Task processing method and device, program product and electronic equipment

Abstract

The invention provides a task processing method and device, a program product and electronic equipment, and relates to the technical field of computers. The method comprises the steps of receiving a to-be-processed task; determining an associated workflow corresponding to the to-be-processed task according to the service information of the to-be-processed task; wherein the associated workflow is obtained from an enhanced memory system in which workflows of various tasks are stored; and determining a target workflow of the to-be-processed task according to the associated workflow, and executing the to-be-processed task according to the target workflow. According to the method, the target workflow is quickly created by determining the associated workflow corresponding to the to-be-processed task, and compared with a processing mode that an intelligent agent establishes the workflow from the beginning each time in the related technology, the efficiency of creating the target workflow is greatly improved, and the processing efficiency of the to-be-processed task is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a task processing method, device, program product and electronic device. Background Art

[0002] An agent is a software entity based on a language model that performs actions according to instructions provided in natural language, such as web navigation or interaction with a website. Specifically, agents based on large language models rely on large language models to understand natural language instructions and generate corresponding actions based on the input.

[0003] Although intelligent agents based on large language models have demonstrated certain capabilities in multi-step reasoning and task execution, they need to generate solutions from scratch each time a task is performed, making it difficult to achieve cross-task generalization and efficiency improvements. Summary of the invention

[0004] The present disclosure provides a task processing method, a task processing device, a computer program product and an electronic device, so as to quickly and accurately create a target workflow corresponding to a task and improve the processing efficiency of the task.

[0005] According to a first aspect of the present disclosure, a task processing method is provided, the method comprising:

[0006] Receive pending tasks;

[0007] Determine, according to the business information of the task to be processed, an associated workflow corresponding to the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks;

[0008] According to the associated workflow, a target workflow of the task to be processed is determined, and according to the target workflow, the task to be processed is executed.

[0009] In a possible implementation, the method further includes:

[0010] Determine whether the action trajectory corresponding to the task to be processed is reusable;

[0011] When it is determined that the motion track corresponding to the task to be processed is reusable, the motion track corresponding to the task to be processed is converted into a new workflow, and the new workflow is stored in the enhanced memory system.

[0012] In a possible implementation, before determining whether the motion trajectory corresponding to the task to be processed is reusable, the method further includes:

[0013] Determining observation information corresponding to the interactive environment when executing the task to be processed; and determining specific action information triggered based on the observation information;

[0014] An action trajectory corresponding to the task to be processed is generated according to the observation information and the corresponding specific action information.

[0015] In a possible implementation, converting the action trajectory corresponding to the task to be processed into a new workflow includes:

[0016] The action track corresponding to the task to be processed is abstracted into a general workflow step, and placeholders are used to replace values ​​to obtain a new workflow.

[0017] In a possible implementation, determining whether the action trajectory corresponding to the task to be processed is reusable includes:

[0018] Determine whether the motion trajectory corresponding to the execution of the task to be processed meets a preset stop condition; the preset stop condition is used to filter the motion trajectory corresponding to the successful execution result of the task to be processed;

[0019] When it is determined that the motion trajectory corresponding to the task to be processed satisfies a preset stop condition, it is determined that the motion trajectory corresponding to the task to be processed is reusable.

[0020] In a possible implementation, determining the target workflow of the task to be processed according to the associated workflow includes:

[0021] A target workflow for the task to be processed is created according to the associated workflow, the preset semantic understanding capability and the contextual reasoning capability.

[0022] In a possible implementation, before receiving the task to be processed, the method further includes:

[0023] Obtaining workflows of various tasks in the target project, and / or obtaining pre-collected annotated data sets; wherein the tasks to be processed belong to subtasks triggered in the target project;

[0024] The workflows of the various tasks and / or the workflows generated based on the data sets are stored in the enhanced memory system.

[0025] According to a second aspect of the present disclosure, there is provided a task processing device, the device comprising:

[0026] A receiving unit, used for receiving tasks to be processed;

[0027] A determination unit, configured to determine, according to the business information of the task to be processed, an associated workflow corresponding to the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks;

[0028] The processing unit is used to determine the target workflow of the task to be processed according to the associated workflow, and execute the task to be processed according to the target workflow.

[0029] In a possible implementation manner, the device further includes a creating unit, configured to:

[0030] Determine whether the action trajectory corresponding to the task to be processed is reusable;

[0031] When it is determined that the motion track corresponding to the task to be processed is reusable, the motion track corresponding to the task to be processed is converted into a new workflow, and the new workflow is stored in the enhanced memory system.

[0032] In a possible implementation manner, the device further includes a generating unit, configured to:

[0033] Determining observation information corresponding to the interactive environment when executing the task to be processed; and determining specific action information triggered based on the observation information;

[0034] An action trajectory corresponding to the task to be processed is generated according to the observation information and the corresponding specific action information.

[0035] In a possible implementation manner, the device further includes a creating unit, configured to:

[0036] The action track corresponding to the task to be processed is abstracted into a general workflow step, and placeholders are used to replace values ​​to obtain a new workflow.

[0037] In a possible implementation manner, the device further includes a creating unit, configured to:

[0038] Determine whether the motion trajectory corresponding to the execution of the task to be processed meets a preset stop condition; the preset stop condition is used to filter the motion trajectory corresponding to the successful execution result of the task to be processed;

[0039] When it is determined that the motion trajectory corresponding to the task to be processed satisfies a preset stop condition, it is determined that the motion trajectory corresponding to the task to be processed is reusable.

[0040] In a possible implementation manner, the processing unit is configured to:

[0041] A target workflow for the task to be processed is created according to the associated workflow, the preset semantic understanding capability and the contextual reasoning capability.

[0042] In a possible implementation manner, the device further includes a creating unit, configured to:

[0043] Obtaining workflows of various tasks in the target project, and / or obtaining pre-collected annotated data sets; wherein the tasks to be processed belong to subtasks triggered in the target project;

[0044] The workflows of the various tasks and / or the workflows generated based on the data sets are stored in the enhanced memory system.

[0045] According to a third aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method of the first aspect and possible implementations thereof are implemented.

[0046] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method of the above-mentioned first aspect and its possible implementation methods by executing the executable instructions.

[0047] The technical solution disclosed in this disclosure has the following beneficial effects:

[0048] In the disclosed embodiment, a task to be processed can be received, and then, based on the business information of the task to be processed, an associated workflow corresponding to the task to be processed can be determined; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various types of tasks; and, based on the associated workflow, a target workflow of the task to be processed can be determined, and the task to be processed can be executed based on the target workflow. It can be seen that in the disclosed embodiment, the intelligent agent can directly obtain the associated workflow from the enhanced memory system constructed by extracting and reusing based on past experience, and then determine the target workflow of the task to be processed based on the associated workflow, so as to quickly process the task to be processed and improve the efficiency of task execution.

[0049] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by practicing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments of the present disclosure will be briefly introduced below. Obviously, the drawings introduced below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A schematic diagram of an application scenario in this exemplary embodiment is shown;

[0052] Figure 2 A schematic diagram showing the use of a workflow in this exemplary embodiment;

[0053] Figure 3 A schematic diagram showing a workflow for obtaining the workflow in this exemplary embodiment is shown;

[0054] Figure 4 A schematic diagram of obtaining an offline workflow in this exemplary embodiment is shown;

[0055] Figure 5 A schematic diagram of obtaining an online workflow in this exemplary embodiment is shown;

[0056] Figure 6 A flowchart showing a task processing method in this exemplary embodiment is shown;

[0057] Figure 7 A schematic diagram showing the structure of a task processing device in this exemplary embodiment is shown;

[0058] Figure 8 A schematic structural diagram of an electronic device in this exemplary embodiment is shown. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the technical solutions 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. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure. In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be arbitrarily combined with each other. In addition, although the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in an order different from that here.

[0060] The term "comprising" and any variations thereof in the specification and claims of the present disclosure are intended to cover non-exclusive protection. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0061] In the embodiments of the present disclosure, one or more, "many" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0062] It should be noted that the terms "first", "second", "third", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order, sequence, size, and priority. For example, the first task to be processed and the second task to be processed in the embodiment of the present disclosure are only used to distinguish different tasks to be processed. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0063] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, which are schematic diagrams of the present disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the accompanying drawings may be functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in hardware modules or integrated circuits, or in networks, processors or microcontrollers. The implementation method can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein. The features, structures or characteristics described in the present disclosure may be combined in one or more implementations in any suitable manner. In the following description, many specific details are provided to provide a full description of the implementation methods of the present disclosure. However, those skilled in the art should be aware that one or more specific details may be omitted when implementing the technical solution of the present disclosure, or one or more specific details may be replaced by other methods, components, devices, steps, etc.

[0064] It should be noted that in the embodiments of the present disclosure, some existing solutions in the industry such as software, components, models, etc. may be mentioned, which should be considered as exemplary, and their purpose is only to illustrate the feasibility of implementing the technical solution of the present disclosure, but it does not mean that the applicant has or will necessarily use the solution. In the technical solution of the present disclosure, the collection, dissemination, and use of data are in compliance with relevant national laws and regulations.

[0065] In the embodiments of the present disclosure, in order to better understand the solutions provided by the present disclosure, some key names involved are introduced below, as follows:

[0066] (1) Agent Workflow Memory (AWM): is a framework designed to enhance the performance and adaptability of language model-based agents by enabling them to extract, store, and reuse workflow-structured action sequences derived from past task experience. It operates in both offline and online modes: in offline mode, workflows are pre-generated from annotated datasets, while in online mode, workflows are dynamically induced by real-time task execution. These workflows are integrated into the agent's memory, allowing it to generalize across different tasks and domains, improve task efficiency, and reduce redundant learning. Over time, the agent's memory continues to grow as new workflows are continuously evolved, ensuring that it can perform complex multi-step tasks more effectively in future scenarios.

[0067] (2) Agent: A software entity based on a language model that performs operations according to instructions provided in natural language, such as web navigation or interaction with a website.

[0068] (3) Workflow: A reusable sequence of operations or routines derived from task experience, representing the structured steps required to complete a task.

[0069] (4) Agents Coordination (AC): refers to the process of synchronizing multiple agents or components to work together efficiently to achieve complex goals by sharing workflows, dividing tasks, or cross-domain collaboration.

[0070] (5) Memory: A repository that stores workflows and is used to guide future task solving processes. This memory can be enhanced offline (from pre-training examples) or online (from new tasks performed in real time).

[0071] Language model-based agents rely on large language models (such as the GPT series) to understand natural language instructions and generate corresponding actions based on the input. Language model-based agents have demonstrated certain capabilities in multi-step reasoning and task execution, but they still face challenges when facing complex and long-term tasks, especially the inability to effectively reuse knowledge between different tasks.

[0072] Specifically, although the intelligent agent based on the large language model can strengthen learning and summarize task trajectories, traditional reinforcement learning mainly focuses on how to adjust the strategy from reward feedback, and its induction and reuse of task trajectories are relatively limited. In addition, the intelligent agent requires generating solutions from scratch every time a task is performed, making it difficult to achieve cross-task generalization and efficiency improvement.

[0073] In view of one or more of the above problems, an exemplary embodiment of the present disclosure provides a task processing method, through which a task to be processed can be received, and then an associated workflow corresponding to the task to be processed can be determined based on the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system that stores workflows of various types of tasks; and the target workflow of the task to be processed can be determined based on the associated workflow, and the task to be processed can be executed based on the target workflow. It can be seen that in the embodiment of the present disclosure, the intelligent agent can directly obtain the associated workflow from the enhanced memory system constructed by extracting and reusing based on past experience, and then determine the target workflow of the task to be processed based on the associated workflow, so as to quickly process the task to be processed and improve the efficiency of task execution.

[0074] In order to better understand the technical solutions provided by the embodiments of the present disclosure, the following briefly introduces the application scenarios to which the technical solutions provided by the embodiments of the present disclosure are applicable. It should be noted that the application scenarios introduced below are only used to illustrate the embodiments of the present disclosure and are not intended to limit the present disclosure. In specific implementation, the technical solutions provided by the embodiments of the present disclosure can be flexibly applied according to actual needs.

[0075] In the embodiments of the present disclosure, the task processing technology can be applied to community marketing, video recommendation, business processing, and other scenarios based on intelligent agent execution of web page navigation or website interaction, which is not limited in the embodiments of the present disclosure.

[0076] See also Figure 1 As shown, Figure 1 This is an application scenario to which the technical solution of the embodiment of the present disclosure can be applied. In the scenario diagram, multiple terminal devices 101 and electronic devices 102 are included. Among them, the terminal device 101-1, the terminal device 101-2, ..., the terminal device 101-n can be used by different users.

[0077] The terminal device 101 and the electronic device 102 , as well as each terminal device 101 , may be directly or indirectly connected to each other through one or more networks 103 .

[0078] In an embodiment of the present disclosure, a user can trigger a task processing request based on the terminal device 101, and then the terminal device 101 sends the task processing request to the electronic device 102, so that the electronic device 102 receives the task to be processed, and then determines the associated workflow corresponding to the task to be processed based on the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system that stores the workflows of various tasks; and, based on the associated workflow, the target workflow of the task to be processed can be determined, and the task to be processed can be executed based on the target workflow.

[0079] In the disclosed embodiment, Figure 1 The terminal devices 101 may be mobile phones, tablet computers (PADs), personal computers (PCs), smart TVs, smart watches, smart speakers, smart car devices, wearable devices, etc., but are not limited thereto. These devices may have the function of supporting web navigation and website interaction.

[0080] as well as, Figure 1 The electronic device 102 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It may also be a cloud server or cloud server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), as well as big data and artificial intelligence platforms, but is not limited to these.

[0081] Of course, the method provided in the embodiment of the present disclosure is not limited to Figure 1 The application scenarios shown may also be used in other possible application scenarios, such as application scenarios where the task processing method is implemented only by electronic devices, and the embodiments of the present disclosure are not limited thereto.

[0082] To further illustrate the technical solution provided by the embodiments of the present disclosure, this is described in detail below in conjunction with the accompanying drawings and specific implementation methods. Although the embodiments of the present disclosure provide method operation steps as shown in the following embodiments or drawings, more or fewer operation steps may be included in the method based on routine or no creative labor. In the steps where there is no necessary causal relationship logically, the execution order of these steps is not limited to the execution order provided by the embodiments of the present disclosure. The method may be executed in the order of the method shown in the embodiments or drawings or in parallel during the actual processing process or when the device is executed.

[0083] In the embodiment of the present disclosure, before introducing the method for processing tasks, the agent workflow memory AWM framework provided by the embodiment of the present disclosure is first introduced. Specifically, an enhanced memory system can be constructed based on the agent workflow memory AWM framework, wherein the enhanced memory system can be a collection of workflows integrated in the memory of the agent.

[0084] See also Figure 2 As shown, Figure 2 The key components in AWM and their interrelationships are presented. Figure 2 It describes how intelligent agents work with the environment, augmented memory systems, and large model bases to achieve accumulation and reuse of task execution and workflows.

[0085] In the disclosed embodiment, when the agent interacts with the environment, it completes the task by performing a series of actions and generates observation results, which are used to guide the next action. In addition, the execution of the action can be based on the semantic understanding ability and contextual reasoning ability provided by the large model base. In other words, the large model base can guide the agent to make reasonable action decisions.

[0086] In the disclosed embodiment, when a task is completed, the agent can extract the workflow of the task and add it to the enhanced memory system. That is, the workflow in the enhanced memory system in the disclosed embodiment can be derived from a set of tasks that have been completed in the past and stored in the memory in a reusable form, so as to provide a reference for similar tasks in the future.

[0087] In the disclosed embodiment, when executing a new task, the agent will query the associated workflow in the enhanced memory system, and apply the associated workflow to the context of the current task to determine the target workflow. It can be seen that the enhanced memory system in the disclosed embodiment can provide the agent with existing workflows, help optimize task execution, and reduce the agent's repeated learning and redundant operations. In addition, through the aforementioned workflows added to the enhanced memory system and the enhanced memory system providing the agent with a loop mechanism of existing workflows, the enhanced memory system can continuously accumulate new workflows, realize the agent's continuous learning and optimization, and enable the agent to gradually improve the task success rate and execution efficiency in a complex environment.

[0088] See also Figure 3 , Figure 3Schematic diagram of the workflow induction process in the embodiment of the present disclosure. In the embodiment of the present disclosure, during the task execution and observation collection stage, the agent perceives the current state of the environment by interacting with the environment. Specifically, at each time step, the agent can obtain new information from the environment, such as web page content or interface elements. The information obtained by the agent from the environment is used as an observation result. Then, based on these observation results, the agent performs specific actions, such as clicking a button or scrolling a page, so as to advance the progress of the task. Furthermore, after the agent performs a specific action, the state of the environment will change due to the interaction between the agent and the environment, so new information can be obtained as a new observation result for use in the next action.

[0089] During the interaction between the agent and the environment, the agent records the generated "action-observation" pairs as trajectories. Each "action-observation" pair contains three parts: the state description of the current environment, the specific action performed by the agent, and the reasoning process of the agent taking this action. For example, the state description of the current environment is: the agent sees the order list on the order page, the specific action performed by the agent is to click on the order link to view the details, and the reasoning process of the agent taking this action is that if the Dingdang page displays an order list, it may want to view the order details.

[0090] In the disclosed embodiment, when a task is close to completion, the agent will determine whether the stop condition has been reached, such as the task has been successfully completed or the STOP action has been executed. For example, the goal of a task is to find a specific order. When the agent successfully finds the order and sends the result to the user, the task can be considered to have been successfully completed. It can be seen that the stop condition can be understood as the agent achieving the task goal or reaching a predetermined termination condition.

[0091] In the disclosed embodiment, after completing a task, the agent will evaluate the entire trajectory to determine whether the task was successfully solved. If the task is successfully completed, the trajectory will be marked as being usable for generating a workflow; otherwise, the agent may adjust its strategy, modify its action sequence and try again. In other words, the agent in the disclosed embodiment can summarize the workflow by determining whether the trajectory is successful, thereby providing a valuable reference for future related tasks.

[0092] In the disclosed embodiment, when the agent determines that the trajectory is marked as being usable for generating a workflow, the agent can abstract the action sequence and observation information of the task into a common workflow step, and use placeholders to replace specific values, such as "{order-id}". The workflow obtained in this way has stronger generalization ability in subsequent tasks and can adapt to the changing requirements of different tasks.

[0093] In the disclosed embodiment, the summarized workflow can be stored in the enhanced memory system for future use in executing other tasks. It can be seen that the workflow stored in the enhanced memory system not only saves the execution steps of the task, but also retains the reasoning logic of the execution, which can be quickly called in future task execution to improve the execution efficiency of the task. Optionally, the aforementioned workflow can also be used as part of the agent knowledge base to provide support for other functions of the agent or the model associated with the agent.

[0094] In the disclosed embodiment, when executing future tasks, the agent will actively apply the workflow trajectory stored in the augmented memory system. In this way, when encountering similar tasks, the agent will query the associated workflow in the augmented memory system and integrate it into the context of the current task. This approach ensures that the agent can complete tasks quickly, reduce redundant operations, and greatly improve the efficiency and success rate of task execution.

[0095] In the disclosed embodiment, the AWM framework supports offline workflow induction and online workflow induction. That is, the induction of workflow can be divided into offline workflow induction and online workflow induction. Among them, offline workflow induction and online workflow induction can determine when and how to induce and utilize workflows, such as inducing and utilizing workflows before task execution (offline) or during task execution (online). That is, the aforementioned offline workflow induction and online workflow induction differ in terms of the time and availability of workflow generation data.

[0096] For example, see Figure 4 , Figure 4 The diagram is a summary of the offline workflow in the embodiment of the present disclosure.

[0097] In the disclosed embodiment, the workflow obtained by offline workflow induction is induced from the annotated examples collected in advance during the training phase. Optionally, a pre-collected annotated data set can be obtained, and a workflow can be generated based on the data set, and the workflow can be stored in the enhanced memory system. That is, the agent can process multiple training data sets, identify common task routines, and store them as offline workflows in its enhanced memory system based on the training data sets and task routines.

[0098] In the disclosed embodiment, the workflow obtained by offline workflow induction can also be induced from the tasks that have been executed in the project. Optionally, the workflow of each type of task in the target project can be obtained; wherein the task to be processed belongs to the subtask triggered in the target project; and the workflow of each type of task is stored in the enhanced memory system. In other words, the agent can obtain the workflow of each completed task in the target project to which the task to be processed belongs, and store the workflow as an offline workflow in its enhanced memory system.

[0099] Obviously, once the workflow is summarized, the workflow can be stored in the enhanced memory system (M+W) of the agent, where M represents the original memory and W represents the newly generated workflow. That is to say, the workflow stored in the enhanced memory system in the embodiment of the present disclosure can be in accordance with the initial workflow, the first version of the initial workflow derived workflow, the second version of the initial workflow derived workflow, ..., the Nth version of the initial workflow derived workflow, N is a positive integer. The pre-stored workflow is used as the knowledge base of the agent, so that it can efficiently handle related tasks in the testing or deployment stage without having to learn new patterns from scratch.

[0100] It can be seen that the pre-processing method of pre-collecting annotated data sets allows the agent to obtain the corresponding workflow from the enhanced memory system when performing new tasks. For example, the agent learns workflows such as "searching for products" or "navigating to hotel lists" by analyzing past annotated web navigation data. In this way, when high-quality, domain-specific workflows are available, the agent can directly determine the target workflow for new tasks based on the workflow. In other words, the agent obtains a certain basic workflow from the beginning, so as to effectively perform new tasks, that is, the workflow corresponding to the new task is determined by using the pre-learned routines, without further adaptation or creating the workflow from scratch.

[0101] For example, see Figure 5 , Figure 5 A schematic diagram summarizing the online workflow in an embodiment of the present disclosure.

[0102] In the disclosed embodiment, online workflow induction focuses on real-time workflow induction during task execution. Figure 5 In the process, the agent can continuously receive test instruction streams and try to complete the task by generating action trajectories. The action trajectories will be evaluated for success. The successful action trajectories will be converted into workflows through language model-based modules, and then the newly generated workflows will be integrated into the agent's enhanced memory system to enhance the agent's ability to handle subsequent tasks.

[0103] It can be seen that when performing tasks, the agent can dynamically learn workflows based on the results of successful task execution and store these workflows in the augmented memory system for future use. In other words, the agent can continuously update the workflows in its augmented memory system when encountering new situations, and dynamically improve the workflows in its augmented memory system. For example, when browsing a travel website, the agent may discover a new workflow, such as "find the nearest airport", and then immediately store this workflow in the augmented memory system and reuse it in subsequent similar tasks.

[0104] In addition, because the workflow stored in the agent's augmented memory system continues to expand over time, it ensures that the agent remains adaptable and efficient even without prior exposure to all task instructions or examples, and continuously improves its performance and generalization capabilities in interaction with new environments. In other words, this induction of online workflows ensures adaptive learning even in the absence of pre-labeled data, enabling the agent to evolve in new or unfamiliar environments.

[0105] It can be seen that in the disclosed embodiment, each successfully completed task can contribute to the agent's enhanced memory system by adding a new induced workflow. This dynamic update allows the agent to maintain relevance and effectiveness, and continuously learn from successes and failures. As workflows accumulate, the agent can build more complex routines, such as combining earlier workflows (e.g., "find a place by name", followed by "get its zip code"), etc., to obtain more comprehensive or longer sub-process workflows.

[0106] In the disclosed embodiment, two workflow induction methods are provided, namely online workflow induction and offline workflow induction. Among them, the effect of offline workflow induction is ideal when annotated data is readily available, and this method can provide performance improvement from the beginning. In addition, since the agent continuously improves its memory to match the ever-changing task environment, the method based on online workflow induction can be applied to various new situations. In this way, the workflow within the enhanced memory system created based on the combination of the aforementioned two workflow induction methods in the disclosed embodiment can ensure that the agent achieves optimal performance in various scenarios. Obviously, the disclosed embodiment provides a flexible and adaptive problem-solving framework for the agent, ensuring efficient, robust generalization and excellent task performance across various fields.

[0107] In the disclosed embodiment, enhanced memory systems for different fields or projects can be created for the intelligent agent based on the aforementioned workflow induction method, and then the obtained enhanced memory systems can be aggregated into the overall enhanced memory system of the intelligent agent.

[0108] In the embodiment of the present disclosure, after introducing the construction of the enhanced memory system, see Figure 6 , Figure 6 FIG. 1 is a flowchart of a task processing method in an embodiment of the present disclosure. The method process may be executed by an electronic device using an enhanced memory system, and the electronic device may be Figure 1 The electronic device 102 in the embodiment is executed, and the specific implementation process of the method is as follows:

[0109] Step 601: Receive tasks to be processed.

[0110] In an embodiment of the present disclosure, an agent may receive a task to be processed, and the task to be processed may be determined based on a model corresponding to the agent.

[0111] Optionally, if the agent is an agent based on a large language model, the task to be processed may be a natural language task, such as “find nearby restaurants”.

[0112] Step 602: Determine the associated workflow corresponding to the task to be processed according to the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks.

[0113] In the disclosed embodiment, the agent can search the enhanced memory system for an associated workflow corresponding to the task to be processed based on the business information of the task to be processed. That is, the associated workflow can be a stored workflow related to the current task to be processed.

[0114] Step 603: Determine the target workflow of the task to be processed according to the associated workflow, and execute the task to be processed according to the target workflow.

[0115] In the disclosed embodiment, the intelligent agent may also collect observation information from the environment, such as web page content or interactive elements, and assist in making decisions on actions to be performed based on the observation information.

[0116] In the disclosed embodiment, the agent can create a target workflow for the task to be processed based on the associated workflow, preset semantic understanding ability, and contextual reasoning ability. The agent can gradually execute actions (such as CLICK, FILL, SCROLL) based on the target workflow and observation information, or generate new actions when necessary. In addition, the agent monitors environmental feedback information to ensure that each action is consistent with the target action when executing the task to be processed, and adjusts the action when necessary to complete the execution of the task to be processed.

[0117] It can be seen that when encountering a task similar to the task to be processed, the agent will query the associated workflow in the memory system and integrate it into the context of the task to be processed. This approach not only ensures that the agent can complete the task quickly and reduce redundant operations, but also greatly improves the efficiency and success rate of task execution.

[0118] In the disclosed embodiment, after executing the task to be processed, the intelligent agent can determine whether the action trajectory corresponding to the task to be processed is reusable; when it is determined that the action trajectory corresponding to the task to be processed is reusable, the action trajectory corresponding to the task to be processed is converted into a new workflow, and the new workflow is stored in the enhanced memory system.

[0119] Optionally, the agent can determine the observation information corresponding to the interactive environment when executing the pending task; and determine the specific action information triggered based on the observation information; generate the action trajectory corresponding to the pending task according to the observation information and the corresponding specific action information. And, determine whether the action trajectory corresponding to the execution of the pending task meets the preset stop condition; the preset stop condition is used to filter the action trajectory corresponding to the successful execution result of the pending task; when it is determined that the action trajectory corresponding to the execution of the pending task meets the preset stop condition, it is determined that the action trajectory corresponding to the execution of the pending task can be reused. Among them, the preset stop condition can be the achievement of the task goal of the pending task or the receipt of a stop instruction.

[0120] In the disclosed embodiment, by extracting reusable workflows from the process of the agent performing tasks and storing them in an enhanced memory system, the accumulation and generalization of the agent's knowledge is achieved. Compared with the solution of performing tasks from scratch in the related art, this method can abstract the action trajectory of a successful task into a workflow with a universal placeholder, ensuring that it can adapt to changes in different task scenarios.

[0121] It can be seen that in the disclosed embodiments, the performance of language model-based agents is enhanced, enabling them to extract, store and reuse workflows derived from past task experience. Specifically, the agent can not only generate initial workflows from pre-collected data sets in offline mode, but also continuously summarize new workflows and update memory system solutions through real-time task execution in online mode. Through this memory-enhanced integration, the agent can quickly retrieve and apply existing workflows in future tasks, improve its adaptability and execution efficiency for complex tasks, and ensure that the agent maintains efficient and stable performance in a constantly changing environment.

[0122] The exemplary embodiment of the present disclosure also provides a task processing device. Figure 7 As shown, the task processing device 700 includes the following program modules:

[0123] A receiving unit 701, used for receiving tasks to be processed;

[0124] The determining unit 702 is used to determine the associated workflow corresponding to the task to be processed according to the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks;

[0125] The processing unit 703 is used to determine the target workflow of the task to be processed according to the associated workflow, and execute the task to be processed according to the target workflow.

[0126] In a possible implementation manner, the device further includes a creating unit, configured to:

[0127] Determine whether the action trajectory corresponding to the task to be processed is reusable;

[0128] When it is determined that the motion track corresponding to the task to be processed is reusable, the motion track corresponding to the task to be processed is converted into a new workflow, and the new workflow is stored in the enhanced memory system.

[0129] In a possible implementation manner, the device further includes a generating unit, configured to:

[0130] Determining observation information corresponding to the interactive environment when executing the task to be processed; and determining specific action information triggered based on the observation information;

[0131] An action trajectory corresponding to the task to be processed is generated according to the observation information and the corresponding specific action information.

[0132] In a possible implementation manner, the device further includes a creating unit, configured to:

[0133] The action track corresponding to the task to be processed is abstracted into a general workflow step, and placeholders are used to replace values ​​to obtain a new workflow.

[0134] In a possible implementation manner, the device further includes a creating unit, configured to:

[0135] Determine whether the motion trajectory corresponding to the execution of the task to be processed meets a preset stop condition; the preset stop condition is used to filter the motion trajectory corresponding to the successful execution result of the task to be processed;

[0136] When it is determined that the motion trajectory corresponding to the task to be processed satisfies a preset stop condition, it is determined that the motion trajectory corresponding to the task to be processed is reusable.

[0137] In a possible implementation manner, the processing unit 703 is configured to:

[0138] A target workflow for the task to be processed is created according to the associated workflow, the preset semantic understanding capability and the contextual reasoning capability.

[0139] In a possible implementation manner, the device further includes a creating unit, configured to:

[0140] Obtaining workflows of various tasks in the target project, and / or obtaining pre-collected annotated data sets; wherein the tasks to be processed belong to subtasks triggered in the target project;

[0141] The workflows of the various tasks and / or the workflows generated based on the data sets are stored in the enhanced memory system.

[0142] The specific details of each part of the above-mentioned device have been described in detail in the implementation method of the method part. The undisclosed details can be found in the implementation method of the method part, so they will not be repeated here.

[0143] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0144] The exemplary embodiments of the present disclosure also provide a computer program product, which includes a computer program, and when the computer program is executed by a processor, the task processing method described above is implemented.

[0145] In one embodiment, the computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing a computer program. The readable storage medium may be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid-state drive (SSD), and the like. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing a computer program, such as a read-only memory, a NAND flash memory, and the like.

[0146] In one embodiment, the computer program product may be an intangible product including a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as a digital file storing an executable file, an installation package, etc. of the computer program.

[0147] The code of the computer program can be written in one or more programming languages. Programming languages ​​such as C language, Java, C++, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., an Internet connection provided by an operator).

[0148] Computer programs can be carried or transmitted through electrical, magnetic, optical, electromagnetic, infrared and other signals. Electronic devices can convert signals carrying computer programs into digital signals to run computer programs. When a computer program runs on an electronic device, its code is used to enable the electronic device to execute (more specifically, it can enable the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure, such as the above-mentioned task processing method, which includes the following steps: Step 601: Receive a task to be processed; Step 602: Determine the associated workflow corresponding to the task to be processed based on the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various types of tasks; Step 603: Determine the target workflow of the task to be processed based on the associated workflow, and execute the task to be processed based on the target workflow.

[0149] By implementing the above method steps through a computer program, it is possible to receive a task to be processed, and then determine the associated workflow corresponding to the task to be processed based on the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system that stores workflows of various tasks; and the target workflow of the task to be processed can be determined based on the associated workflow, and the task to be processed can be executed based on the target workflow. It can be seen that in the disclosed embodiment, the intelligent agent can directly obtain the associated workflow from the enhanced memory system constructed by extracting and reusing based on past experience, and then determine the target workflow of the task to be processed based on the associated workflow, so as to quickly process the task to be processed and improve the efficiency of task execution.

[0150] The exemplary embodiments of the present disclosure also provide an electronic device, which may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the method steps of various exemplary embodiments of the present disclosure by executing the executable instructions.

[0151] Reference below Figure 8 , the electronic device is exemplarily described in the form of a general-purpose computing device. It should be understood that Figure 8 The electronic device 800 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0152] like Figure 8 As shown, the electronic device 800 may include: a processor 810 , a memory 820 , a bus 830 , an I / O (input / output) interface 840 , a network adapter 850 , and a display 880 .

[0153] The memory 820 may include a volatile memory, such as a RAM 821, a cache unit 822, and may also include a non-volatile memory, such as a ROM 823. The memory 820 may also include one or more program modules 824, such program modules 824 include but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof may include the implementation of a network environment. For example, the program module 824 may include each module in the above-mentioned device.

[0154] The processor 810 may include one or more processing units. For example, the processor 810 may include an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor and / or an NPU (Neural-Network Processing Unit), etc.

[0155] The processor 810 can be used to execute executable instructions stored in the memory 820, such as the above-mentioned task processing method, which includes the following steps: Step 601: Receive the task to be processed; Step 602: Determine the associated workflow corresponding to the task to be processed based on the business information of the task to be processed; wherein the associated workflow is obtained from an enhanced memory system that stores the workflows of various tasks; Step 603: Determine the target workflow of the task to be processed based on the associated workflow, and execute the task to be processed based on the target workflow.

[0156] By executing the above method steps through the processor 810, a task to be processed can be received, and then, based on the business information of the task to be processed, an associated workflow corresponding to the task to be processed can be determined; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks; and, based on the associated workflow, a target workflow of the task to be processed can be determined, and the task to be processed can be executed based on the target workflow. It can be seen that in the embodiment of the present disclosure, the intelligent agent can directly obtain the associated workflow from the enhanced memory system constructed by extracting and reusing based on past experience, and then determine the target workflow of the task to be processed based on the associated workflow, so as to quickly process the task to be processed and improve the efficiency of task execution.

[0157] The bus 830 is used to realize the connection between different components of the electronic device 800, and may include a data bus, an address bus, and a control bus.

[0158] The electronic device 800 can communicate with one or more external devices 900 (eg, a keyboard, a mouse, an external controller, etc.) through the I / O interface 840 .

[0159] The electronic device 800 can communicate with one or more networks through the network adapter 850. For example, the network adapter 850 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, near field communication, etc. The network adapter 850 can communicate with other modules of the electronic device 800 through the bus 830.

[0160] The electronic device 800 may display observation information and the like through the display 880 .

[0161] although Figure 8 Not shown, other hardware and / or software modules may also be provided in the electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0162] As can be seen from the above, the technical solution of the present disclosure can be implemented as a method, an apparatus, a system, a computer program product, a storage medium, an electronic device, etc. Those skilled in the art can understand that various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, such as being respectively referred to as a "circuit", "module" or "system".

[0163] It should be understood that the present disclosure is not limited to the specific method steps or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. Those skilled in the art will easily think of other embodiments based on the specific embodiments provided by the present disclosure. Therefore, the specific embodiments provided by the present disclosure are only exemplary, and the scope and spirit of the present disclosure are indicated by the claims, and any variations, uses or adaptive changes of the present disclosure should be covered, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the technical field that are not disclosed in the present disclosure.

Claims

1. A task processing method, characterized in that: The method comprises: Receive pending tasks; Determine, according to the business information of the task to be processed, an associated workflow corresponding to the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks; According to the associated workflow, a target workflow of the task to be processed is determined, and according to the target workflow, the task to be processed is executed.

2. The method according to claim 1, characterized in that The method further comprises: Determine whether the action trajectory corresponding to the task to be processed is reusable; When it is determined that the motion track corresponding to the task to be processed is reusable, the motion track corresponding to the task to be processed is converted into a new workflow, and the new workflow is stored in the enhanced memory system.

3. The method according to claim 2, characterized in that Before determining whether the motion trajectory corresponding to the task to be processed is reusable, the method further includes: Determining observation information corresponding to the interactive environment when executing the task to be processed; and determining specific action information triggered based on the observation information; An action trajectory corresponding to the task to be processed is generated according to the observation information and the corresponding specific action information.

4. The method according to claim 2, characterized in that: Converting the action trajectory corresponding to the task to be processed into a new workflow includes: The action track corresponding to the task to be processed is abstracted into a general workflow step, and placeholders are used to replace values ​​to obtain a new workflow.

5. The method according to any one of claims 1 to 4, characterized in that: Determining whether the action trajectory corresponding to the task to be processed is reusable includes: Determine whether the motion trajectory corresponding to the execution of the task to be processed meets a preset stop condition; the preset stop condition is used to filter the motion trajectory corresponding to the successful execution result of the task to be processed; When it is determined that the motion trajectory corresponding to the task to be processed satisfies a preset stop condition, it is determined that the motion trajectory corresponding to the task to be processed is reusable.

6. The method according to any one of claims 1 to 4, characterized in that: Determining the target workflow of the task to be processed according to the associated workflow includes: A target workflow for the task to be processed is created according to the associated workflow, the preset semantic understanding capability and the contextual reasoning capability.

7. The method according to any one of claims 1 to 4, characterized in that: Before receiving the task to be processed, the method further includes: Obtaining workflows of various tasks in the target project, and / or obtaining pre-collected annotated data sets; wherein the tasks to be processed belong to subtasks triggered in the target project; The workflows of the various tasks and / or the workflows generated based on the data sets are stored in the enhanced memory system.

8. A task processing device, characterized in that: The device comprises: A receiving unit, used for receiving tasks to be processed; A determination unit, configured to determine, according to the business information of the task to be processed, an associated workflow corresponding to the task to be processed; wherein the associated workflow is obtained from an enhanced memory system storing workflows of various tasks; The processing unit is used to determine the target workflow of the task to be processed according to the associated workflow, and execute the task to be processed according to the target workflow.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: processor; A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 7 by executing the executable instructions.

Images