CN-122021698-A - Task processing method and device, electronic equipment and storage medium
Abstract
The embodiment of the application discloses a task processing method, a device, electronic equipment and a storage medium, relates to the technical field of artificial intelligence, and can realize efficient processing of multiple-loop complex tasks and improve the reliability of task processing results. The method comprises the steps of receiving a target task processing request input by a user, determining a thinking chain skill adapted to the target task based on the target task processing request, wherein the thinking chain skill describes a solution of the target task, arranging an execution flow of the target task by using the thinking chain skill according to the solution described by the thinking chain skill, wherein the execution flow comprises at least one step needing to be executed by using an atomic skill, executable logic codes for realizing a preset function are packaged in the atomic skill, the executable logic codes are called when the atomic skill is used for realizing the preset function, and the atomic skill execution step is used according to the arranged execution flow so as to obtain a processing result of the target task. The method is suitable for complex task processing scenes of the intelligent agent.
Inventors
- Zou Guanyun
- LIU ZEMING
Assignees
- 物自体(上海)科技有限公司
Dates
- Publication Date
- 20260512
- Application Date
- 20251231
Claims (10)
- 1. A method of task processing, the method comprising: receiving a target task processing request input by a user; Determining, based on the target task processing request, a mental chain skill adapted to the target task, the mental chain skill describing a solution to the target task; Applying the mental chain skills to arrange an execution flow of the target task based on a solution described by the mental chain skills, wherein the execution flow comprises at least one step needing to be executed by using atomic skills, executable logic codes for realizing preset functions are packaged in the atomic skills, and the executable logic codes are called when the atomic skills are applied to realize the preset functions; And executing the steps by using the atomic skills according to the arranged execution flow so as to obtain the processing result of the target task.
- 2. The task processing method according to claim 1, wherein the mental chain skills include a mental chain description including at least one of checklists, step sequences, branching conditions, and feedback loop information, wherein: the checklist describes the content to be checked by the checkpoints in the solution or the input information to be acquired in the solution; the sequence of steps describes the execution steps in the solution; the branch condition describes logic judgment conditions of different execution paths in the solution; the feedback loop information describes the processing principles for the predetermined event.
- 3. The task processing method according to claim 1, wherein the atomic skills include a data retrieval skill, a form statistical analysis skill, a report generation skill, an image generation skill, and a video script microscope generation skill.
- 4. The task processing method according to claim 1, characterized in that, The determining, based on the target task processing request, a mental chain skill adapted to the target task includes: Dividing the target task into at least two task stages based on the target task processing request; determining, for each of the task phases, a mental chain skill adapted to the task phase; The employing the mental chain skills to orchestrate execution flow of the target task with a solution based on the mental chain skill descriptions includes: Arranging an execution flow of the task stage by using the thought chain skills matched with the task stage, wherein the execution flow of the task stage comprises at least one step needing to be executed by using atomic skills; The step of executing the steps by using the atomic skills according to the scheduled execution flow to obtain the processing result of the target task comprises the following steps: Aiming at each task stage, executing the steps by using the atomic skills according to the scheduled execution flow of the task stage so as to acquire an execution result of each task stage; And acquiring an execution result of the target task based on the execution result of each task stage.
- 5. The task processing method according to claim 4, characterized in that, The task stages comprise a first task stage and a second task stage which are sequentially executed; after obtaining the execution result of the first task stage, the method further includes: Evaluating whether the execution result of the first task stage meets a predetermined quality condition; if the execution result of the first task stage meets the preset quality condition, executing the step by using the atomic skills according to the scheduled execution flow of the second task stage so as to acquire the execution result of the second task stage; And if the execution result of the first task stage does not meet the preset quality condition, determining a reason for the execution result not meeting the preset quality condition, and determining a subsequent execution path based on the reason to adjust the execution result of the first task stage.
- 6. The task processing method according to claim 1, wherein the determining, based on the target task processing request, a mental chain skill adapted to the target task includes: Providing the target task processing request to a mental chain skill agent, so that the mental chain agent determines the mental chain skill adapted to the target task based on the task processing request, and further applies the mental chain skill to program the execution flow of the target task based on the solution described by the mental chain skill.
- 7. The task processing method according to claim 6, wherein said executing said step using said atomic skills according to said scheduled execution flow comprises: Receiving the arranged execution flow provided by the thinking chain skill agent, and providing the arranged execution flow for an atomic skill agent so that the atomic skill agent executes the steps by using the atomic skill according to the arranged execution flow; Or alternatively And providing the arranged execution flow to an atomic skill agent through the thinking chain agent, so that the atomic skill agent executes the steps by using the atomic skill according to the arranged execution flow.
- 8. A task processing device, the device comprising: The data receiving module is used for receiving a target task processing request input by a user; A first processing module for determining, based on the target task processing request, a mental chain skill adapted to the target task, the mental chain skill describing a solution to the target task; A second processing module, configured to apply the mental chain skill to arrange an execution flow of the target task based on a solution described by the mental chain skill, where the execution flow includes at least one step that needs to be performed using an atomic skill, where the atomic skill is encapsulated with executable logic code for implementing a predetermined function, and when the atomic skill is applied, the executable logic code is called to implement the predetermined function; and the first execution module is used for executing the steps by utilizing the atomic skills according to the arranged execution flow so as to acquire the processing result of the target task.
- 9. An electronic device comprising a processor and a memory, wherein the memory is configured to store executable program code, and wherein the processor is configured to execute a program corresponding to the executable program code by reading the executable program code stored in the memory, and to perform the task processing method according to any one of the preceding claims 1 to 7.
- 10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores one or more programs, the one or more programs being executable by one or more processors to implement the task processing method of any of the preceding claims 1 to 7.
Description
Task processing method and device, electronic equipment and storage medium Technical Field The present application relates to the field of artificial intelligence technologies, and in particular, to a task processing method, a task processing device, an electronic device, and a storage medium. Background With the rapid progress of large language model (Large Language Model, LLM) technology, agents are increasingly being applied to diversified business scenarios by virtue of autonomous sensing, decision making, execution and other capabilities, and intelligently process various tasks in the business scenarios. Taking content marketing as an example, part of the link's intelligent work can be undertaken by the agent, for example, generating an activity document, writing product introduction, providing choice advice, etc. In the process of processing tasks, the intelligent agent can call a series of available tools, but when processing content marketing tasks of multi-link complex engineering, such as market analysis, a plurality of links such as data acquisition, cleaning, analysis and text formation are needed, the intelligent agent is required to understand task targets and is responsible for selecting and combining tools, deviation easily occurs in the task processing process, the processing efficiency is low, and the reliability of task processing results is low. Disclosure of Invention In view of the above, the embodiment of the application provides a task processing method, which can realize the efficient processing of multiple-loop complex tasks and improve the reliability of task processing results. In a first aspect, an embodiment of the present application provides a task processing method, where the method includes receiving a target task processing request input by a user, determining, based on the target task processing request, a mental chain skill adapted to the target task, where the mental chain skill describes a solution of the target task, applying the mental chain skill to arrange an execution flow of the target task based on the solution described by the mental chain skill, where the execution flow includes at least one step that needs to be performed using an atomic skill, where the atomic skill is packaged with executable logic code for implementing a predetermined function, and when the atomic skill is applied, the executable logic code is called to implement the predetermined function, and executing the step using the atomic skill according to the arranged execution flow to obtain a processing result of the target task. According to one embodiment of the application, the thinking chain skill comprises a thinking chain description, wherein the thinking chain description comprises at least one of a checklist, a step sequence, a branch condition and feedback loop information, the checklist describes content to be checked by checkpoints in the solution or input information to be acquired in the solution, the step sequence describes execution steps in the solution, the branch condition describes logic judgment conditions of different execution paths in the solution, and the feedback loop information describes processing principles for preset events. According to one embodiment of the present application, the atomic skills include data retrieval skills, form statistical analysis skills, report generation skills, image generation skills, video script microscope generation skills. According to one embodiment of the application, the method for determining the mind chain skills adaptive to the target task based on the target task processing request comprises dividing the target task into at least two task stages based on the target task processing request, determining the mind chain skills adaptive to the task stages for each task stage, wherein the step of using the mind chain skills to arrange the execution flow of the target task based on the solutions described by the mind chain skills comprises the step of using the mind chain skills adaptive to the task stages to arrange the execution flow of the task stages, wherein the execution flow of the task stages comprises at least one step needing to be executed by using atomic skills, and the step of using the atomic skills to acquire the processing results of the target task comprises the step of using the atomic skills to acquire the execution results of each task stage for each task stage according to the arranged execution flow of the task stage, and the step of using the atomic skills to acquire the execution results of each task stage based on the acquired task execution results of the target task stage. According to one embodiment of the application, the task stages comprise a first task stage and a second task stage which are sequentially executed, after the execution result of the first task stage is obtained, the method further comprises the steps of evaluating whether the execution result of the first task stage meets a preset quality