Process scheduling and choreography method, device, system and equipment based on big model

Abstract

The present disclosure provides a process scheduling and choreography method, apparatus, system and equipment based on a big model, wherein the method comprises: determining text description information that matches the target business process in response to input question content; determining the step to be executed and the step description information at the current moment through the big model based on the text description information, the question content and the step jump information at the current moment, and determining the step calling interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step; determining the step execution result of the step to be executed by calling the step calling interface, and determining the process execution result of the target business process according to the step execution result of the last executed step, and sending the execution process information of each executed step.

Description

Technical Field

[0001] The present disclosure relates to the technical field of artificial intelligence, and in particular, to a process scheduling and orchestration method, apparatus, system and device based on a large model. Background Art

[0002] At present, different scenarios in professional fields have a set of complex business processes. In order to realize the functions of different scenarios, all processes of the problem content can be generated by intelligent agents. However, due to the complexity of the business processes of different detailed scenarios in professional fields, most of them are a huge inverted tree structure, and there is no large-scale corpus for large model training in proprietary subdivisions, so it is very difficult to generate all processes through intelligent agents, and this technical solution has problems such as low accuracy of output results and unstable output results. Summary of the invention

[0003] The embodiments of the present disclosure at least provide a process scheduling and orchestration method, apparatus, system, and device based on a large model.

[0004] In a first aspect, an embodiment of the present disclosure provides a process scheduling and orchestration method based on a large model, including:

[0005] In response to the input question content, determining text description information matching the target business process;

[0006] Determine the step to be executed at the current moment and the step description information based on the text description information, the question content and the step jump information at the current moment through the big model, and determine the step calling interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step;

[0007] The step execution result of the step to be executed is determined by calling the step calling interface, and the process execution result of the target business process is determined according to the step execution result of the last execution step, and the execution process information of each execution step is sent.

[0008] In an optional implementation manner, determining the text description information of the matching target business process includes:

[0009] determining the user intention of the user based on the question content;

[0010] A target business process matching the user intention is determined in at least one preset business process, and text description information of the target business process is determined.

[0011] In an optional implementation manner, determining the step description information of the step to be executed based on the text description information and the jump information of the step to be executed by the large model includes:

[0012] In the case where the step to be executed is not the first executed step, determining the step description information and step execution result of the last executed step;

[0013] The question content, the text description information, the step description information of the last executed step and the step execution result are processed by the large model to obtain the step description information of the step to be executed.

[0014] In an optional implementation, when the step to be executed is the first execution step, the step jump information is a null value.

[0015] In an optional implementation manner, the step of determining the step to be executed calling an interface includes:

[0016] Determine the interface description information of the interface called by the target business process;

[0017] The problem content, the interface description information, the steps to be executed and their step description information are processed through the large model to obtain the step calling interface of the steps to be executed.

[0018] In an optional implementation manner, determining the step execution result of the step to be executed by calling the step calling interface includes:

[0019] Determining an interface call link of the step calling the interface according to an interface parameter description of the interface called by the target business process;

[0020] The interface call link is called through the interface call template to determine the step execution result of the step to be executed through the step call interface.

[0021] In an optional implementation manner, determining the interface call link of the step calling the interface according to the interface parameter description of the interface called by the target business process includes:

[0022] Extracting parameter values ​​of target interface parameters of the step calling interface from the problem content according to the interface parameter description by using the large model;

[0023] Based on the parameter value of the target interface parameter of the step calling interface, a uniform resource locator is generated, and the uniform resource locator is determined as the interface calling link.

[0024] In an optional implementation manner, the method further includes:

[0025] In the case that the parameter value of the target interface parameter is not extracted from the question content, sending parameter supplementary information to the front-end interaction device;

[0026] Acquire the parameter value of the target interface parameter sent by the front-end interaction device based on the parameter supplementary information.

[0027] In an optional implementation manner, determining the process execution result of the target business process according to the step execution result of the last execution step includes:

[0028] In the case where it is determined that the step to be executed is the last execution step of the target business process, determining the step execution result of the step to be executed as the process execution result;

[0029] When it is determined that the step to be executed is not the last execution step of the target business process, step jump information of the next step to be executed is determined based on the step description information of the step to be executed and the step execution result.

[0030] In an optional implementation manner, the method further includes:

[0031] Before detecting the question content input by the user, determine the text description information of each business process; wherein the text description information of the business process includes one or more of the following: process flow logic, step identification, step processing content, step flow condition, step end condition;

[0032] Determine the interface description information of all interfaces called by each of the business processes; wherein the interface description information includes one or more of the following: interface implementation function, interface application scenario, interface implementation purpose, and interface call address;

[0033] Determine the interface parameter descriptions of all interfaces called by the business process.

[0034] In an optional implementation manner, the sending of execution process information of each execution step includes:

[0035] Send the execution process information of each execution step to the front-end interactive device; wherein the execution process information includes one or more of the following: each execution step, the step calling interface of each execution step, the step execution result of each execution step, and the analysis process information of the step execution result.

[0036] In an optional implementation, the step of determining text description information matching the target business process in response to the input question content includes:

[0037] Detecting the voice information input by the user through the front-end interactive device to obtain the question content;

[0038] Extracting question keywords from the question content;

[0039] The text description information of the target business process is determined based on the question keywords, and the question keywords are sent to the front-end interactive device so that the front-end interactive device displays the question keywords.

[0040] In a second aspect, the present disclosure provides a process scheduling and orchestration system based on a large model, including:

[0041] A process selection module, used to determine text description information matching the target business process in response to the input question content;

[0042] A thinking module, used for determining the step description information of the step to be executed based on the text description information and the step jump information of the step to be executed at the current moment through a large model;

[0043] An action module is used to determine the step calling interface of the step to be executed, and to determine the step execution result of the step to be executed by calling the step calling interface, and to determine the process execution result of the target business process according to the step execution result of the last execution step, and to send the execution process information of each execution step; wherein the step jump information is used to indicate the association with the step execution result of the previous execution step.

[0044] In an optional implementation, the action module includes: an interface selection submodule, a parameter selection submodule, an interface splicing and calling submodule, and an interface output result conversion submodule;

[0045] The interface selection submodule is used to process the problem content, the interface description information of the interface called by the target business process, and the steps to be executed and their step description information by calling the large model to obtain the step calling interface of the steps to be executed;

[0046] The parameter selection submodule is used to extract the parameter value of the target interface parameter of the step calling interface in the question content according to the interface description information by calling the large model;

[0047] The interface splicing and calling submodule is used to generate a uniform resource locator based on the parameter value of the target interface parameter of the step calling interface; and call the uniform resource locator through the interface calling template to determine the step execution result of the step to be executed through the step calling interface;

[0048] The interface output result conversion submodule is used to replace the target indicator expressed in the first manner in the step execution result with the target indicator expressed in the second manner according to the indicator mapping relationship preset in the parameter library.

[0049] In an optional implementation, the system further includes: a memory module, configured to:

[0050] The step identifier, step description information, and historical step information of the step to be executed are stored.

[0051] In a third aspect, the present disclosure provides a process scheduling and orchestration device based on a large model, including:

[0052] A first determination unit, configured to determine text description information matching a target business process in response to an input question content;

[0053] A second determination unit is used to determine the step to be executed at the current moment and the step description information based on the text description information, the question content and the step jump information at the current moment through the large model, and determine the step calling interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step;

[0054] The third determination unit is used to determine the step execution result of the to-be-executed step by calling the step calling interface, and determine the process execution result of the target business process according to the step execution result of the last execution step, and send the execution process information of each execution step.

[0055] In a fourth aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the above-mentioned first aspect, or any possible implementation of the first aspect are performed.

[0056] In a fifth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned first aspect or any possible implementation of the first aspect are executed.

[0057] In a sixth aspect, an embodiment of the present disclosure further provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the above-mentioned first aspect, or the steps in any possible implementation of the first aspect.

[0058] In an embodiment of the present disclosure, after detecting the input question content, the text description information that matches the target business process is determined; then, the step to be executed at the current moment and the step description information are determined through the large model based on the text description information, the question content and the step jump information at the current moment, and the step calling interface of the step to be executed is determined; wherein the step jump information is associated with the step execution result of the last executed step; finally, the step execution result of the step to be executed is determined through the step calling interface of the step to be executed, and the user's process execution result for the target business process is determined based on the step execution result of the last executed step.

[0059] In the above implementation, by describing the business process in text form and obtaining text description information, the big model can be mapped to the corresponding business process according to the user's intention, thereby solving the problem of the difficulty and low accuracy of the big model in automatically generating business process solutions for different business scenarios. Through the above method, the big model can be guided step by step to realize the flow, scheduling and arrangement of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0060] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0062] Figure 1 A flowchart of a process scheduling and choreography method based on a large model provided in an embodiment of the present disclosure is shown;

[0063] Figure 2 A schematic diagram of a work order assistant page provided by an embodiment of the present disclosure is shown;

[0064] Figure 3 A logical flow chart of a business process provided by an embodiment of the present disclosure is shown;

[0065] Figure 4 A schematic diagram of an input page for question content in a front-end interaction page of a front-end interaction device provided by an embodiment of the present disclosure is shown;

[0066] Figure 5 A schematic diagram of a display page for parameter supplementary information in a front-end interaction page of a front-end interaction device provided by an embodiment of the present disclosure is shown;

[0067] Figure 6 A schematic diagram of a page for adding a new business process in a front-end interactive page provided by an embodiment of the present disclosure is shown;

[0068] Figure 7 A schematic diagram of a display page for parameter supplementary information in a front-end interaction page of another front-end interaction device provided by an embodiment of the present disclosure is shown;

[0069] Figure 8 A flowchart of another large model-based process scheduling and orchestration method provided by an embodiment of the present disclosure is shown;

[0070] Fig. 9 A schematic diagram of a process scheduling and orchestration system based on a large model provided by an embodiment of the present disclosure is shown;

[0071] Fig.10 A schematic diagram of a process scheduling and arranging device for a large model provided by an embodiment of the present disclosure is shown;

[0072] Fig.11 A schematic diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0074] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0075] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0076] Research has found that different scenarios in professional fields currently have a set of complex business processes. In order to realize the functions of different scenarios, the complex business processes for each scenario are solidified by IT; or all processes are generated by intelligent agents. However, for the IT solidification solution, if the original process needs to be changed, the code needs to be modified, resulting in repeated development, low R&D efficiency, and poor generalization. Since the business processes of different detailed scenarios in professional fields are very complex, most of them are a huge inverted tree structure, and there is no large-scale corpus in the proprietary subdivision field to feed the large model for training, it is very difficult to generate all processes through intelligent agents. This solution has problems such as low accuracy of output results and unstable output results.

[0077] Based on the above research, the present disclosure provides a process scheduling and choreography method, device, system and equipment based on a big model. In the embodiment of the present disclosure, after detecting the input question content, determine the text description information that matches the target business process; then, through the big model based on the text description information, the question content and the step jump information at the current moment, determine the step to be executed at the current moment and the step description information, and determine the step call interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step; finally, determine the step execution result of the step to be executed through the step call interface of the step to be executed, and determine the user's process execution result for the target business process according to the step execution result of the last executed step.

[0078] In the above implementation, by describing the business process in text form and obtaining text description information, the big model can be mapped to the corresponding business process according to the user's intention, thereby solving the problem of the difficulty and low accuracy of the big model in automatically generating business process solutions for different business scenarios. Through the above method, the big model can be guided step by step to realize the flow, scheduling and arrangement of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0079] To facilitate understanding of this embodiment, a process scheduling and choreography method based on a large model disclosed in the embodiment of the present disclosure is first introduced in detail. The execution subject of the process scheduling and choreography method based on a large model provided in the embodiment of the present disclosure is generally an electronic device with certain computing capabilities. In some possible implementations, the process scheduling and choreography method based on a large model can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0080] See also Figure 1 As shown, it is a flowchart of a process scheduling and choreography method based on a large model provided by an embodiment of the present disclosure, and the method includes steps S101 to S103, wherein:

[0081] S101: In response to input question content, determine text description information matching the target business process.

[0082] In the disclosed embodiment, the user's input question can be detected and parsed to obtain the question content, wherein the question content can indicate the user's intention. After the question content is detected, the matching target business process and the text description information of the target business process can be determined according to the question content.

[0083] Here, for various business scenarios, especially complex business scenarios, corresponding business processes can be set, and the business processes can be described in text form, thereby obtaining text description information of the business processes. The text description information includes one or more of the following: process flow logic, step identifier (for example, step number), step processing content, step flow condition, and step end condition.

[0084] After detecting the question content input by the user, a business process matching the question content may be determined from multiple business processes as a target business process.

[0085] In an optional implementation, the user intention may be determined based on the question content, and then a matching target business process may be determined from multiple business processes according to the user intention.

[0086] In another optional implementation, semantic matching analysis may be performed between the question content and the brief description of each business process and / or the process keywords, thereby obtaining a semantic matching result between the question content and each business process, for example, the semantic matching result between the question content and business process A is 90%. Afterwards, the business process corresponding to the highest semantic matching result may be determined as the matching target business process.

[0087] In another optional implementation, question keywords can also be extracted from the question content, and then semantic matching analysis can be performed between the question keywords and the brief description of each business process and / or the process keywords to obtain a semantic matching result between the question content and each business process; thereafter, the matching target business process can be determined based on the semantic matching result.

[0088] After the target business process is determined, text description information of the target business process can be obtained in the process library.

[0089] S102: Determine the steps to be executed at the current moment and the step description information based on the text description information, the question content and the step jump information at the current moment through the big model, and determine the step calling interface of the steps to be executed; wherein the step jump information is associated with the step execution result of the last executed step.

[0090] Here, the big model can be understood as a pre-trained language big model, through which the text description information, question content and step jump information at the current moment can be semantically analyzed, so as to obtain the steps to be executed at the current moment and their step description information. Among them, the step description information is used to indicate the step content of the step to be executed, for example, including: step identification, step processing content, step flow condition, step end condition, etc.

[0091] For example, suppose that if the content of a certain execution step is:

[0092] step1: query the status of XX;

[0093] If the value of YY returned is not empty, execute step 2;

[0094] If the result returns that the value of YY is empty, execute step 3.

[0095] At this time, step1 is the step identifier, "query the status of XX" is the processing content of the step, and "if the value of YY returned in the result is not empty, execute step2; if the value of YY returned in the result is empty, execute step3" is the flow condition of the step; there is no end condition for this step.

[0096] Here, the step to be executed at the current moment can be understood as the step to be executed at the current moment. The step jump information is the guidance information for the large model to determine the step to be executed and its step description information. The step jump information can guide the large model to determine the next step to be executed (that is, the step to be executed at the current moment). Among them, if the step to be executed at the current moment is not the first execution step, then the step jump information is associated with the step execution result of the previous execution step. The step jump information can be the step description information of the previous execution step, as well as the step execution result of the previous execution step (that is, the interface return value of the step call interface corresponding to the execution step).

[0097] After determining the step to be executed at the current moment and its step description information, the step calling interface of the step to be executed can be determined according to the step description information.

[0098] S103: Determine the step execution result of the to-be-executed step by calling the step calling interface, determine the process execution result of the target business process according to the step execution result of the last executed step, and send the execution process information of each executed step.

[0099] After determining the step to be executed and its step calling interface, the step can be executed by calling the step calling interface, thereby obtaining the step execution result (i.e., the interface return value). Here, when it is determined that the step to be executed is the last execution step of the target business process, the step execution result of the step to be executed is determined as the process execution result; when it is determined that the step to be executed is not the last execution step of the target business process, the step jump information of the next step to be executed is determined based on the step description information and step execution result of the step to be executed.

[0100] That is, if it is determined that the step to be executed is the last execution step, the step execution result of the step to be executed can be determined as the process execution result of the user for the target business process; otherwise, continue to execute the next execution step until the last execution step is executed, and the step execution result of the last execution step is determined as the process execution result of the user for the target business process.

[0101] In the disclosed embodiment, after obtaining the process execution result of the target business process, the backend service platform can return the execution process information of each execution step to the frontend interactive device; wherein the execution process information includes one or more of the following: each execution step, the step call interface of each execution step, the step execution result of each execution step, and the analysis process information of the step execution result. After obtaining the execution process information, the frontend interactive device can display the execution process information on the frontend interactive page. For example, Figure 2As shown in the figure, the execution process information can be displayed in the section corresponding to the "Work Order Assistant". Figure 2 The “thinking process” in the above is the analysis process information of the step execution results.

[0102] In the disclosed embodiment, if the key fields in the data returned by the back-end service platform to the front-end interactive device are empty or there are problems with the format, the following prompt information may be displayed through the front-end interactive page: We are very sorry, the model needs some information to help it think. In order not to affect your work, it is recommended to transfer to the comprehensive adjustment seat to help handle it. If there is an abnormality in the call of the large model by the back-end service platform, the following prompt information may be displayed through the front-end interactive page: We are very sorry, the model thinking has encountered some problems. In order not to affect your work, it is recommended to transfer to the comprehensive adjustment seat to help handle it.

[0103] In the disclosed embodiment, each execution step is implemented by calling an external interface or tool, rather than calling a function through code. In this way, by configuring the interface or tool, the integration can be completed quickly, and the difficulty of calling the interface is greatly reduced compared to calling the function through code, further ensuring the feasibility of the solution.

[0104] In the above implementation, by describing the business process in text form and obtaining text description information, the big model can be mapped to the corresponding business process according to the user's intention, thereby solving the problem of the difficulty and low accuracy of the big model in automatically generating business process solutions for different business scenarios. Through the above method, the big model can be guided step by step to realize the flow, scheduling and arrangement of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0105] In the related art, IT solidification is performed in advance for the complex business processes of each scenario. If the original process is changed, the code needs to be modified to implement it. If a new scenario is added, the development of a new process also needs to be added, resulting in repeated development, low R&D efficiency, and poor generalization. The technical solution disclosed in the present invention uses a large model to think about each execution step and the way each execution step calls the interface according to the problem content, and there is no need for IT solidification of the business process. For new scenarios, new business processes can be added without modifying the code, which improves generalization.

[0106] In the embodiment of the present disclosure, before performing process orchestration for a user according to the large model-based process scheduling and orchestration method described above, the following contents need to be prepared: text description information of the business process, interface description information of the interface called by the business process, and interface parameter description of the called interface.

[0107] In an optional embodiment, the method further comprises the following steps:

[0108] Step S1: before detecting the question content input by the user, determining the text description information of each business process; wherein the text description information of the business process includes one or more of the following: process flow logic, step identification, step processing content, step flow condition, and step end condition;

[0109] Step S2: Determine the interface description information of all interfaces called by each of the business processes; wherein the interface description information includes one or more of the following: interface implementation function, interface applicable scenario, interface implementation purpose, and interface call address;

[0110] Step S3: Determine the interface parameter descriptions of all interfaces called by the business process.

[0111] In the embodiment of the present disclosure, the business process can be divided from multiple dimensions to obtain multiple business processes, wherein the dimensions include at least one of the following: business scenario, business type, business region, business complexity, etc., and the dimensions are not specifically limited here.

[0112] For each complex business process, corresponding text description information is set, for example Figure 3 The figure shows a process logic diagram of an optional business process. The text description information of the business process can be described as follows:

[0113] step1:Query XX.

[0114] If the value of YY returned is not empty, execute step 2;

[0115] If the result returns that the value of YY is empty, execute step 3.

[0116] step2: Determine MM.

[0117] If AA is returned, it is considered to be BB, and step 4 is executed;

[0118] If CC is returned, execute step 5;

[0119] step3: Determine EE.

[0120] If FF is returned, execute step 18;

[0121] If HH is returned, execute step 17;

[0122] step4:Judge KK.

[0123] If QQ is returned, execute step 6;

[0124] If RR is returned, execute step 7;

[0125] step5: Check PP.

[0126] If LL is returned, execute step 22;

[0127] If SS is returned, execute step 23; ......

[0129] step22: Output: "VV", the process ends;

[0130] step23: Determine NN.

[0131] If JJ is returned, execute step 24;

[0132] If ZZ is returned, execute step 25;

[0133] step24: Output: "OO", the process ends.

[0134] step25:Call GG.

[0135] Output: TT, and return words: "UU", the process ends.

[0136] In the disclosed embodiment, the user may upload a description file of the business process to the backend service platform through the frontend interactive device; the backend service platform may generate the text description information based on the description information.

[0137] In addition, users can also fill in the step description information of each step step by step through the front-end interactive device, and the back-end service platform can generate text description information based on the content filled in by the user. In addition, the text description information can also be the description information pre-set in the process library by the back-end operation personnel.

[0138] After generating the text description information of each business process, it is also necessary to generate interface description information for all interfaces called by each business process. The interface description information can explain the functions of each interface. The interface description information includes one or more of the following: interface implementation function, interface applicable scenarios, interface implementation purpose, and interface call address.

[0139] For example, the above interface description information can be roughly described as follows:

[0140] Interface A: Use interface A to check whether two pieces of information are consistent. Or use interface A to check whether a certain interface is consistent with the interface information in the acceptance document. The interface call address of interface A can be ip1:port1 / url1.

[0141] Interface B: Through this interface B, the current link information of the in-transit work order flow can be queried, which is convenient for installation and maintenance to determine whether the work order flow is normal. Among them, the interface call address of this interface B can be ip2:port2 / url2.

[0142] After generating the interface description information of all interfaces called by each business process, the interface parameter description of all interfaces called by the business process can be determined. The interface parameter description includes but is not limited to the following fields: parameter name, parameter Chinese name, whether it is required, type, value requirement, sample, etc.

[0143] For example, the information content of the interface parameter description of interface A may be the content in Table 1.

[0144] Table 1

[0145]

[0146] In the above implementation, by setting more detailed text description information for each business process and setting more detailed interface description information for all interfaces called by the business process, the large model can select an interface with a better matching function based on the text description information and interface description information of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0147] After the above preparation process is completed, the process scheduling can be arranged based on the above preparation content.

[0148] Here, it should be noted that in the process of scheduling and arranging the process according to the above-prepared content, users can be supported to adjust the text description information, interface description information and interface parameter description of the business process; wherein the adjustment includes at least one of modification, addition and deletion.

[0149] When implementing it, Figure 4 The front-end interactive interface of the front-end interactive device is shown. Here, an adjustment button can be added to the front-end interactive interface; wherein the adjustment button can be as follows Figure 4 The floating movable button shown may also be a button of other forms, and the present disclosure does not specifically limit this. After detecting the user's triggering operation on the adjustment button, you can enter the information adjustment page, wherein the information adjustment page may include: adjustment of text description information, adjustment of interface description information, and adjustment of interface parameter description. After detecting the user's triggering operation on the adjustment of the text description information, you can enter the page for adjusting the text description information, at this time, you can display in the page: a new button for the text description information, a modify button for the text description information, and a delete button for the text description information.

[0150] For example, after detecting the trigger operation of the new button, you can enter the following Figure 6 In the information adding page shown, a text description information editing window or a file uploading window for text description information can be provided to the user in the information adding page (in Figure 6 (not shown). At this time, you can enter "Process Overview" and "Process Details" in the editing window in sequence; then enter the "Function Description" and "Interface Address" of the interface called by the business process, and finally, add the parameter description of the interface, namely "Input Parameter Description" and "Output Parameter Description".

[0151] For another example, after detecting the triggering operation of the modify button, the number information of the business process entered by the user in the search box can be detected, and then the text description information of the business process to be adjusted can be displayed to the user based on the number information. At this time, the user's editing operation on the text description information can be detected, and the text description information of the business process can be updated according to the edited content.

[0152] In an optional implementation, the above step S101 determines the text description information of the matching target business process, specifically including the following steps:

[0153] Step S11: determining the user intention of the user based on the question content;

[0154] Step S12: determining a target business process that matches the user intention in at least one preset business process, and determining text description information of the target business process.

[0155] In the disclosed embodiment, the user intention can be determined based on the question content, and then the user intention is semantically matched with multiple business processes, so as to determine the matching target business process according to the semantic matching analysis result. For example, the matching target business process can be determined based on the user intention through a large model.

[0156] Here, the process feature information of the business process to be matched can be determined, and then the process feature information and the question content entered by the user are added to the prompt and input into the big model. The big model can understand the user's intention based on the question content, and then map the question content to a matching process, that is, the target business process, according to the user's intention. Among them, the process feature information can be understood as information used to describe the function and / or principle of the business process to be matched, and the process feature information can include one or more of the following: a brief description of the business process to be matched, process keywords and other information.

[0157] For example, the content of the question is that the broadband account is "XXX", the optical modem sn is "YYY", the service type is "activation order", and the registration card is 20%.

[0158] In the embodiments of the present disclosure, in addition to understanding the user intention through the big model and obtaining the target business process, the text description information of the matching target business process can also be determined in the following ways:

[0159] Determine the intent label of the business process to be matched; calculate the similarity between the user intent and each intent label; then, determine the business process to be matched corresponding to the intent label with the highest similarity as the target business process.

[0160] and / or

[0161] Determine the process feature information of the business process to be matched; then, calculate the similarity between the user intention and the process feature information; then, determine the business process to be matched corresponding to the process feature information with the highest similarity as the target business process.

[0162] In the disclosed embodiment, before determining the process feature information of the business process to be matched, it is also necessary to determine the business process to be matched in at least one preset business process. The preset business process is the business process prepared by the above-mentioned preparation process. Here, the user's permission information can be determined; wherein the permission information is used to indicate the user's permission to use the business process. Then, according to the permission information, the business process that the user has permission to use in all business processes is determined as the business process to be matched.

[0163] Through the above implementation, the business process that matches the user's intention can be quickly and accurately located, improving the correctness of the overall solution.

[0164] In an optional implementation, the above step S101 determines text description information matching the target business process in response to the input question content, and specifically includes the following steps:

[0165] Step S21: Detect the voice information input by the user through the front-end interactive device to obtain the question content;

[0166] Step S22: extracting question keywords from the question content;

[0167] Step S23: Determine text description information of the target business process based on the question keywords, and send the question keywords to the front-end interactive device so that the front-end interactive device displays the question keywords.

[0168] In the disclosed embodiment, a voice input button can be set in the front-end interaction page of the front-end interaction device. After detecting the user's triggering operation on the voice input button, the front-end interaction device starts to collect the user's voice information and converts the voice information into text to obtain the question content. Figure 5As shown, the interaction process between the user and the backend service platform can be displayed in the front-end interaction page. For example, the user can input voice information through the front-end interaction device by means of a voice input identifier.

[0169] Here, if the recognition of the voice information input by the user fails, for example, the voice information is not recognized, the user can be allowed to re-enter N times; where N can be set according to business needs, such as N=2. In the case of voice recognition failure, AI can display the following message to the user through the front-end interactive page: The voice information you said was not recognized, please say it again. If the recognition still fails for the third time, AI can display the following message to the user: No information was recognized, please enter text.

[0170] After obtaining the question content based on the voice information input by the user, the content keywords of the question content can be extracted; wherein the content keywords can indicate the key information of the question content, so that the matching target business process can be determined according to the key information. For example, the question keyword can be "XXX fault".

[0171] After obtaining the question keywords, the matching target business process can be determined based on the question keywords, and the text description information of the target business process can be determined, including the following methods:

[0172] Method 1:

[0173] The big model can be used to perform semantic matching analysis on the problem keywords and the process feature information of the business process to be matched, so as to determine the matching target business process in the business process to be matched according to the semantic analysis result. The process feature information may include one or more of the following: a brief description of the business process to be matched, process keywords, and other information.

[0174] Here, the process feature information of the to-be-matched business process may be based on the text description information of the business process. For example, a brief process description and / or process keywords of the to-be-matched business process may be extracted from the text description information.

[0175] Method 2:

[0176] The similarity between the question keywords and the process feature information of the to-be-matched business process may be calculated, and then the matching target business process may be determined in the to-be-matched business process according to the similarity, and the text description information of the target business process may be determined.

[0177] After determining the question keywords of the question content, the backend service platform can feedback the question keywords to the frontend interactive device, so that the frontend interactive device displays the question keywords on the frontend interactive page. Figure 5As shown, the displayed question keywords can be expressed as "Identified that your question is: XXXX".

[0178] In the above implementation, by collecting user voice information and converting the voice information into text to obtain the question content, a diversified question input method can be provided to the user, thereby meeting the various input requirements of different users. The method of determining the target business process and its text description information through the question keywords of the question content and the process feature information of the business process to be matched can quickly and accurately locate the business process that matches the user's intention, thereby improving the correctness of the overall solution.

[0179] In an optional implementation, the above step S102 determines the step description information of the step to be executed based on the text description information and the jump information of the step to be executed through the big model, and specifically includes the following steps:

[0180] Step S31: when the step to be executed is not the first execution step, determine the step description information and step execution result of the previous execution step; wherein, when the step to be executed is the first execution step, the step jump information is a null value;

[0181] Step S32: The question content, the text description information, the step description information of the last executed step and the step execution result are processed by the large model to obtain the step description information of the step to be executed.

[0182] In the disclosed embodiment, first, the step jump information of the step to be executed at the current moment needs to be determined, through which the identifier of the step to be executed after the last executed step, that is, the step number of the step to be executed at the current moment, can be determined.

[0183] Here, the step jump information can be determined according to the step sequence of the steps to be executed at the current moment; wherein, if the step to be executed at the current moment is determined to be the first execution step based on the step sequence, the step jump information can be set as preset information, for example, it can be set to a null value. If the step to be executed at the current moment is determined not to be the first execution based on the step sequence, the step jump information can be determined based on the step description information of the last execution step and its step execution result.

[0184] After determining the step jump information, the text description information, the question content and the step jump information at the current moment can be added to the prompt to instruct the large model to think about the next execution step and the step description information of the execution step based on the above input.

[0185] For example, when the step to be executed is not the first execution step, the text description information, question content, step description information of the previous execution step and the step execution result can be input into the large model for processing to obtain the step number of the step to be executed and its step description information.

[0186] When the step to be executed is the first execution step, text description information, question content, and step jump information of empty values ​​can be input into the large model for processing to obtain the step number of the step to be executed and its step description information.

[0187] For example, there is a prompt template: for a specific business process, please answer questions according to the scheduling process, and return the final output when the process ends. Suppose the logic of a business process is as follows:

[0188] step1:Query XX.

[0189] If the value of YY returned is not empty, execute step 2;

[0190] If the result returns that the value of YY is empty, execute step 3.

[0191] step2:......

[0192] The above is the entire business logic. At this point, you need to select the complete fragment of the next step based on the business logic. Assume that the next step is the first execution step. At this point, in addition to the text description information described above, the data input to the big model can also include the following:

[0193] Current execution steps:

[0194] Current step execution result:

[0195] Question content: The broadband account is "XXX", the optical modem sn is "YYY", the service type is "QQQ", and the registration card is 20%.

[0196] Here, "Current Execution Step" and "Current Step Execution Result" are the step jump information described above. Since the next step is the first execution step, the values ​​of "Current Execution Step" and "Current Step Execution Result" are both null values.

[0197] At this point, the large model can return a result, where the returned result is the step number of the step to be executed and its step description information. For example, the following result can be returned:

[0198] step1:Query XX.

[0199] If the value of YY returned is not empty, execute step 2;

[0200] If the result returns that the value of YY is empty, execute step 3.

[0201] In the above implementation, by guiding the big model to solve the scheduling and orchestration flow problems of multi-branch complex processes in professional and detailed fields in a step-by-step manner, the accuracy and stability of the big model output can be improved, thereby improving the accuracy of the overall solution.

[0202] In an optional implementation, the above step S102 determines the step calling interface of the step to be executed, which specifically includes the following steps:

[0203] Step S41: Determine the interface description information of the interface called by the target business process;

[0204] Step S42: The problem content, the interface description information, the steps to be executed and their step description information are processed by the large model to obtain a step calling interface of the steps to be executed.

[0205] In the embodiment of the present disclosure, after the step description information of the step to be executed is determined, the step calling interface of the step to be executed can also be determined based on the step description information.

[0206] In an optional implementation, the question content, step description information of the steps to be executed, and interface description information of the interface called by the target business process can be added to the prompt to indicate the tools or interfaces that need to be called for the steps to be executed in the large model.

[0207] For example, a prompt template: Please select a suitable tool to meet the current needs according to the current needs and the description in the tool. Supported tools include: Interface A: Use interface A to check whether two pieces of information are consistent. Or use interface A to check whether a certain interface is consistent with the interface information in the acceptance document. Among them, the interface call address of interface A can be ip1:port1 / url1. Interface B: Through this interface B, you can query the current link information of the in-transit work order flow, which is convenient for installation and maintenance to determine whether the work order flow is normal. Among them, the interface call address of interface B can be ip2:port2 / url2.

[0208] The current requirements (i.e. the content of the question) are: the broadband account is "XXX", the optical modem sn is "YYY", the service type is "QQQ", and the registration card is 20%. The current execution steps are: step 1: query XX. If the value of YY returned in the result is not empty, execute step 2; if the value of YY returned in the result is empty, execute step 3. After obtaining the above data, the large model can determine that the interface to be called in the step 1 to be executed is interface A.

[0209] In another optional embodiment, after determining the interface description information, the interface label of the interface description information can also be determined, wherein the interface label can be the key information of the interface description information, such as a keyword; the interface label can also be the application scenario of each interface determined based on the interface description information. Next, the problem content, the interface label, the steps to be executed and their step description information can be processed by the large model to obtain the step calling interface of the steps to be executed. In addition, the interface label can also be matched with the problem content and the step description information, so that the matched interface can be determined as the step calling interface of the steps to be executed. For example, the degree of association between the key information of the interface description information and the problem content and the step description information can be calculated respectively, and then the interface with the highest degree of association can be used as the step calling interface of the steps to be executed. For another example, the application scenario of each interface can be matched with the step description information, so as to determine whether the application scenario of the interface matches the scenario corresponding to the step description information according to the matching result; if it matches, the interface is determined as the step calling interface of the steps to be executed.

[0210] In the disclosed embodiment, each execution step is implemented by calling an external interface or tool, rather than calling a function through code. In this way, by configuring the interface or tool, the integration can be completed quickly, and the difficulty of calling the interface is greatly reduced compared to calling the function through code, further ensuring the feasibility of the solution.

[0211] In an optional implementation, the above step S103 determines the step execution result of the step to be executed by calling the step calling interface, which specifically includes the following steps:

[0212] Step S51: determining an interface call link of the interface called by the step according to the interface parameter description of the interface called by the target business process;

[0213] Step S52: calling the interface calling link through the interface calling template to determine the step execution result of the step to be executed through the step calling interface.

[0214] In the embodiment of the present disclosure, the interface parameter description of the interface called by the target business process can be determined; however, the interface call link of the step calling the interface is determined according to the interface parameter description.

[0215] In specific implementation, the big model can be used to process the problem content, the step call interface and the interface parameter description of the determined step to be executed, and then determine the interface call link of the step call interface according to the processing result, wherein the interface call link can be a uniform resource locator URL (Uniform Resource Locator).

[0216] After determining the interface call link, you can call the uniform resource locator URL output by the above steps according to the provided template of the calling interface or tool (i.e., the interface call template). After that, you can receive the return value of the interface (i.e., the step execution result). If an error occurs when calling the interface or tool, the process ends, and the interface call failure and the reason for the failure are returned to the front-end interactive device. For example, the following prompt message can be returned to the front-end interactive device: I am very sorry, there are some problems with the system. In order not to affect your work, it is recommended to transfer to the comprehensive adjustment seat to help deal with it. If the call interface or tool is successful, the return result is output, that is, the step execution result. Next, you can replace the English indicators of the step call interface API output result (i.e., the step execution result) with Chinese indicators through the description of the output parameters, which is convenient for the large model to understand the output results.

[0217] In the embodiment of the present disclosure, the above steps determine the interface call link of the interface called by the step according to the interface parameter description of the interface called by the target business process, and specifically include the following steps:

[0218] Step S511: extracting parameter values ​​of target interface parameters of the interface called in the step from the question content according to the interface parameter description through the large model;

[0219] Step S512: Generate a uniform resource locator based on the parameter value of the target interface parameter of the interface called in the step, and determine the uniform resource locator as the interface call link.

[0220] In specific implementation, the question content, the step call interface of the determined step to be executed and the interface parameter description can be added to the prompt, instructing the large model to extract the parameter value of the target interface parameter of the step call interface from the question content according to the interface parameter description.

[0221] In an embodiment of the present disclosure, when the parameter value of the target interface parameter is not extracted from the question content, parameter supplementary information is sent to the front-end interaction device; and the parameter value of the target interface parameter sent by the front-end interaction device based on the parameter supplementary information is obtained.

[0222] Here, if the large model does not extract the corresponding parameter value from the question content, it can start multiple rounds of dialogue with the user to guide the user to supplement their parameter value. Figure 7 As shown, parameter supplementary information may be sent to the front-end interactive device, for example, "Please provide XXXX data of the device". The front-end interactive device may Figure 7 The parameter supplementary information is displayed in the front-end interactive page shown. The user can supplement the "XXXX data" according to the parameter supplementary information. For example, Figure 7As shown, the user can upload the supplementary "XXXX data" to the back-end service platform through the front-end interactive page; the data format of the supplementary data uploaded by the user can be pictures, voice, text, video, etc. There is no specific limitation on the data format here, which is subject to what can be implemented.

[0223] For supplementary data in the form of images, if the backend service platform fails to identify the supplementary data from the image, it can provide users with N retries, where N can be set according to business needs, such as N=2. In the case of image recognition failure, AI can display the following information to users through the front-end interactive page: The scene was not recognized, please provide a new scene image. If the third image is still not recognized, AI can display the following information to users: The scene was not recognized, please transfer to manual processing.

[0224] For example, a prompt template: Please obtain the target interface parameters required according to the interface parameter description, and analyze it in combination with the actual problem to obtain the input parameters of the calling function (that is, the target interface parameters) and parameter values.

[0225] The output format of this parameter value is: a list consisting of multiple dictionaries. The key of each dictionary in the list should be the input parameter of the function obtained from the interface parameter description, and the value in the dictionary is the actual value understood from the problem content. If there are multiple input parameters, you can configure multiple dictionaries according to the actual number of input parameters and put them into the list to ensure that each dictionary can only have one key.

[0226] The interface parameters are described as:

[0227] Parameter: onu_sn; Field Chinese name: AAAA; Required: Yes; Type: string; Value or description: BB>=6 digits; Data sample: For example: FHTXXX; The question content is: M is "XXX", sn is "YYY", and P is "QQQ".

[0228] The result (i.e. parameter value) returned by the large model is: {'onu_N': 'YYY'}.

[0229] From the above description, it can be seen that the large model can extract the target interface parameters based on the interface parameter description and determine the parameter values ​​of the target interface parameters from the question content. For example, the onu_sn matched to the user is "YYY".

[0230] After determining the parameter value of the target interface parameter, the parameter value can be mapped to its URL, and the target interface parameter and parameter value can be concatenated into the URL. For example, the generated URL is as follows: http:ip1:port1 / XXX?onu_sn=YYY.

[0231] In the above implementation, by describing the business process in text form and obtaining text description information, the big model can be mapped to the corresponding business process according to the user's intention, thereby solving the problem of the difficulty and low accuracy of the big model in automatically generating business process solutions for different business scenarios. Through the above method, the big model can be guided step by step to realize the flow, scheduling and arrangement of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0232] See also Figure 8 As shown, it is a flowchart of another process scheduling and choreography method based on a large model provided by an embodiment of the present disclosure, the method includes steps S801 to S811, wherein:

[0233] Step S801, obtaining the question content input by the user;

[0234] Step S802: determining the user intention based on the question content, and determining a target business process that matches the user intention;

[0235] Step S803: Input the question content and the text description information of the target business process into the large model for processing to obtain the processing result;

[0236] Step S804: determine whether to end the process according to the judgment result; if it is judged to be yes, the process is ended and the final step execution result is returned; if it is judged to be no, the steps to be executed at the current moment and the step description information are obtained, and step S805 is executed;

[0237] Step S805: input the steps to be executed, the step execution results, and the interface description information into the large model for processing to obtain the step calling interface of the steps to be executed;

[0238] Step S806, extracting the parameter value of the target interface parameter from the question content according to the interface parameter description of the step calling interface;

[0239] Step S807: Map the step call interface input by the large model to a URL, and concatenate the parameter value into the URL;

[0240] Step S808, calling the interface through the URL calling step;

[0241] Step S809, determining whether the step calling interface is successfully called; if it is determined that the call is successful, executing step S810; otherwise, executing step S811;

[0242] Step S810: Replace the English index of the output result of the step calling interface with the Chinese index.

[0243] Here, the result of step S810 is returned to step S811 and serves as step jump information for the next step to be executed.

[0244] Step S811, end the process and return the interface call failure and the reason for the failure.

[0245] The specific implementation process of steps S801 to S811 is as described in the above embodiment and will not be described in detail here.

[0246] Reference Fig. 9 As shown, it is a schematic diagram of a process scheduling and choreography system based on a large model provided by an embodiment of the present disclosure, wherein the system includes: a process selection module 91, a thinking module 92, and an action module 93; wherein,

[0247] The process selection module 91 is used to determine text description information matching the target business process in response to the input question content.

[0248] The process selection module can access the process library to obtain the text description information of all business processes connected to the current system, and splice the obtained text description information with the question content carried in the user request into the prompt, and then access the "big model module" to obtain the target business process and text description information mapped to the question content entered by the user.

[0249] Specific implementation method As described in step S101 in the above embodiment, it will not be described in detail here.

[0250] The thinking module 92 is used to determine the step description information of the step to be executed based on the text description information and the step jump information of the step to be executed at the current moment through the large model.

[0251] The thinking module is responsible for calling the big model to determine the steps to be executed at the current moment. The output of the process selection module includes the question content, the corresponding text description information, and the step jump information memorized by the "memory module". At this time, the question content, text description information, and step jump information can be added to the prompt to call the "big model module". Based on the above input, the big model can think about the next step to be executed and the step description information of the step.

[0252] Specific implementation method As described in step S102 in the above embodiment, it will not be described in detail here.

[0253] Action module 93 is used to determine the step calling interface of the step to be executed, and to determine the step execution result of the step to be executed by calling the step calling interface, and to determine the process execution result of the target business process according to the step execution result of the last execution step, and to send the execution process information of each execution step; wherein the step jump information is used to indicate the association with the step execution result of the previous execution step.

[0254] The action module can analyze the steps to be executed at the current moment, select the corresponding tool / interface, determine the parameter values ​​of the target interface parameters, and concatenate the target interface parameters and parameter values ​​to obtain a URL, then call the interface or work through the URL and convert the final result.

[0255] Specific implementation method As described in the above embodiment, step S102 and step S103 are not described in detail here.

[0256] In the above implementation, by describing the business process in text form and obtaining text description information, the big model can be mapped to the corresponding business process according to the user's intention, thereby solving the problem of the difficulty and low accuracy of the big model in automatically generating business process solutions for different business scenarios. Through the above method, the big model can be guided step by step to realize the flow, scheduling and arrangement of the business process, thereby improving the accuracy of each execution step and improving the correctness of the overall solution.

[0257] In an optional embodiment, if Fig. 9 As shown, the action module 93 includes: an interface selection submodule 931 , a parameter selection submodule 932 , an interface splicing and calling submodule 933 , and an interface output result conversion submodule 934 .

[0258] The interface selection submodule 931 is used to process the problem content, the interface description information of the interface called by the target business process, and the steps to be executed and their step description information by calling the large model to obtain the step calling interface of the steps to be executed.

[0259] The interface selection submodule can read all interfaces involved in the target business process and obtain the interface description information of the interface. The obtained interface description information, the question content entered by the user, and the output of the thinking module are added to the prompt, and the "big model module" is called. The big model outputs the name of the step calling interface or tool that needs to be called for the step to be executed.

[0260] The parameter selection submodule 932 is used to extract the parameter value of the target interface parameter of the step calling interface from the problem content according to the interface description information by calling the large model.

[0261] The parameter selection submodule can query the parameter library to obtain the interface parameter description of the step calling interface or tool according to the output of the interface selection submodule "name of the step calling interface or tool". Add the step calling interface, the interface parameter description corresponding to the step calling interface, and the question content entered by the user to the prompt, call the "big model module", and the big model extracts the parameter value of the target interface parameter from the question content entered by the user according to the parameter requirements of the interface. If the corresponding parameter value cannot be extracted from the question content entered by the user, multiple rounds of dialogue with the user can be started to guide the user to supplement their parameter values.

[0262] The interface splicing and calling submodule 933 is used to generate a uniform resource locator based on the parameter value of the target interface parameter of the step calling interface; and call the uniform resource locator through the interface calling template to determine the step execution result of the step to be executed through the step calling interface.

[0263] The interface splicing and calling submodule can map the step calling interface output by the interface selection submodule to its URL according to the interface parameter description of the step calling interface, and splice the target interface parameters and parameter values ​​output by the parameter selection submodule into the URL, and then call the interface or tool according to the URL.

[0264] The interface output result conversion submodule 934 is used to replace the target indicator expressed in the first way in the step execution result with the target indicator expressed in the second way according to the indicator mapping relationship preset in the parameter library.

[0265] The interface output result conversion submodule can query the parameter library to obtain the correspondence between English indicators and Chinese indicators, and then replace the English indicators (target indicators expressed in the first way) of the step execution results output by the interface splicing and calling submodules with Chinese indicators (target indicators expressed in the second way). This processing method can facilitate the large model to understand the meaning of each field and make correct judgments based on the process flow conditions.

[0266] like Fig. 9 As shown, the system further includes a memory module 94 for storing the step identifier, step description information, and historical step information of the step to be executed.

[0267] In the disclosed embodiment, the memory module can save the step number of the current step to be executed, the step description information of the current step to be executed, and the user's multi-round historical step information for use by the multi-round thinking module.

[0268] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0269] Based on the same inventive concept, the embodiment of the present disclosure also provides a process scheduling and orchestration device based on a big model corresponding to the process scheduling and orchestration method based on a big model. Since the principle of solving the problem by the device in the embodiment of the present disclosure is similar to the above-mentioned process scheduling and orchestration method based on a big model in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0270] Reference Fig.10 , which is a schematic diagram of a process scheduling and orchestration device based on a large model provided by an embodiment of the present disclosure, the device includes: a first determining unit 101, a second determining unit 102 and a third determining unit 103;

[0271] A first determining unit 101 is used to determine text description information matching a target business process in response to an input question content;

[0272] The second determining unit 102 is used to determine the step to be executed at the current moment and the step description information based on the text description information, the question content and the step jump information at the current moment through the big model, and determine the step calling interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step;

[0273] The third determination unit 103 is used to determine the step execution result of the to-be-executed step by calling the step calling interface, determine the process execution result of the target business process according to the step execution result of the last execution step, and send the execution process information of each execution step.

[0274] In a possible implementation manner, the first determining unit 101 is configured to:

[0275] determining the user intention of the user based on the question content;

[0276] A target business process matching the user intention is determined in at least one preset business process, and text description information of the target business process is determined.

[0277] In a possible implementation manner, the second determining unit 102 is configured to:

[0278] In the case where the step to be executed is not the first executed step, determining the step description information and step execution result of the last executed step;

[0279] The question content, the text description information, the step description information of the last executed step and the step execution result are processed by the large model to obtain the step description information of the step to be executed.

[0280] In a possible implementation manner, when the step to be executed is the first execution step, the step jump information is a null value.

[0281] In a possible implementation manner, the second determining unit 102 is configured to:

[0282] Determine the interface description information of the interface called by the target business process;

[0283] The problem content, the interface description information, the steps to be executed and their step description information are processed through the large model to obtain the step calling interface of the steps to be executed.

[0284] In a possible implementation manner, the third determining unit 103 is configured to:

[0285] Determining an interface call link of the step calling the interface according to an interface parameter description of the interface called by the target business process;

[0286] The interface call link is called through the interface call template to determine the step execution result of the step to be executed through the step call interface.

[0287] In a possible implementation manner, the third determining unit 103 is configured to:

[0288] Extracting parameter values ​​of target interface parameters of the step calling interface from the problem content according to the interface parameter description by using the large model;

[0289] Based on the parameter value of the target interface parameter of the step calling interface, a uniform resource locator is generated, and the uniform resource locator is determined as the interface calling link.

[0290] In a possible implementation manner, the device is also used for:

[0291] In the case that the parameter value of the target interface parameter is not extracted from the question content, sending parameter supplementary information to the front-end interaction device;

[0292] Acquire the parameter value of the target interface parameter sent by the front-end interaction device based on the parameter supplementary information.

[0293] In a possible implementation manner, the third determining unit 103 is configured to:

[0294] In the case where it is determined that the step to be executed is the last execution step of the target business process, determining the step execution result of the step to be executed as the process execution result;

[0295] When it is determined that the step to be executed is not the last execution step of the target business process, step jump information of the next step to be executed is determined based on the step description information of the step to be executed and the step execution result.

[0296] In a possible implementation manner, the device is also used for:

[0297] Before detecting the question content input by the user, determine the text description information of each business process; wherein the text description information of the business process includes one or more of the following: process flow logic, step identification, step processing content, step flow condition, step end condition;

[0298] Determine the interface description information of all interfaces called by each of the business processes; wherein the interface description information includes one or more of the following: interface implementation function, interface application scenario, interface implementation purpose, and interface call address;

[0299] Determine the interface parameter descriptions of all interfaces called by the business process.

[0300] In a possible implementation manner, the device is also used for:

[0301] Send the execution process information of each execution step to the front-end interactive device; wherein the execution process information includes one or more of the following: each execution step, the step calling interface of each execution step, the step execution result of each execution step, and the analysis process information of the step execution result.

[0302] In a possible implementation manner, the first determining unit 101 is configured to:

[0303] Detecting the voice information input by the user through the front-end interactive device to obtain the question content;

[0304] Extracting question keywords from the question content;

[0305] The text description information of the target business process is determined based on the question keywords, and the question keywords are sent to the front-end interactive device so that the front-end interactive device displays the question keywords.

[0306] For descriptions of the processing flow of each module in the device and the interaction flow between each module, reference may be made to the relevant descriptions in the above method embodiment, which will not be described in detail here.

[0307] Corresponds to Figure 1The process scheduling and arranging method based on a large model in the present disclosure also provides an electronic device 1100, such as Fig.11 FIG. 1 is a schematic diagram of the structure of an electronic device 1100 provided in an embodiment of the present disclosure, including:

[0308] Processor 111, memory 112, and bus 113; memory 112 is used to store execution instructions, including internal memory 1121 and external memory 1122; the internal memory 1121 is also called internal memory, which is used to temporarily store the operation data in the processor 111 and the data exchanged with the external memory 1122 such as a hard disk. The processor 111 exchanges data with the external memory 1122 through the internal memory 1121. When the electronic device 1100 is running, the processor 111 communicates with the memory 112 through the bus 113, so that the processor 111 executes the following instructions:

[0309] In response to the input question content, determining text description information matching the target business process;

[0310] Determine the step to be executed at the current moment and the step description information based on the text description information, the question content and the step jump information at the current moment through the big model, and determine the step calling interface of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step;

[0311] The step execution result of the step to be executed is determined by calling the step calling interface, and the process execution result of the target business process is determined according to the step execution result of the last execution step, and the execution process information of each execution step is sent.

[0312] The present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the large model-based process scheduling and orchestration method described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0313] The present disclosure also provides a computer program product that carries a program code. The program code includes instructions that can be used to execute the steps of the large model-based process scheduling and orchestration method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0314] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is implemented as a computer storage medium. In another optional embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0315] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0316] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0317] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0318] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0319] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A process scheduling and orchestration method based on a large model, characterized in that: include: In response to the input question content, determining text description information matching the target business process from multiple business processes according to the user's intention; The step to be executed at the current moment and the step description information are determined through the large model based on the text description information, the question content and the step jump information at the current moment, and the step calling interface of the step to be executed is determined according to the step description information of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step, and if the step to be executed at the current moment is not the first executed step, then the step jump information is the step description information of the last executed step, and the step execution result of the last executed step; The step execution result of the step to be executed is determined by calling the step calling interface, and the process execution result of the target business process is determined according to the step execution result of the last execution step, and the execution process information of each execution step is sent.

2. The method according to claim 1, characterized in that The determining of the text description information of the matching target business process includes: determining a user intent of the user based on the question content; A target business process matching the user intention is determined in at least one preset business process, and text description information of the target business process is determined.

3. The method according to claim 1, characterized in that The step of determining the step description information of the step to be executed at the current moment based on the text description information, the question content and the step jump information at the current moment by using the large model includes: In the case where the step to be executed is not the first executed step, determining the step description information and step execution result of the last executed step; The question content, the text description information, the step description information of the last executed step and the step execution result are processed by the large model to obtain the step description information of the step to be executed.

4. The method according to claim 3, characterized in that: In the case where the step to be executed is the first execution step, the step jump information is a null value.

5. The method according to claim 1, characterized in that The step of determining the step calling interface of the step to be executed according to the step description information of the step to be executed includes: Determine the interface description information of the interface called by the target business process; The problem content, the interface description information, the steps to be executed and their step description information are processed through the large model to obtain the step calling interface of the steps to be executed.

6. The method according to claim 1, characterized in that Determining the step execution result of the step to be executed by calling the step calling interface includes: Determining an interface call link of the step calling the interface according to an interface parameter description of the interface called by the target business process; The interface call link is called through the interface call template to determine the step execution result of the step to be executed through the step call interface.

7. The method according to claim 6, characterized in that The step of determining the interface calling link of the step calling the interface according to the interface parameter description of the interface called by the target business process includes: Extracting parameter values ​​of target interface parameters of the step calling interface from the problem content according to the interface parameter description by using the large model; Based on the parameter value of the target interface parameter of the step calling interface, a uniform resource locator is generated, and the uniform resource locator is determined as the interface calling link.

8. The method according to claim 7, characterized in that The method further comprises: In the case that the parameter value of the target interface parameter is not extracted from the question content, sending parameter supplementary information to the front-end interaction device; Acquire the parameter value of the target interface parameter sent by the front-end interaction device based on the parameter supplementary information.

9. The method according to claim 1, characterized in that: Determining the process execution result of the target business process according to the step execution result of the last execution step includes: In the case where it is determined that the step to be executed is the last execution step of the target business process, determining the step execution result of the step to be executed as the process execution result; When it is determined that the step to be executed is not the last execution step of the target business process, step jump information of the next step to be executed is determined based on the step description information of the step to be executed and the step execution result.

10. The method according to claim 1, characterized in that The method further comprises: Before detecting the question content input by the user, determine the text description information of each business process; wherein the text description information of the business process includes one or more of the following: process flow logic, step identification, step processing content, step flow condition, step end condition; Determine the interface description information of all interfaces called by each of the business processes; wherein the interface description information includes one or more of the following: interface implementation function, interface application scenario, interface implementation purpose, and interface call address; Determine the interface parameter descriptions of all interfaces called by the business process.

11. The method according to claim 1, characterized in that: The sending of the execution process information of each execution step includes: Send the execution process information of each execution step to the front-end interactive device; wherein the execution process information includes one or more of the following: each execution step, the step calling interface of each execution step, the step execution result of each execution step, and the analysis process information of the step execution result.

12. The method according to claim 1, characterized in that The step of determining text description information matching the target business process in response to the input question content includes: Detecting the voice information input by the user through the front-end interactive device to obtain the question content; Extracting question keywords from the question content; The text description information of the target business process is determined based on the question keywords, and the question keywords are sent to the front-end interactive device so that the front-end interactive device displays the question keywords.

13. A process scheduling and orchestration system based on a large model, characterized in that: include: A process selection module, for responding to the input question content and determining text description information matching the target business process from multiple business processes according to the user's intention; A thinking module, used for determining the step description information of the step to be executed based on the text description information and the step jump information of the step to be executed at the current moment through a large model; An action module is used to determine the step calling interface of the step to be executed according to the step description information of the step to be executed, and to determine the step execution result of the step to be executed by calling the step calling interface, and to determine the process execution result of the target business process according to the step execution result of the last execution step, and to send the execution process information of each execution step; wherein the step jump information is used to indicate the association with the step execution result of the previous execution step. If the step to be executed at the current moment is not the first execution step, then the step jump information is the step description information of the previous execution step, and the step execution result of the previous execution step.

14. The system according to claim 13, characterized in that The action module includes: an interface selection submodule, a parameter selection submodule, an interface splicing and calling submodule, and an interface output result conversion submodule; The interface selection submodule is used to process the problem content, the interface description information of the interface called by the target business process, and the steps to be executed and their step description information by calling the large model to obtain the step calling interface of the steps to be executed; The parameter selection submodule is used to extract the parameter value of the target interface parameter of the step calling interface in the question content according to the interface description information by calling the large model; The interface splicing and calling submodule is used to generate a uniform resource locator based on the parameter value of the target interface parameter of the step calling interface; and call the uniform resource locator through the interface calling template to determine the step execution result of the step to be executed through the step calling interface; The interface output result conversion submodule is used to replace the target indicator expressed in the first manner in the step execution result with the target indicator expressed in the second manner according to the indicator mapping relationship preset in the parameter library.

15. The system according to claim 13, characterized in that The system further comprises a memory module, configured to: The step identifier, step description information, and historical step information of the step to be executed are stored.

16. A process scheduling and arranging device based on a large model, characterized in that: include: A first determination unit, configured to determine text description information matching a target business process from among a plurality of business processes in response to the input question content and according to the user's intention; A second determination unit is used to determine the step to be executed and the step description information at the current moment based on the text description information, the question content and the step jump information at the current moment through a large model, and determine the step calling interface of the step to be executed according to the step description information of the step to be executed; wherein the step jump information is associated with the step execution result of the last executed step, and if the step to be executed at the current moment is not the first executed step, then the step jump information is the step description information of the last executed step and the step execution result of the last executed step; The third determination unit is used to determine the step execution result of the to-be-executed step by calling the step calling interface, and determine the process execution result of the target business process according to the step execution result of the last execution step, and send the execution process information of each execution step.

17. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the large model-based process scheduling and orchestration method as described in any one of claims 1 to 12 are performed.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the large-model-based process scheduling and choreography method as described in any one of claims 1 to 12.

19. A computer program product, characterized in that The computer program product is stored in a storage medium, and the program product is executed by at least one processor to implement the steps of the process scheduling and choreography method based on a large model as described in any one of claims 1 to 12.

Images