identifier of each process node.

10. An electronic device, characterized in that, The electronic device includes at least a processor and a memory, the processor being used to implement the method as described in claim 9 when executing a computer program stored in the memory.

Images