Task processing method and device, equipment, storage medium and product

Abstract

The invention discloses a task processing method and device, equipment, a storage medium and a product, and relates to the technical field of artificial intelligence, and the method comprises the steps: responding to a to-be-executed task input by a user, inputting the to-be-executed task and a preset cue word into a large language model, and obtaining a task analysis result output by the large language model; distributing each sub-task corresponding to the to-be-executed task to a corresponding service agent based on the task analysis result; a technical agent corresponding to the subtask is scheduled through the service agent to execute the subtask, and subtask execution results are summarized; and determining a task execution result corresponding to the to-be-executed task based on the sub-task execution result. Compared with the mode that an existing automatic robot can only execute the set task according to the fixed processing flow, the method has the advantages that the to-be-executed task can be fully understood, the to-be-executed task can be split, the to-be-executed task is executed in a graded mode through the corresponding service agent scheduling technology agent, human intervention can be reduced, and the user satisfaction degree can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to task processing methods, devices, equipment, storage media and products. Background Art

[0002] In today's era, the rapid progress of artificial intelligence technology is driving the innovation of enterprise operations. As an important tool to improve production efficiency and reduce costs, Robotic Process Automation (RPA) technology has been adopted by more and more companies. However, traditional RPA platforms usually need to be driven by professional execution software and require users to have an in-depth understanding of the platform, which undoubtedly sets a certain learning threshold and makes the operation process complicated. In addition, traditional RPA systems often rely on fixed rules and codes, which to a certain extent limits their scope of application and flexibility. Summary of the invention

[0003] The main purpose of this application is to provide a task processing method, device, equipment, storage medium and product, aiming to solve the technical problems of high threshold for use and poor flexibility of existing robotic process automation.

[0004] To achieve the above objectives, the present application proposes a task processing method, which includes:

[0005] In response to a task to be performed input by a user, the task to be performed and a preset prompt word are input into a large language model to obtain a task analysis result output by the large language model;

[0006] Allocate each subtask corresponding to the task to be executed to the corresponding business agent based on the task analysis result;

[0007] The business agent schedules the technical agent corresponding to the subtask to execute the subtask, and summarizes the subtask execution results;

[0008] The task execution result corresponding to the task to be executed is determined based on the subtask execution result.

[0009] Optionally, the step of allocating each subtask corresponding to the task to be executed to a corresponding business agent based on the task analysis result includes:

[0010] Determine the subtask corresponding to the task to be performed according to the task analysis result;

[0011] Obtaining business agent information of each business agent;

[0012] Determining the relevance between each business agent and the subtask based on the business agent information;

[0013] Allocate each subtask corresponding to the task to be executed to a corresponding business agent according to the correlation degree.

[0014] Optionally, the step of scheduling the technical agent corresponding to the subtask to execute the subtask through the business agent and summarizing the subtask execution results includes:

[0015] Determine, according to the business agent, the execution process corresponding to executing the subtask;

[0016] Determine the execution tool and execution parameters corresponding to executing the subtask based on the execution process;

[0017] The corresponding technical agent is scheduled to execute the subtask according to the execution tool and the execution parameters, and the subtask execution result fed back by the technical agent is obtained.

[0018] Optionally, the step of inputting the task to be performed and a preset prompt word into a large language model in response to the task to be performed input by the user, and obtaining the task analysis result output by the large language model includes:

[0019] Responding to the task to be performed input by the user, obtaining a preset prompt word;

[0020] Constructing an input text based on the preset prompt word and the task to be performed;

[0021] The input text is input into a large language model to obtain a task analysis result output by the large language model.

[0022] Optionally, the step of determining the task execution result corresponding to the to-be-executed task based on the subtask execution result includes:

[0023] Determine the task execution order according to the task analysis results;

[0024] The task execution result corresponding to the task to be executed is determined based on the task execution order and the subtask execution result.

[0025] Optionally, the large language model includes an intention analysis module, a task decomposition module and a task scheduling module;

[0026] The step of inputting the task to be executed and the preset prompt word into the large language model to obtain the task analysis result output by the large language model includes:

[0027] Performing an intention analysis on the task to be executed by the intention analysis module using the preset prompt word to obtain an intention analysis result;

[0028] Decomposing the to-be-executed task based on the intention analysis result by the task decomposition module to obtain a task decomposition result;

[0029] The task scheduling module schedules the tasks to be executed based on the task decomposition results to obtain task analysis results.

[0030] In addition, to achieve the above-mentioned purpose, the present application also proposes a task processing device, the task processing device comprising:

[0031] A response module, used for responding to a task to be executed input by a user, inputting the task to be executed and a preset prompt word into a large language model, and obtaining a task analysis result output by the large language model;

[0032] A task assignment module, used to assign each subtask corresponding to the task to be executed to a corresponding business agent based on the task analysis result;

[0033] A task execution module, used to schedule the technical agent corresponding to the subtask to execute the subtask through the business agent, and summarize the subtask execution results;

[0034] The task summary response module is used to determine the task execution result corresponding to the task to be executed based on the subtask execution result.

[0035] In addition, to achieve the above objectives, the present application also proposes a task processing device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the task processing method described above.

[0036] In addition, to achieve the above objectives, the present application also proposes a storage medium, which is 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 task processing method described above are implemented.

[0037] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the task processing method described above are implemented.

[0038] In response to the task to be executed input by the user, this application inputs the task to be executed and the preset prompt words into the large language model, and obtains the task analysis result output by the large language model; based on the task analysis result, each subtask corresponding to the task to be executed is assigned to the corresponding business intelligent body; the business intelligent body schedules the technical intelligent body corresponding to the subtask to execute the subtask, and summarizes the subtask execution results; based on the subtask execution results, the task execution result corresponding to the task to be executed is determined. Since this application analyzes the task to be executed through a large language model, and then executes the task to be executed hierarchically through business intelligent bodies and technical intelligent bodies, compared with the existing automated robots that can only execute set tasks according to a fixed processing flow, the above method of this application can fully understand the task to be executed, split the task to be executed, and schedule the technical intelligent body to execute the task to be executed hierarchically through the corresponding business intelligent body, which can reduce human intervention and improve user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0041] Figure 1 A flowchart of the first embodiment of the task processing method of this application is provided;

[0042] Figure 2 A schematic diagram of a three-layer architecture of an intelligent agent provided in Example 1 of the task processing method of this application;

[0043] Figure 3 A flowchart of the second embodiment of the task processing method of this application is provided;

[0044] Figure 4 A flowchart of the third embodiment of the task processing method of this application is provided;

[0045] Figure 5 This is a schematic diagram of the module structure of the task processing device according to an embodiment of the present application;

[0046] Figure 6 Schematic diagram of the device structure of the hardware operating environment involved in the task processing method in the embodiment of the present application.

[0047] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0048] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0049] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0050] The main solution of the embodiment of the present application is: in response to the task to be executed input by the user, the task to be executed and the preset prompt word are input into the large language model to obtain the task analysis result output by the large language model; based on the task analysis result, each subtask corresponding to the task to be executed is assigned to the corresponding business intelligent body; the technical intelligent body corresponding to the subtask is scheduled by the business intelligent body to execute the subtask, and the subtask execution results are summarized; based on the subtask execution results, the task execution result corresponding to the task to be executed is determined. Since the present application analyzes the task to be executed through a large language model, and then executes the task to be executed hierarchically through the business intelligent body and the technical intelligent body, compared with the existing automated robot that can only execute the set tasks according to a fixed processing flow, the above method of the present application can fully understand the task to be executed, split the task to be executed, and schedule the technical intelligent body to execute the task to be executed hierarchically through the corresponding business intelligent body, which can reduce human intervention and improve user satisfaction.

[0051] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, an RPA system, etc. The following takes the RPA system as an example to illustrate this embodiment and the following embodiments.

[0052] Based on this, the present application embodiment provides a task processing method, referring to Figure 1 , Figure 1 A flowchart diagram of the first embodiment of the task processing method of this application is provided.

[0053] In this embodiment, the task processing method includes the following steps:

[0054] Step S10, in response to the task to be performed input by the user, inputting the task to be performed and the preset prompt word into the large language model, and obtaining the task analysis result output by the large language model;

[0055] It should be noted that the task to be executed may be a processing task issued by the user through voice, text, etc. For example, it may be a user input to help me search for official documents related to "foreign affairs" and send them to me by email. The preset prompt word may be a pre-set intelligent prompt word used to prompt the large language model to generate an output result that meets the requirements. The large language model (Large Language Model, LLM) may be a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. The inputting the task to be executed and the preset prompt word into the large language model to obtain the task analysis result output by the large language model may be to input the task to be executed and the preset prompt word into the large language model, so that the large language model analyzes the user intention corresponding to the task to be executed, and disassembles the task and arranges the task according to the understood user intention and the task to be executed, to obtain the task analysis result. If we analyze the intent of the task in the example above, we can get: this task involves searching for official documents on OA and sending emails, which belongs to the office field. We need to call the intelligent agent in the office field to complete the task. By disassembling the task, we can get: to complete the task in the office field, we need to do the following: 1. Log in to the OA webpage; 2. Find the official document area; 3. Search for official documents with the keyword "foreign affairs"; 4. Download the searched official documents; 5. Compress and send them to the corresponding users by email. Arranging the above tasks can be to plan the execution order of the above disassembled tasks to obtain the final task analysis results.

[0056] Furthermore, in order to accurately understand the user's intention, the large language model includes an intention analysis module, a task decomposition module, and a task scheduling module; the step S10 includes: performing intention analysis on the task to be executed by the intention analysis module using the preset prompt words to obtain an intention analysis result;

[0057] Decomposing the to-be-executed task based on the intention analysis result by the task decomposition module to obtain a task decomposition result;

[0058] The task scheduling module schedules the tasks to be executed based on the task decomposition results to obtain task analysis results.

[0059] It should be noted that the intention analysis of the task to be executed may be to use the natural language understanding capability of the large language model to perform the intention analysis on the task to be executed. The task decomposition result obtained by the task decomposition module based on the intention analysis result may be to divide the task to be executed into a plurality of subtasks according to the processing flow of the task to be executed. The task scheduling of the task to be executed may be to schedule the execution order of each subtask.

[0060] Step S20, allocating each subtask corresponding to the task to be executed to a corresponding business agent based on the task analysis result;

[0061] It should be noted that this embodiment realizes hierarchical management of intelligent agents by dividing them into three levels. Each level assumes specific roles and responsibilities, ensuring clear division of labor and efficient collaboration between intelligent agents. The intelligent agents in this embodiment are composed of three levels: control agents, business agents, and technical agents. As the most basic unit, technical agents are responsible for performing specific technical tasks on executors in the production environment. By combining multiple technical agents, business agents for specific vertical industries can be constructed. These business agents are able to understand and handle complex business processes in specific fields. When users input instructions, the control agent receives and parses these instructions, and then arranges and dispatches the corresponding business agents according to the content of the instructions. This process is like a control center commanding a team of professional agents, each member focuses on his or her own task, and works together to complete the tasks specified by the user. The specific relationship between the three can be referred to. Figure 2 , Figure 2 A schematic diagram of a three-layer architecture of an intelligent agent provided in Example 1 of the task processing method of this application; Figure 2 The control agent in the system is used to analyze the intent, parse tasks, and make dynamic decisions on other instructions such as tasks to be executed input by users, and then assign the disassembled subtasks to the corresponding business agents. The business agents then call the technical agents corresponding to the subtasks to perform specific technical tasks. This design not only improves the intelligence level of the system, but also greatly improves the system's processing capabilities and flexibility for complex business processes. Through the layering and combination of agents, the system can adapt to various business scenarios and provide users with efficient and accurate services.

[0062] In the specific implementation, the control agent: as the top-level controller of the system, the control agent is responsible for the scheduling and planning of subsequent agents. The control agent undertakes the task of planning the execution sequence and scheduling process. Business agent: controlled by the control agent, the business agent focuses on knowledge processing in a specific field. Compared with the control agent, its notable feature is that it has a domain-specific knowledge base. Each business agent is equipped with a corresponding knowledge base, which stores the information and parameters required to perform key tasks to assist in task execution. Technical agent: The technical agent is responsible for planning and scheduling appropriate tools based on the task name and parameters provided by the business agent to ensure that the execution end can successfully complete the final execution of the task. In this architecture, agents at each level perform their respective duties, support each other, and jointly promote the completion of tasks.

[0063] Step S30, scheduling the technical agent corresponding to the subtask to execute the subtask through the business agent, and summarizing the subtask execution results;

[0064] It should be noted that the business agent scheduling the technical agent corresponding to the subtask to execute the subtask and summarizing the execution results of the subtask can be to send the subtask to the corresponding business agent. After receiving the corresponding subtask, the business agent analyzes the operations that need to be performed in the subtask, determines the technical agent that needs to be called and the corresponding parameter information according to the operations that need to be performed, and sends the parameter information to the corresponding technical agent (which can also be understood as a tool, such as Excel, email, Word and other tools), so that the technical agent can feedback the execution results after execution, and summarize the execution results fed back by each technical agent.

[0065] Further, in order to improve the execution efficiency of the task, the step S30 may include: determining the execution process corresponding to executing the subtask according to the business agent;

[0066] Determine the execution tool and execution parameters corresponding to executing the subtask based on the execution process;

[0067] The corresponding technical agent is scheduled to execute the subtask according to the execution tool and the execution parameters, and the subtask execution result fed back by the technical agent is obtained.

[0068] It should be noted that the business agent may also be integrated with a large language model of the corresponding field, which is obtained by training through the corresponding field knowledge base equipped by the business agent. The subtasks are parsed and analyzed through the large language model of the specific field in the business agent, and the execution tools and execution parameters that need to be called can be determined. Then, at least one technical agent that needs to be scheduled is determined based on the execution tools and the execution parameters. The technical agent calls at least one execution tool to jointly execute the subtask, and obtains the subtask execution results fed back by each called technical agent.

[0069] Step S40: determining the task execution result corresponding to the task to be executed based on the subtask execution result.

[0070] It should be noted that determining the task execution result corresponding to the task to be executed based on the subtask execution results may be to sort the subtask execution results according to the arrangement order in the task analysis result to obtain the task execution result corresponding to the task to be executed.

[0071] In a specific implementation, the task processing flow in this embodiment can be: 1) Construct prompt words: After receiving the user input, the control agent constructs prompt words for intention analysis, task decomposition and task scheduling according to the predefined prompt word template. These prompt words will guide the large language model to correctly understand and respond to the user's request.

[0072] 2) Big model intention analysis: After receiving the prompt word, the RPA big model (i.e. the big language model) will conduct an in-depth analysis of the user input to identify the user's intention and generate the result of the intention analysis. This step involves natural language processing and machine learning technology to ensure that the user's intention is accurately captured.

[0073] 3) Large model task decomposition: Based on the results of intent analysis, the RPA large model decomposes complex tasks into smaller and more specific subtasks. These subtasks are further converted into executable instructions to prepare for task execution.

[0074] 4) Large model task orchestration: The RPA large model uses the results of task decomposition to orchestrate subtasks and determine the execution order and dependencies. At the same time, the large language model identifies and specifies the skill tools and parameters required to execute each subtask.

[0075] 5) Control Agent Task Parsing: The control agent receives the final orchestration results of the RPA big model and parses them into specific execution instructions. This step ensures that the agent can correctly understand and execute the output of the big model.

[0076] 6) Business agent scheduling: The control agent schedules the corresponding business agent based on the results of the RPA big model, and encapsulates the output results of the big model and passes them to the business agent. This ensures that the correct business agent is activated and ready to perform tasks.

[0077] 7) Multi-skill agent scheduling: The business agent schedules technical agents with multiple skills based on the results of task orchestration. These technical agents may need to work together to complete more complex tasks.

[0078] 8) Action execution: Each technical agent calls the required action tools in turn according to the guidance of the task arrangement, performs specific operations, and completes the final task. This step involves interaction with external systems or devices, as well as feedback and records after the task is completed. Through the above series of detailed steps in this embodiment, the system can ensure that user requests are accurately understood and completed efficiently and accurately. This process design not only improves the intelligence level of the system, but also enhances the adaptability and reliability of the system, and can provide excellent services in various business scenarios.

[0079] The intelligent agent in this embodiment is a program or system that can autonomously perceive the environment, make independent decisions and execute tasks. It includes a memory module, a planning module, a tool and an action module in addition to the large model as the core controller; the memory module makes up for the statelessness of the large model, the planning module stimulates the task decomposition and planning capabilities of the large model through the prompt word technology, and the tool and action module are responsible for environmental interaction or special capabilities. Among them, the control agent, as the core control part of the entire system, understands the natural language input by the user by calling the RPA large model, identifies the user's intention, and then performs task decomposition and arrangement according to the user's intention. Finally, the RPA large model outputs the final task plan. After the control agent receives the task plan generated by the RPA large model, it will first parse the task plan, and then schedule and assign specific tasks to the business agent according to the results of the analysis.

[0080] The business intelligence body in this embodiment is based on the Retrieval Augmented Generation (RAG) architecture, which slices the knowledge of various business fields and stores it in a vector knowledge base. Then, when used, it constructs prompt words by retrieving knowledge in various fields to create a business intelligence body, thereby realizing the planning of business field tasks and parameter extraction, etc. It mainly includes the following core functional modules.

[0081] 1. Business intelligence body construction: Use RAG technology to build a business intelligence body based on RAG. Slice the knowledge of each business field and store it in the vector knowledge base. Construct prompt words by retrieving knowledge in each field to create a business intelligence body.

[0082] 2. Business knowledge base construction: Text feature extraction: Use text parsing and feature extraction technology to extract key features from text data. Word vector encoding: Convert text features into word vectors to facilitate storage and retrieval of the vector knowledge base. Attention encoding: Use the attention mechanism to encode word vectors to improve the retrieval efficiency of the knowledge base.

[0083] 3. Knowledge base indexing: Use loaders to take knowledge and convert it into a single document. Use segmenters to convert documents into small chunks or fragments. Embedders convert fragments into vectors that can be used for semantic search. Save the embedded vectors in a vector database along with their text fragments.

[0084] 4. Retrieval process: A question is asked and sent to the vector database using the same embedding machine. The best matching fragment is determined to answer the question.

[0085] 5. Retrieval-enhanced answer generation: Format knowledge snippets together with customized system prompts and questions, and ultimately get context-specific answers.

[0086] Through the above core functional modules, business agents based on the RAG architecture can efficiently retrieve and utilize knowledge in various business fields, refine and generate task names and task parameters in the business fields required by technical agents (such as the address of financial accounting, the address of OA office, etc.). This enables business agents to understand user intentions more accurately and generate solutions for specific business scenarios.

[0087] The technical agent is configured to be able to use a variety of tools to achieve the final execution of the action end. The configuration of the tool is mainly achieved through prompt words. The role of the prompt word is to inform the large language model of the currently available built-in tools, the functional description of each tool, and the required parameters. By analyzing the instructions, the appropriate built-in tool is selected to complete the task. The current prompt words mainly include the following parts: Role: describes the role of the agent in a specific scenario, for example: you are an office personal assistant who can help users automatically handle daily work tasks; Skills: Skills are external tools that the agent can call, such as: functions or methods (Application Programming Interface, API), robotic process automation, etc. When configuring the prompt word skill, it is necessary to include information such as skill name, functional description, examples, parameters, etc. Take the OA official document download digital employee as an example: Skill ID: 001; Skill name: OA official document download; Function description: This is an OA official document download digital employee, which can search for specified keywords and related official documents within a limited number of days in OA smart search, and send them to the user by email (the user's mailbox has been preset in the tool and does not need to be provided again). When the user needs to check what official documents are on OA, he can use this digital employee. Usage example: I searched for documents related to "foreign affairs" in OA documents in the past week and sent them to me. Parameters: Parameter 1: Parameter variable: keyword; Parameter description: Official document search keyword; Parameter type: str (string); Default value: null. Parameter 2: Parameter variable: limit_day; Parameter description: The number of days from the current OA document issuance; Parameter type: str (string); Default value: null; Restrictions: Set the rules and restrictions that the agent should abide by, for example: only handle tasks related to the current work area, refuse to handle irrelevant issues; be polite and professional in the conversation, and strictly prohibit any form of discrimination or offensive behavior. Based on the above fields, the prompt words of the technical agent can be constructed, and the execution of the execution-end tool can be driven by the skill prompt words. This design enables the agent to respond to various tasks more flexibly and improves the efficiency and accuracy of problem solving.

[0088] In response to the task to be executed input by the user, this embodiment inputs the task to be executed and the preset prompt words into the large language model to obtain the task analysis result output by the large language model; based on the task analysis result, each subtask corresponding to the task to be executed is assigned to the corresponding business agent; the business agent schedules the technical agent corresponding to the subtask to execute the subtask, and summarizes the subtask execution results; based on the subtask execution results, the task execution result corresponding to the task to be executed is determined. Since this embodiment analyzes the task to be executed through a large language model, and then executes the task to be executed hierarchically through business agents and technical agents, compared to the existing automated robots that can only execute set tasks according to a fixed processing flow, the above method of this embodiment can fully understand the task to be executed, split the task to be executed, and schedule the technical agent to execute the task to be executed hierarchically through the corresponding business agent, which can reduce human intervention and improve user satisfaction.

[0089] This embodiment is a multi-agent collaborative RPA system based on a large language model. It aims to use the powerful ability of the large language model in natural language understanding to understand and parse the instructions input by the user, and send the parsed instructions to the agent according to a predetermined format template. The agent uses its own perception, action and memory capabilities to make autonomous decisions and executions to ultimately complete the user's task requirements. This multi-agent collaborative RPA system based on a large language model realizes the seamless connection between the RPA process and natural language processing, greatly improving the usability and adaptability of RPA technology. Users do not need to have a deep understanding of the internal mechanism of the RPA platform. They only need to issue instructions like communicating with a real assistant to complete complex automation processes. This not only reduces the difficulty of operation, but also expands the scope of application of RPA technology, enabling it to respond more flexibly to changing corporate needs and environments. In short, by combining large language models and agent technology with RPA, users are provided with a more intelligent, efficient and easy-to-use automation solution. In order to improve the automation and intelligence of the system, this embodiment also integrates powerful software operation capabilities, which can perform automated operations on PC and mobile phones. Through the RPA process, the system can realize the automated control of various software, including but not limited to office software, customer management system (CRM), enterprise resource management system (ERP), etc. In addition, by using deep learning technologies such as recurrent neural network (RNN), long short-term memory network (LSTM), optical character recognition (OCR), etc., the system also realizes and integrates a variety of commonly used AI capabilities, such as speech recognition, business card recognition, invoice recognition, social security card recognition, marriage certificate recognition, etc., making business processing more convenient and accurate. The system also uses the YOLO algorithm to detect element targets in images, and then inputs the elements into the MobileNet network and OCR for element image classification and text recognition, so as to obtain the semantic information of the elements, so as to have perceptual intelligence and can quickly and accurately recognize and process image information.

[0090] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 3 , Figure 3 The flowchart provided for the second embodiment of the task processing method of the present application, the step S20 further includes the following steps:

[0091] Step S201: determining the subtask corresponding to the task to be executed according to the task analysis result;

[0092] It should be noted that the subtask corresponding to the task to be performed determined according to the task analysis result may be a plurality of subtasks obtained after dividing the task to be performed by a large language model according to the task analysis result.

[0093] Step S202: Acquire business agent information of each business agent;

[0094] It should be noted that the business agent information may include information such as the current load, historical performance, domain, and task type that the business agent is good at.

[0095] Step S203: determining the relevance between each business agent and the subtask based on the business agent information;

[0096] It should be noted that the task to be executed T may include multiple subtasks, which are represented by T = {T 1 ,T 2 ,…,T i}, the determination of the relevance between each business agent and the subtask based on the business agent information may be performed by determining the relevance between each business agent S by the following formula: j With the subtask T i Degree of association:

[0097]

[0098] Among them, Affinity ij For subtask T i With business intelligence S j Capability ij For business intelligence entity S j Process subtask T i Capability, Load j For business intelligence entity S j Current load, Performance j For business intelligence entity S j Historical performance, Match ij For subtask T i Type and Business Intelligence j Good at matching the task type, w 1 ,w 2 ,w 3 ,w 4 It is a pre-set weight coefficient used to balance the influence of various factors.

[0099] Step S204: allocating each subtask corresponding to the task to be executed to a corresponding business agent according to the correlation degree.

[0100] It should be noted that allocating each subtask corresponding to the to-be-executed task to the corresponding business agent according to the correlation may be allocating the subtask to the business agent with the highest correlation therewith.

[0101] Furthermore, in order to monitor the progress of task execution, the technical agent can also regularly feedback the current status and task progress to the control agent.

[0102]

[0103] Among them, Progress j Used to characterize the task progress of technical agent j, Completed j Total is used to represent the number of tasks completed by technical agent j. j The total number of tasks used to characterize technical agent j.

[0104] This embodiment determines the subtasks corresponding to the task to be executed according to the task analysis result; obtains the business agent information of each business agent; determines the relevance between each business agent and the subtask based on the business agent information; and assigns each subtask corresponding to the task to be executed to the corresponding business agent according to the relevance. This embodiment can improve the processing efficiency of the subtasks and thus improve the execution efficiency of the task to be executed by assigning the subtasks to the business agent with the highest relevance to the subtasks.

[0105] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the above embodiments can be referred to the above introduction, and will not be described in detail later. Figure 4 , Figure 4 The flowchart provided for the third embodiment of the task processing method of the present application, the step S10 further includes the following steps:

[0106] Step S101: in response to a task to be executed input by a user, obtaining a preset prompt word;

[0107] It should be noted that the preset prompt word may be an intelligent prompt word that is preset to prompt the large language model to generate an output result that meets the requirements.

[0108] Step S102: constructing an input text based on the preset prompt word and the task to be performed;

[0109] It should be noted that constructing the input text based on the preset prompt word and the task to be performed may be concatenating the preset prompt word and the task to be performed to obtain the input text.

[0110] Step S103: inputting the input text into a large language model to obtain a task analysis result output by the large language model.

[0111] It should be noted that the preset prompt word in this embodiment can be an automatic chain of thought (Chain of Thought, CoT). The CoT prompt word is a prompt word method used for a large language model, which encourages the model to explain its reasoning process.

[0112] In specific implementation, the CoT prompt word construction strategy of this embodiment can refer to the following mode: Task example: organize the financial data for the decade from 2014 to 2024 into a financial report and send it to the project manager via email. CoT prompt word design:

[0113] Step 1: Task understanding and orchestration of control agents.

[0114] 1. First of all, what is the mission objective?

[0115] The task objective is to collate the financial data for the decade from 2014 to 2024 and generate a financial report, which will then be sent to the project manager via email.

[0116] 2. What are the main steps needed to achieve this goal?

[0117] Organize and analyze financial data. Generate financial reports in the required format. Send reports to project managers via email.

[0118] 3. How to break down tasks into smaller subtasks?

[0119] Analyze data preparation. Data processing and organization. Generate financial reports. Review financial reports. Email delivery of reports.

[0120] 4. What resources or expertise are needed to complete these tasks?

[0121] Financial data analysis (completed by financial agents). Auditing (completed by audit agents). Report generation (completed by office agents and word operation agents). Data table operations (completed by excel operation agents). Email operations (completed by email operation agents).

[0122] 5. Which business agents and technical agents should be assigned to each subtask?

[0123] Financial data processing: financial agent + excel operation agent. Audit report: audit agent. Report generation: office agent + word operation agent. Send report: email operation agent.

[0124] 6. What is the order in which tasks are executed?

[0125] (1) The financial agent reads the financial data and analyzes it (obtaining data through the Excel operation agent). (2) The audit agent reviews the data to ensure its accuracy. (3) The office agent collaborates with the word operation agent to generate a financial report. (4) The email operation agent prepares the email and sends the report to the project manager.

[0126] Step 2: Task division and execution of business intelligence entities.

[0127] 1. First of all, what financial data needs to be processed?

[0128] The financial agent needs to process financial data from 2014 to 2024, involving income, expenditure, profit, assets and liabilities, etc.

[0129] 2. What operations need to be performed on the data?

[0130] Filter and summarize financial data for ten years. Calculate key financial indicators (such as profit growth rate, current ratio, etc.). Generate financial analysis tables.

[0131] 3. How to verify and audit data?

[0132] The audit agent is responsible for reviewing the reports generated by the financial agent to ensure that the financial data meets the audit standards.

[0133] 4. What should the final financial report format be?

[0134] The office agent needs to generate a formal financial report including a cover page, table of contents, main text and conclusion. The work operation agent is responsible for generating a report document that meets the format requirements based on the input of the financial agent and the audit agent.

[0135] 5. What tools are needed to generate reports and process data?

[0136] The Excel operation agent is used to import, organize and process data tables. The Word operation agent is used to generate financial reports.

[0137] Step 3: Task execution by technical agents.

[0138] 1. What tools are needed to generate and send reports?

[0139] Use Excel Agent to process financial data. Use Word Agent to generate report documents. Use Mail Agent to prepare and send mail.

[0140] 2. How do technical agents interact with business agents?

[0141] The Excel operation agent collaborates with the financial agent to extract and analyze data. The Word operation agent collaborates with the office agent to generate a formatted report document. The email operation agent collaborates with the office agent to attach the report and send it to the project manager.

[0142] 3. What is the format and standard of the generated files?

[0143] Financial reports should be in a standardized Word document (.docx). The file name of the attachment should comply with the company naming convention. The email content should include a brief description and copy to the relevant departments.

[0144] 4. What is the key information required for email sending?

[0145] The email address of the project manager. The subject of the email should be "Financial Report 2014-2024". The body of the email should briefly describe the content of the report and attach the report.

[0146] In this embodiment, in response to the task to be performed input by the user, a preset prompt word is obtained; an input text is constructed based on the preset prompt word and the task to be performed; the input text is input into the large language model to obtain the task analysis result output by the large language model. In this embodiment, the preset prompt word is used to strengthen the reasoning process of the large language model, so that a more accurate task analysis result can be obtained, thereby improving the user experience.

[0147] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the task processing method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0148] This application also provides a task processing device, please refer to Figure 5 , the task processing device comprises:

[0149] A response module 10 is used to respond to the task to be executed input by the user, input the task to be executed and the preset prompt word into the large language model, and obtain the task analysis result output by the large language model;

[0150] A task assignment module 20, configured to assign each subtask corresponding to the task to be executed to a corresponding business agent based on the task analysis result;

[0151] The task execution module 30 is used to schedule the technical agent corresponding to the subtask to execute the subtask through the business agent, and summarize the subtask execution results;

[0152] The task summary response module 40 is used to determine the task execution result corresponding to the task to be executed based on the subtask execution result.

[0153] In response to the task to be executed input by the user, this embodiment inputs the task to be executed and the preset prompt words into the large language model to obtain the task analysis result output by the large language model; based on the task analysis result, each subtask corresponding to the task to be executed is assigned to the corresponding business agent; the business agent schedules the technical agent corresponding to the subtask to execute the subtask, and summarizes the subtask execution results; based on the subtask execution results, the task execution result corresponding to the task to be executed is determined. Since this embodiment analyzes the task to be executed through a large language model, and then executes the task to be executed hierarchically through business agents and technical agents, compared to the existing automated robots that can only execute set tasks according to a fixed processing flow, the above method of this embodiment can fully understand the task to be executed, split the task to be executed, and schedule the technical agent to execute the task to be executed hierarchically through the corresponding business agent, which can reduce human intervention and improve user satisfaction.

[0154] The task processing device provided by the present application adopts the task processing method in the above embodiment, which can solve the technical problems of high threshold for use and poor flexibility of existing robotic process automation. Compared with the prior art, the beneficial effects of the task processing device provided by the present application are the same as the beneficial effects of the task processing method provided by the above embodiment, and the other technical features in the task processing device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0155] The present application provides a task processing device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the task processing method in the above-mentioned embodiment one.

[0156] Reference below Figure 6 , which shows a schematic diagram of the structure of a task processing device suitable for implementing an embodiment of the present application. The task processing device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The task processing device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0157] like Figure 6 As shown, the task processing device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. Various programs and data required for the operation of the task processing device are also stored in the RAM 1004. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 may allow the task processing device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a task processing device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0158] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0159] The task processing device provided by the present application adopts the task processing method in the above embodiment, which can solve the technical problems of high threshold for use and poor flexibility of existing robotic process automation. Compared with the prior art, the beneficial effects of the task processing device provided by the present application are the same as the beneficial effects of the task processing method provided by the above embodiment, and the other technical features in the task processing device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0160] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0161] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0162] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the task processing method in the above-mentioned embodiment.

[0163] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0164] The computer-readable storage medium may be included in the task processing device; or may exist independently without being installed in the task processing device.

[0165] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0166] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0167] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0168] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned task processing method, and can solve the technical problems of high threshold for use and poor flexibility of existing robotic process automation. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the task processing method provided by the above-mentioned embodiment, and will not be elaborated here.

[0169] The present application also provides a computer program product, including a computer program, which implements the steps of the task processing method as described above when executed by a processor.

[0170] The computer program product provided by this application can solve the technical problems of high threshold for use and poor flexibility of existing robotic process automation. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the task processing method provided by the above embodiment, which will not be repeated here.

[0171] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A task processing method, characterized in that: The task processing method comprises the following steps: In response to a task to be performed input by a user, the task to be performed and a preset prompt word are input into a large language model to obtain a task analysis result output by the large language model; Allocate each subtask corresponding to the task to be executed to the corresponding business agent based on the task analysis result; The business agent schedules the technical agent corresponding to the subtask to execute the subtask, and summarizes the subtask execution results; The task execution result corresponding to the task to be executed is determined based on the subtask execution result.

2. The task processing method according to claim 1, characterized in that: The step of allocating each subtask corresponding to the task to be executed to the corresponding business agent based on the task analysis result includes: Determine the subtask corresponding to the task to be performed according to the task analysis result; Obtaining business agent information of each business agent; Determining the relevance between each business agent and the subtask based on the business agent information; Allocate each subtask corresponding to the task to be executed to a corresponding business agent according to the correlation degree.

3. The task processing method according to claim 1, characterized in that: The step of scheduling the technical agent corresponding to the subtask to execute the subtask through the business agent and summarizing the subtask execution results includes: Determine, according to the business agent, the execution process corresponding to executing the subtask; Determine the execution tool and execution parameters corresponding to executing the subtask based on the execution process; The corresponding technical agent is scheduled to execute the subtask according to the execution tool and the execution parameters, and the subtask execution result fed back by the technical agent is obtained.

4. The task processing method according to claim 1, characterized in that: The step of inputting the task to be performed and the preset prompt word into the large language model in response to the task to be performed input by the user, and obtaining the task analysis result output by the large language model, comprises: Responding to the task to be performed input by the user, obtaining a preset prompt word; Constructing an input text based on the preset prompt word and the task to be performed; The input text is input into a large language model to obtain a task analysis result output by the large language model.

5. The task processing method according to any one of claims 1 to 4, characterized in that: The step of determining the task execution result corresponding to the task to be executed based on the subtask execution result includes: Determine the task execution order according to the task analysis results; The task execution result corresponding to the task to be executed is determined based on the task execution order and the subtask execution result.

6. The task processing method according to any one of claims 1 to 4, characterized in that: The large language model includes an intention analysis module, a task decomposition module and a task scheduling module; The step of inputting the task to be executed and the preset prompt word into the large language model to obtain the task analysis result output by the large language model includes: Performing an intention analysis on the task to be executed by the intention analysis module using the preset prompt word to obtain an intention analysis result; Decomposing the to-be-executed task based on the intention analysis result by the task decomposition module to obtain a task decomposition result; The task scheduling module schedules the tasks to be executed based on the task decomposition results to obtain task analysis results.

7. A task processing device, characterized in that: The task processing device comprises: A response module, used for responding to a task to be executed input by a user, inputting the task to be executed and a preset prompt word into a large language model, and obtaining a task analysis result output by the large language model; A task allocation module, used to allocate each subtask corresponding to the task to be executed to a corresponding business agent based on the task analysis result; A task execution module, used to schedule the technical agent corresponding to the subtask to execute the subtask through the business agent, and summarize the subtask execution results; The task summary response module is used to determine the task execution result corresponding to the task to be executed based on the subtask execution result.

8. A task processing device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the task processing method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the task processing method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the task processing method according to any one of claims 1 to 6 are implemented.

Images