Action sequence execution method based on agent, related device and computer program product

Abstract

The invention provides an action sequence execution method and device based on an agent, electronic equipment, a computer readable storage medium and a computer program product, and relates to the technical field of artificial intelligence such as deep learning, data processing and intelligent scheduling. According to one specific embodiment, the method comprises the steps that in response to the fact that a first confidence coefficient of an agent processing result of an action node in an action sequence is smaller than a confidence coefficient threshold value, whether an updated agent exists or not is inquired, and the updated agent can be used for achieving an agent calling intention indicated by the action node; in response to the existence of the update agent, calling the update agent, and determining an update agent processing result corresponding to the action node; and in response to the fact that the second confidence of the updated agent processing result is greater than or equal to the confidence threshold, replacing the agent processing result with the updated agent processing result. Therefore, 'results which do not meet requirements' can be prevented from flowing into subsequent action nodes, and the quality of execution results obtained by executing the action sequence based on the intelligent agent is guaranteed and improved while the waste of operation resources is reduced.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, specifically to the field of artificial intelligence technology such as deep learning, data processing and intelligent scheduling, and especially to an action sequence execution method, device, electronic device, computer-readable storage medium and computer program product based on an intelligent agent. Background Art

[0002] Intelligent agents, for example, can be based on neural network algorithms to mimic the way the human brain processes information. Intelligent agents can be trained using large amounts of data to have the ability to handle tasks and implement functions. For example, after training, intelligent agents can have the ability to perform complex tasks such as image recognition, natural language processing, and predictive analysis by recognizing patterns and features.

[0003] In this context, in order to improve the capabilities of intelligent agents and enable them to handle larger and more complex tasks, as one of the strategies, we can choose to use multiple intelligent agents in a joint and continuous manner, and use multiple different intelligent agents in the form of "intelligent agent groups" or "intelligent agent sets" to improve processing performance. For example, the order of calling the intelligent agents and the specific intelligent agents to be called can be indicated by editing the action sequence. Then, the action sequence can be used to instruct the electronic device to call these intelligent agents continuously in the calling order. Summary of the invention

[0004] The embodiments of the present disclosure provide an agent-based action sequence execution method, device, electronic device, computer-readable storage medium, and computer program product.

[0005] In a first aspect, an embodiment of the present disclosure proposes an agent-based action sequence execution method, comprising: in response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence threshold, querying whether there is an update agent, wherein the update agent can be used to implement the agent calling intention indicated by the action node; in response to the existence of an update agent, calling the update agent to determine the update agent processing result corresponding to the action node; in response to a second confidence level of the update agent processing result being greater than or equal to the confidence threshold, replacing the agent processing result with the update agent processing result.

[0006] In the second aspect, an embodiment of the present disclosure proposes an agent-based action sequence execution device, including: an update agent query unit, configured to query whether an update agent exists in response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence threshold, wherein the update agent can be used to implement the agent calling intention indicated by the action node; an update result determination unit, configured to call the update agent in response to the existence of the update agent, and determine the update agent processing result corresponding to the action node; an update result replacement unit, configured to replace the agent processing result with the update agent processing result in response to a second confidence level of the update agent processing result being greater than or equal to the confidence threshold.

[0007] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: 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 when the at least one processor executes, it is able to implement the agent-based action sequence execution method described in any implementation method in the first aspect.

[0008] In a fourth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, which are used to enable a computer to implement the agent-based action sequence execution method described in any implementation method of the first aspect when executed.

[0009] In a fifth aspect, an embodiment of the present disclosure provides a computer program product comprising a computer program, which, when executed by a processor, can implement the agent-based action sequence execution method as described in any implementation manner in the first aspect.

[0010] The agent-based action sequence execution method, apparatus, electronic device, computer-readable storage medium and computer program product provided by the embodiments of the present disclosure query whether there is an updated agent in response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence threshold, wherein the updated agent can be used to implement the agent calling intention indicated by the action node; in response to the existence of an updated agent, the updated agent is called to determine the updated agent processing result corresponding to the action node; in response to a second confidence level of the updated agent processing result being greater than or equal to the confidence threshold, the updated agent processing result is replaced with the updated agent processing result.

[0011] The present disclosure can monitor the calling and using results of the agent in the process of calling and using the agent based on the action sequence. Then, when it is detected that the calling and using results of the agent of the action node do not meet the requirements, other updated agents are used to reprocess the agent calling intention indicated by the action node to achieve timely update of the results that do not meet the requirements. In this way, it is possible to avoid "errors" and "results that do not meet the requirements" from erroneously flowing into subsequent action nodes, while reducing the waste of computing resources, ensuring and improving the quality of the execution results obtained by executing the action sequence based on the agent.

[0012] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Other features, objects and advantages of the present disclosure will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0014] Figure 1 is an exemplary system architecture in which the present disclosure may be applied;

[0015] Figure 2 A flowchart of an agent-based action sequence execution process provided in an embodiment of the present disclosure;

[0016] Figure 3 A flowchart of a concurrent agent-based action sequence execution process provided by an embodiment of the present disclosure;

[0017] Figure 4 A flowchart of a process of generating an action sequence provided by an embodiment of the present disclosure;

[0018] Figure 5 A schematic diagram of a flow chart of an agent-based action sequence execution process in an application scenario provided by an embodiment of the present disclosure;

[0019] Figure 6 A structural block diagram of an agent-based action sequence execution device provided in an embodiment of the present disclosure;

[0020] Figure 7 A schematic diagram of the structure of an electronic device suitable for executing an agent-based action sequence execution method provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. It should be noted that the embodiments in the present disclosure and the features in the embodiments may be combined with each other without conflict.

[0022] In addition, in the technical solutions involved in the present disclosure, the acquisition, storage, use, processing, transportation, provision and disclosure of user personal information involved (for example, in some scenarios, the action sequence involved in the present disclosure may be provided by the user) shall comply with the relevant laws and regulations and shall not violate public order and good morals.

[0023] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the agent-based action sequence execution method, apparatus, electronic device, and computer-readable storage medium of the present disclosure can be applied.

[0024] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0025] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Terminal devices 101, 102, 103 and server 105 may be installed with various applications for realizing information communication between the two, such as task processing type applications.

[0026] Terminal devices 101, 102, 103 and server 105 can be hardware or software. When terminal devices 101, 102, 103 are hardware, they can be various electronic devices with display screens, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, etc.; when terminal devices 101, 102, 103 are software, they can be installed in the electronic devices listed above, which can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here. When server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server; when the server is software, it can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here.

[0027] The server 105 can provide various services through various built-in applications. Taking a task processing application that can provide task processing based on at least an agent as an example, the server 105 can achieve the following effects when running the task processing application: first, the pending tasks to be processed are obtained from the terminal devices 101, 102, and 103 through the network 104, and then a corresponding action sequence is generated based on the pending tasks; then, the server 105 implements the corresponding action (for example, calling the agent) based on the instructions in the action sequence; in this process, the server 105 responds to the confidence of the agent processing result of the action node in the action sequence (for the convenience of description, it is called the first confidence) being less than the confidence threshold, and inquiries whether there is an updated agent, wherein the updated agent can be used to implement the agent calling intention indicated by the action node; then, in response to the existence of the updated agent, the server 105 calls the updated agent and determines the updated agent processing result corresponding to the action node; finally, in response to the second confidence of the updated agent processing result being greater than or equal to the confidence threshold, the server 105 replaces the agent processing result with the updated agent processing result.

[0028] It should be understood that the above process is only an example. In some scenarios, the server 105 may also "detect and monitor" the action sequence that is not directly executed locally in a similar manner, and update the agent processing results where the first confidence is less than the confidence threshold. For example, the action sequence may be executed by other devices (for example, terminal devices 101, 102, 103).

[0029] It should be noted that, in addition to being obtained from the terminal devices 101, 102, 103 through the network 104, the tasks to be executed can also be pre-stored in the server 105 in various ways. Therefore, when the server 105 detects that the data has been stored locally (for example, before starting to process the tasks to be executed that were previously stored), it can choose to directly obtain the data locally. In this case, the exemplary system architecture 100 may also not include the terminal devices 101, 102, 103 and the network 104.

[0030] Since monitoring action sequences, calling agents, etc. may require more computing resources and stronger computing power, the agent-based action sequence execution method provided in the subsequent embodiments of the present disclosure is generally executed by the server 105 with stronger computing power and more computing resources. Accordingly, the agent-based action sequence execution device is generally also set in the server 105. However, it should also be pointed out that when the terminal devices 101, 102, and 103 also have computing power and computing resources that meet the requirements, the terminal devices 101, 102, and 103 can also complete the above-mentioned various operations that are handed over to the server 105 through the task processing application installed thereon, and then output the same result as the server 105. Especially in the case where there are multiple terminal devices with different computing powers at the same time, but when the task processing application determines that the terminal device where it is located has stronger computing power and more remaining computing resources, it can allow the terminal device to perform the above-mentioned operations, thereby appropriately reducing the computing pressure of the server 105. Accordingly, the agent-based action sequence execution device can also be set in the terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may also not include the server 105 and the network 104 .

[0031] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0032] Please refer to Figure 2 , Figure 2 A flowchart of an agent-based action sequence execution process provided by an embodiment of the present disclosure, wherein process 200 includes the following steps:

[0033] Step 201: In response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence level threshold, querying whether there is an update agent;

[0034] This step is intended to be performed by the execution subject of the agent-based action sequence execution method (e.g. Figure 1The server 105 shown responds when determining that the first confidence of the agent processing result of the action node in the action sequence is less than the confidence threshold to query whether there is an update agent.

[0035] As discussed above, the agent-based action sequence can correspond to the task to be performed. For example, the action sequence can correspond to a task flow chart of the task to be performed, which is used to call the corresponding agent according to the instructions of the task flow chart to utilize one or more agents to "process the task."

[0036] An agent is an intelligent entity, which can be, for example, an algorithm model that performs some intelligent analysis and decision-making work, a large model agent (LLM Agent), or an expert rule-based agent (Rule-based Agent). In some embodiments, the agent can also be a "human agent" implemented based on artificial collaboration.

[0037] Accordingly, the task flow chart can provide and indicate the task flow in a manner of indicating the process of data change. For example, the task flow chart can be composed of multiple task nodes, each of which records the expected "node data" expected in the task node, thereby providing the task flow chart in a manner of data change process.

[0038] For example, in task flow chart A, the first task node may correspond to node data a1, the second task node may correspond to node data a2, and the third task node (for example, the task node may correspond to the task node required to be reached when the task flow chart is completed) may correspond to node data a3. Thus, based on the task flow chart, it can be determined that the purpose of the task to be performed is to change node data a1 into node data a2, and then change and process node data a2 into node data a3. That is, the "task" actually indicated by the task flow chart is to process node data a1 into node data a2 to finally obtain node data a3.

[0039] Accordingly, the execution subject can determine the corresponding data processing actions based on such a change process and flow. For example, for the first task node, it is necessary to call an agent that can process node data a1 into node data a2. Then, the execution subject can determine the action sequence based on these data processing actions. Each action node in the action sequence can correspond to a data processing action (implementation intention), that is, the agent's call action, or the agent's call intention.

[0040] Generally, the agent call intent may indicate the function of the agent that is expected to be called. For example, in the above example, for the first task node, the function of the agent to be called is to process node data a1 into node data a2. Accordingly, the agent call intent may also be determined as "calling an agent that has the function of processing node data a1 into node data a2".

[0041] In some optional implementations of this embodiment, after obtaining the task description text of the task to be performed, the execution subject can generate a task flow chart for the task to be performed based on the task entity recognition result of the task description text. For example, the task description text can be a task flow described in natural language provided by the provider of the task to be performed (e.g., the user). For example, "process node data a1 into node data a3 via node data a2".

[0042] Accordingly, the execution subject can determine the various "task nodes" of the task that the user expects to perform based on the task entity recognition result of the task description text, for example, through an entity recognition model, a natural language processing (Neuro-Linguistic Programming) model (or an "intelligent agent" with such capabilities). For example, the determined "entities" can be data a1, a2, and a3. Then, after the execution subject determines the task nodes based on these entities, it obtains the task flow chart based on the arrangement results of the task nodes.

[0043] Finally, the execution subject can choose to generate an action sequence based on the task flow chart as discussed above, thereby allowing the user to describe the task in natural language, so that the user can provide the task requirements of the task to be executed to the execution subject more simply and efficiently.

[0044] In some embodiments, the action sequence can be pre-generated and edited. Similarly, the action sequence can be obtained by the execution subject directly from a local storage device, or from a non-local storage device (e.g. Figure 1 The local storage device may be a data storage module provided in the execution subject, such as a server hard disk, in which case the action sequence can be quickly read locally; the non-local storage device may also be any other electronic device provided for storing data, such as some user terminals, in which case the execution subject may obtain the required action sequence by sending an acquisition command to the electronic device.

[0045] Next, as discussed above, regardless of whether the action sequence is executed locally by the execution subject or by a device outside the execution subject, the execution subject can detect the first confidence of the intelligent agent processing result output after the action node is executed during the execution of the action sequence.

[0046] In some optional implementations of this embodiment, the agent called by the action sequence can be configured with the ability to generate confidence in the agent's processing results (for example, the agent can be configured with a classification head so that the agent can use the classification head to provide its "confidence" for the output result). For example, an output confidence classification head (often called the classification head or output layer) can be configured in the agent to generate the agent's confidence in the agent's processing results.

[0047] In some optional implementations of this embodiment, the execution entity may also choose to use a scoring agent to "evaluate" the processing results of the agent to determine the confidence level of the agent's processing results.

[0048] Then, the execution subject can respond when it is determined that the first confidence level of the agent processing result is less than the confidence threshold (usually preset based on the standard that the "agent processing result" can be considered "unreliable" or needs to be updated), and inquire whether there is an update agent. The update agent can also be used to implement the agent call intention indicated by the action node. For example, in the above example, for the first action node, the update agent can be another agent different from the currently used agent, but the node data a1 is processed into the node data a2.

[0049] It should be understood that the updated agent may be the same or different in scale from the original agent in use. For example, the updated agent may be another agent with stronger agent processing capabilities, more functions, or a larger scale than the original agent in use.

[0050] Step 202: In response to the existence of an update agent, calling the update agent and determining the update agent processing result corresponding to the action node;

[0051] On the basis of step 201, in this step, if such an update agent can be searched, the execution subject can respond to this, call the update agent, and determine the update agent processing result corresponding to the action node. That is, the execution subject can call the update agent to "re" execute the agent call intention indicated by the action node. For example, in the above example, the node data a1 is reprocessed into node data a2 using the update agent. For ease of understanding, the node data a2 processed by the original agent can be described as a2', while the node data a2 processed by the update agent can be described as a2". a2' and a2" may have different specific contents, but both can be considered as "node data a2" data (for example, a2' and a2" have the same data type, but different specific values).

[0052] In addition, in order to explain it more clearly and in detail, the specific source of updating the intelligent agent will be discussed in more detail below. Therefore, the relevant content can refer to the content discussed below, which will not be expanded here.

[0053] Step 203: In response to the second confidence of the updated agent processing result being greater than or equal to the confidence threshold, the updated agent processing result is used to replace the agent processing result.

[0054] Based on step 202, this step aims to replace the agent processing result with the updated agent processing result when the confidence level of the updated agent processing result obtained by the updated agent (described as the second confidence level for the convenience of description) is greater than or equal to the above-mentioned confidence threshold.

[0055] That is, after calling the update agent to obtain the update agent processing result, the execution subject can still choose to "detect" and "compare" the numerical relationship between the second confidence level (the second confidence level can be generated and determined in a similar manner to the first confidence level, which will not be described here) of the update agent processing result and the confidence threshold. Accordingly, if the second confidence level is greater than or equal to the above confidence threshold, the execution subject can determine that the update agent processing result can meet the "trustworthy" usage requirements, and the execution subject uses the update agent processing result to replace the agent processing result.

[0056] As a result, other action nodes in the action node that follow the action node can use the updated intelligent agent processing result as input to perform subsequent processing (for example, use a" as the source data for subsequent execution and determination of node data a3, rather than a'). In this way, the result detection and update in the "action process" can be achieved without having to make adjustments after all action nodes are completed and the final result is "unqualified and does not meet the requirements of use". This can not only avoid resource losses caused by erroneous calculations using unreliable intelligent agent processing results, but also save the update resources needed to find and troubleshoot "unreliable" and ensure the quality of the output results of the action sequence.

[0057] The agent-based action sequence execution method provided by the disclosed embodiment can monitor the calling and use results of the agent in the process of calling and using the agent based on the action sequence. Then, when it is detected that the agent calling and use results of the action node do not meet the requirements, other updated agents are used to reprocess the agent calling intention indicated by the action node to achieve timely update of the results that do not meet the requirements. In this way, it is possible to avoid "errors" and "results that do not meet the requirements" from erroneously flowing into subsequent action nodes, while reducing the waste of computing resources, ensuring and improving the quality of the execution results obtained by executing the agent-based action sequence.

[0058] In some embodiments, in order to improve the execution efficiency of the action sequence, it is also possible to choose to adjust the execution strategy of the action sequence during the execution of the agent-based action sequence so that the action sequence can be executed more efficiently.

[0059] For this, please refer to Figure 3 , Figure 3 A flowchart of a concurrent agent-based action sequence execution process provided by an embodiment of the present disclosure, wherein process 300 includes the following steps:

[0060] Step 301: Determine a sub-flowchart based on the task dependency relationship in the task flow chart of the task to be executed;

[0061] Specifically, the execution subject can determine the task dependency between each task node in the task flow chart using the task to be executed. For example, the execution subject can express the task flow chart in the form of a directed acyclic graph (DAG). A directed acyclic graph is a graph structure in which each edge has a direction and there is no loop.

[0062] The execution subject can choose to determine the task with no prior dependency from the task flow chart (for example, a task that does not depend on the execution result of any task to start execution). It should be understood that the task with no prior dependency can only be executed without relying on the execution result of other tasks, and it itself can serve as the "prior dependency" of other tasks. The execution subject can take the prior dependent task as the "start" and continuously determine the "complete task chain" based on the subsequent dependency relationship of the task with no prior dependency. That is, the "task chain" starts with the task with no prior dependency and ends with the last dependent task that is not relied on by other tasks, and each task in the "task chain" has a prior and / or subsequent dependency relationship with each other.

[0063] Then, the execution subject can split the task flow chart based on each task chain to obtain a sub-flow chart. The tasks included in the sub-flow chart belong to the same task chain, and the first task in the sub-flow chart is a task without a prior dependency.

[0064] Step 302: Generate an action sequence corresponding to the sub-flowchart;

[0065] Specifically, the execution subject may generate the corresponding action sequence for each sub-flowchart in the manner of generating the action sequence for the task flow chart.

[0066] Step 303: In response to the existence of at least two action sequences, execute the at least two action sequences concurrently.

[0067] Specifically, if the execution subject finally determines more than two sub-flowcharts, that is, there are more than two action sequences corresponding to the sub-flowcharts, then the execution subject can respond to this and execute these action sequences in a concurrent manner (for example, executing the agent call according to the actions executed by each action sequence at the same time). For example, the execution subject can choose to assign corresponding threads to each action sequence, so as to use each thread to execute these action sequences in parallel and concurrently. Therefore, the execution subject can use "no prior dependent tasks" as a benchmark, split out multiple independent and independent task chains, realize concurrency between tasks, and improve execution efficiency.

[0068] In some optional implementations of this embodiment, similar to the above "tasks without prior dependencies", the parallel branches in the task chain are further split to further improve efficiency. For example, the execution subject can further traverse each sub-flowchart to determine the task nodes where branches exist (i.e., a task node is simultaneously dependent on at least two subsequent task nodes).

[0069] If there is such a task node with branches, the execution body may choose to further split the sub-flowchart here (for example, each branch may be regarded as a new sub-flowchart) to achieve concurrency of the "branches".

[0070] In some optional implementations of this embodiment, although the non-previous-dependent tasks do not depend on other tasks, there may be conflicts between the non-previous-dependent tasks in the action sequences corresponding to different sub-flowcharts. For example, the non-previous-dependent tasks in two action sequences may need to call the same target agent.

[0071] For such a situation, for ease of understanding, these two (or more) action sequences can be described as conflicting action sequences. If there are at least two conflicting action sequences, at least two conflicting action sequences simultaneously rely on the same target agent to complete the first action node. The execution subject can select the descending order of the length of the estimated execution time of the conflicting action sequence to determine the order of utilization of the conflicting action sequence for the target agent. The utilization order corresponds to the arrangement order indicated by the descending order of the length.

[0072] Therefore, considering that the final execution result of the action sequence depends on the full completion of the action sequences of each sub-flowchart, the execution subject can choose to allocate the target intelligent agent to the action sequence with long execution time with higher priority, so that the action sequence with long execution time can be actually started and executed earlier, so as to improve the overall execution efficiency.

[0073] In some embodiments, as discussed above, for the convenience of users, users can provide "expectations" to the execution subject by providing task description text. Accordingly, the task flow chart can be generated based on the task entity recognition results of the task description text. In such a case, the execution subject can determine whether the task flow is "reasonable" (for example, whether it can be achieved through a single agent call) based on the degree of association between the adjacent tasks in the task flow chart. Thus, user consumption and user usage requirements can be further reduced. For example, the user does not have to provide the complete data change status to complete the task flow chart (for example, in the above example, the user can provide the requirement for node data a1 to change to node data a3, and then the configuration can be completed).

[0074] Specifically, after generating the task flow chart, the execution subject can detect the task correlation between the first task and the second task that are adjacent to each other (for example, the first task is before the second task and is adjacent to the second task). For example, the execution subject can determine whether two data types can be converted through "one step". Accordingly, the corresponding correlation threshold can also be determined in advance based on the estimation and determination of the capabilities of the intelligent agent (for example, it can be determined based on the degree of correlation that is determined to be impossible to achieve using one intelligent agent). In this case, if the execution subject determines that the task correlation between the first task and the second task in the task flow chart is less than the correlation threshold, it can respond to this and determine a third task for connecting the first task and the second task.

[0075] The third task is a task that serves as an "intermediate state" to connect the first task and the second task. For example, when converting U data to I data, two agents need to be called continuously (for example, in the case of agent A and agent B). The addition of the third task allows the overall task flow to be changed to calling agent A to convert U data into intermediate data such as O data at the first task, and for the third task, agent B can be called to convert O data into I data. Therefore, using the third task to connect the first task and the second task can reduce user consumption and difficulty of use while avoiding action jumps in action nodes caused by the lack of task nodes provided by the user and the inability to determine the specific agent calling intention.

[0076] Accordingly, the execution subject may update the third task to the task flow chart to obtain an updated task flow chart, and when subsequently generating an action sequence based on the task flow chart, the action sequence may be generated based on the updated task flow chart.

[0077] In some embodiments, in order to facilitate users to provide, edit and adjust task flowcharts, the execution entity may also choose to provide users with a visual view of the task flowchart so that users can understand the task flowchart more intuitively and edit it conveniently.

[0078] Specifically, the execution subject can present a visual view of the task flow chart based on the pre-configured first communication path using a first device, for example, a terminal device used by a user that is determined in advance by the protocol (for example, terminal devices 101, 102, and 103). For example, a display component (for example, a screen) of the first device is used to present a visual view of the task flow chart to the user (for example, in which each task node and the relationship between the task nodes are identified by nodes and connecting lines).

[0079] Accordingly, the user can adjust the task flow chart by updating the visual view (for example, adjusting the number, position, specific content and relationship between task nodes in the visual view, etc.). Accordingly, the user can return the obtained updated visual view to the execution subject after updating. The execution subject replaces the visual view with the updated visual view in response to receiving the updated visual view for the task flow chart. And, when subsequently generating an action sequence based on the task flow chart, the task flow chart after the update determined based on the updated visual view is selected, and the action sequence is generated based on the updated task flow chart.

[0080] Furthermore, as discussed above, the execution subject can use the task flow chart to generate the action sequence. Figure 4 , Figure 4 A flowchart of a process of generating an action sequence provided by an embodiment of the present disclosure, wherein process 400 includes the following steps:

[0081] Step 401: determining a data processing action corresponding to a task node based on node data of the task node in the task flow chart and node data of a next task node adjacent to the task node;

[0082] Specifically, after obtaining and determining the task flow chart, the execution subject can compare the node data it has with the node data of the next task node adjacent to the task node for a task node, determine the data processing direction, and implement the data processing action of the processing direction (for example, node data a1 needs to be processed into node data a2). That is, the data processing action is used to process the node data of the task node into the node data of the next task node adjacent to the task node.

[0083] Step 402: Determine an agent that can complete the data processing action from the agent library;

[0084] Specifically, after determining the data action based on the above step 401, the execution subject can search for an agent that can complete the data processing action from the agent library. The agent library can be pre-configured, and the agent library includes one or more agents, and the agents can be used to implement different functions or data processing actions. For example, the agent library can include the storage location of the agent (for example, the location of the code file required to implement the agent). Accordingly, when the agent is determined to be called by the execution subject, the execution subject can obtain the agent based on the storage location and use it to complete the "call".

[0085] In some optional implementations of this embodiment, the agents in the agent library can be correspondingly added with action and function labels pointing to the data processing actions that can be implemented, so that the execution subject can use the action and function labels to determine the agent that can complete the data processing action.

[0086] Step 403: Use the task node as the action node in the action sequence, and use calling the action agent to complete the data processing action as the agent calling intention of the action node to generate the action sequence.

[0087] Specifically, after the intelligent agent is determined based on the above step 402, the executing entity uses the task node as the action node in the action sequence, and at each action node, determines the calling action for the intelligent agent (for example, calling the intelligent agent to process the current node data and output the node data of the next task node) as the intelligent agent calling intention discussed above, and generates an action sequence.

[0088] Therefore, the execution entity can use the pre-maintained intelligent agent library to automatically "input the scheduling behavior for the intelligent agent" based on the intelligent agent library to automatically configure and schedule the intelligent agent, thereby improving the use and scheduling efficiency of the intelligent agent.

[0089] Next, the source of the "update agent" will be discussed in detail.

[0090] In some optional implementations of this embodiment, in scenarios such as the above where an agent is provided based on an agent library, the executing entity may choose to search the agent library for whether such an updated agent exists (for example, the agent library has at least two different agents configured for the same function and the same data processing action).

[0091] Accordingly, when the execution subject queries whether there is an updated agent in response to the first confidence level of the agent processing result of the action node in the action sequence being less than the confidence level threshold, it may choose to query whether there is an updated agent in the agent library.

[0092] If there is an updated agent in the agent library, the execution subject can extract the agent from the agent library and call the updated agent.

[0093] In some optional implementations of this embodiment, if the agent does not exist in the agent library, the execution subject may respond to this by querying the second device based on the second communication path whether the second device can provide an updated agent.

[0094] It should be understood that, in an actual scenario, the second device is actually the same device as the first device, and / or the third device discussed later, and the present disclosure is not intended to be limited in this regard.

[0095] Accordingly, if the second device can provide the update agent, the execution subject can choose to control the second device to call the update agent and determine the update agent processing result corresponding to the action node. Then, the execution subject controls the second device to return the update agent processing result to use the second device to provide the update agent. In this way, the update agent can be provided by an "external device" independent of the execution subject, so as to enrich the source of the update agent and improve the processing capability of the execution subject.

[0096] In some embodiments, after the action sequence is fully executed, the execution subject may determine whether the task processing result corresponding to the action sequence meets the requirements. For example, in a text generation task, the completeness of the generated text meets the predetermined completeness requirement (for example, greater than or equal to a preset completeness threshold). For another example, the execution subject may provide the corresponding task processing result to the first device to request an indication of "whether the requirements are met".

[0097] Correspondingly, if the task processing result corresponding to the action sequence meets the requirements, the execution subject can respond to this and associate the agent used in the action sequence with the action node in the action sequence to generate action sequence storage information. Therefore, the subsequent processing process can be directly "reused" based on the sequence storage information to save the process of generating the action sequence, improve efficiency, and avoid repeated updates of the agents used.

[0098] In some optional implementations of this embodiment, after generating the action sequence storage information, the user can also adjust the action sequence storage information accordingly to meet different action sequence usage requirements. For example, the user can use the first device to communicate with the execution subject and send a node update instruction for the action sequence storage information.

[0099] Accordingly, the execution subject can respond to receiving the node update instruction for the action sequence storage information and update the agent call intention of the indicated action node based on the update agent call intention included in the node update instruction. For example, the execution subject can make personalized adjustments to the number of action nodes, the agent actually pointed to by the agent call intention, and other contents according to the user's instructions.

[0100] Based on any of the above embodiments, if there is no updated agent (for example, neither the agent library nor the second device can provide an updated agent), the execution entity can respond to this by requesting the third device to generate a corrected processing result of the action node based on the third communication path.

[0101] That is, the execution subject may also choose to communicate prompt information to the third device in the absence of an update agent to "notify" the absence of an update agent. Typically, the third device may be a terminal device used by a user with management and editing authority to execute the task.

[0102] Accordingly, a user with management and editing authority can use a third device to directly provide the execution subject with a revised processing result for updating the original agent processing result. Accordingly, the updated agent processing result actually provided by the user can be directly regarded by the execution subject as having a confidence level greater than or equal to the confidence threshold.

[0103] Therefore, in the absence of a newer agent, the executing entity can, through collaboration with users, replace the agent processing results with the corrected processing results, update the agent processing results that do not meet the requirements in the action nodes, and improve the execution quality.

[0104] Accordingly, in some embodiments, when generating an action sequence, if the "original agent" cannot be found for an action node, the action node can be similarly configured so that a third device provides "agent processing results" to assist the user in processing. This avoids the "action sequence" from being unable to be executed due to lack of an agent, thereby improving the stability of the action sequence.

[0105] In some embodiments, if the agent processing result corresponding to the action node cannot be updated (that is, there is no update agent, or the update agent cannot provide an update result that meets the requirements, and the third device cannot provide a corrected processing result), the execution entity may also choose to interrupt the task after the action node (for example, delete subsequent action nodes to adjust the action sequence) to avoid wasting computing resources.

[0106] To deepen understanding, the present disclosure also provides a specific implementation solution in combination with a specific application scenario, see Figure 5 Process 500 is shown.

[0107] For ease of understanding, the above system architecture 100 and a specific “task to be performed” can be used as an example. For example, the task to be performed can generate a slide outline based on a task description text.

[0108] For example, in process 500, a user (not shown) can use terminal device 103 to provide task description text 510 to server 105, so as to use server 105 to call an agent and execute the "task to be executed". Accordingly, server 105 can be exemplified as the above-mentioned "executing subject".

[0109] For example, the description text 510 may be specifically illustrated as follows: Please generate a slide outline, which needs to include two parts: Y1 theme and Y2 theme. In the outline, Y1 theme needs to include J sub-themes, and each sub-theme needs to include keywords of the subject content; Y2 theme needs to include K sub-themes, and each sub-theme needs to include keywords of the subject content, as well as reference materials and summaries of the reference materials for the provided keywords.

[0110] Accordingly, after acquiring the task description text 510 , the server 105 may execute S502 to generate a task flow chart 515 .

[0111] In the task flow chart 515, task nodes 521 to 528 may be included. For example, the node data of task node 521 may be the Y1 theme, the node data of task node 522 may be the J sub-themes under the Y1 theme, and the node data of task node 523 may be the keywords of the theme content that needs to be included in each sub-theme under the J sub-themes.

[0112] Similarly, exemplarily, the node data of task node 524 may be the Y2 theme, the node data of task node 525 may be the K sub-themes under the Y2 theme, the node data of task node 526 may be the keywords of the subject content that needs to be included in each sub-theme under the K sub-themes, the node data of task node 527 may be reference materials (for example, the access address of the reference materials) for the provided keywords, and the node data of task node 528 is a summary of the reference materials.

[0113] Next, the server 105 can split the task flowchart 515 into a sub-flowchart 531 and a sub-flowchart 533 based on the task dependency relationship in the task flowchart 515 and the two tasks without prior dependencies, namely, task node 521 and task node 524. For example, the sub-flowchart 531 can be understood as a part for generating a slide outline related to the Y1 theme, and the sub-flowchart 533 can be understood as a part for generating a slide outline related to the Y2 theme.

[0114] Then, the server 105 can at least use the agent library 540 to execute S504 to generate an action sequence 532 corresponding to the sub-flowchart 531 and an action sequence 534 corresponding to the sub-flowchart 533. The action sequence 532 includes action nodes 551 to 553, and the action sequence 534 includes action nodes 554 to 558.

[0115] Exemplarily, action node 551 may instruct the calling of an intelligent agent to generate J sub-topics under the Y1 topic; action node 552 may instruct the calling of an intelligent agent to generate keywords of the topic content that needs to be included in each of the J sub-topics; and action node 552 may instruct the output of the final execution result.

[0116] Similarly, illustratively, action node 554 can instruct the calling of an intelligent agent to generate K sub-topics under the Y2 topic; action node 555 can instruct the calling of an intelligent agent to generate keywords for the subject content that needs to be included in each of the K sub-topics; action node 555 can instruct the calling of an intelligent agent to find and provide reference materials for the keywords (for example, the access address of the reference materials); action node 557 can instruct the calling of an intelligent agent to generate a summary of the reference materials; and action node 558 can instruct the output of the final execution result.

[0117] Next, the server 105 may execute S505 to execute action sequences 532 and 534 in parallel.

[0118] Exemplarily, in the process of executing action sequences 532 and 534 in parallel, the first confidence of the agent processing results of action nodes 552 and 555 is less than the confidence threshold.

[0119] The server 105 may respond respectively and determine whether there is a corresponding update agent.

[0120] For example, for action node 552, no update agent that can be used can be found, and terminal device 103 can be a terminal device used by a user with management and editing authority. Accordingly, in this case, server 105 can choose to execute S506 to request terminal device 103 to provide correction processing result 563.

[0121] If the terminal device 103 returns the modified processing result 563 to the server 105 by executing S507 , the server 105 may choose to execute S508 to replace the agent processing result of the action node 552 with the modified processing result 563 .

[0122] After replacing the intelligent agent processing result of action node 552 with the modified processing result 563, the server 105 can continue to execute other action nodes (for example, action node 553) located after action node 552, and obtain the execution result 573 after finally completing all action nodes in the action sequence 532.

[0123] As for the action node 555 , in response to being able to determine the update agent 545 from the agent library 540 , the server 105 executes S509 , calls the update agent 545 , and determines the update agent processing result 565 of the action node 555 .

[0124] If the second confidence of the updated agent processing result 565 is greater than the second confidence threshold, the server 105 executes S510 to replace the agent processing result of the action node 555 with the updated agent processing result 565 .

[0125] And similarly, after replacing the agent processing result of action node 555 with the updated agent processing result 565, the server 105 can continue to execute other action nodes (for example, action nodes 556 to 558) located after action node 555, and after finally completing all action nodes in the action sequence 534, obtain the execution result 575.

[0126] After obtaining the execution result 573 and the execution result 575 , the server 105 may execute S511 to provide the execution results 573 and 575 (eg, a slide outline) to the terminal device 103 .

[0127] It should be understood that the method of sending the execution results 573 and 575 at the same time is only an exemplary description. In practice, based on different scenarios and requirements, the server 105 can also directly send the obtained execution result to the terminal device 103 after obtaining the execution result 573 or 575, without waiting for the other to be executed. Accordingly, after the other is executed, it is separately provided to the terminal device 103. In this way, the user can obtain part of the execution results "earlier".

[0128] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an action sequence execution device based on an intelligent agent, and the device embodiment is Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0129] like Figure 6As shown, the agent-based action sequence execution device 600 of this embodiment may include: an update agent query unit 601, an update result determination unit 602, and an update result replacement unit 603. The update agent query unit 601 is configured to query whether there is an update agent in response to the first confidence of the agent processing result of the action node in the action sequence being less than the confidence threshold, wherein the update agent can be used to implement the agent call intention indicated by the action node; the update result determination unit 602 is configured to call the update agent in response to the existence of the update agent, and determine the update agent processing result corresponding to the action node; the update result replacement unit 603 is configured to replace the agent processing result with the update agent processing result in response to the second confidence of the update agent processing result being greater than or equal to the confidence threshold.

[0130] In this embodiment, in the agent-based action sequence execution device 600, the specific processing of the update agent query unit 601, the update result determination unit 602 and the update result replacement unit 603 and the technical effects thereof can be referred to in Figure 2 The relevant descriptions of steps 201 - 203 in the corresponding embodiment are not repeated here.

[0131] In some optional implementations of the present embodiment, the device 600 also includes: a sub-flowchart determination unit, configured to determine a sub-flowchart based on the task dependencies in the task flow chart of the task to be executed, wherein the tasks included in the sub-flowchart belong to the same task chain, and the task at the first position in the sub-flowchart is a task without prior dependencies; a first action sequence generation unit, configured to generate an action sequence corresponding to the sub-flowchart; and an action sequence execution unit, configured to execute at least two action sequences concurrently in response to the existence of at least two action sequences.

[0132] In some optional implementations of the present embodiment, the device 600 also includes: a utilization order determination unit, which is configured to determine the utilization order of the conflicting action sequences for the target intelligent agent in response to the existence of at least two conflicting action sequences, using the length descending order arrangement results based on the estimated execution time of the conflicting action sequences, wherein at least two conflicting action sequences simultaneously rely on the same target intelligent agent to complete the action node located first, and the utilization order corresponds to the arrangement order indicated by the length descending order arrangement results.

[0133] In some optional implementations of this embodiment, the device 600 also includes: a description text acquisition unit, configured to acquire a task description text of the task to be executed; a task flowchart generation unit, configured to generate a task flowchart of the task to be executed based on the task entity recognition result of the task description text; and a second action sequence generation unit, configured to generate an action sequence based on the task flowchart.

[0134] In some optional implementations of this embodiment, the device 600 also includes: a connection task determination unit, configured to determine a third task for connecting the first task and the second task in response to the task correlation between the first task and the second task in the task flow chart being less than a correlation threshold; a task flow chart updating unit, configured to update the third task to the task flow chart to obtain an updated task flow chart; and a second action sequence generating unit, further configured to generate an action sequence based on the updated task flow chart.

[0135] In some optional implementations of this embodiment, the apparatus 600 further includes: a task flow chart presentation unit, configured to present a visual view of the task flow chart using a first device based on a first communication path; a visual view replacement unit, configured to replace the visual view with the updated visual view in response to receiving an updated visual view for the task flow chart; and a second action sequence generation unit, further configured to determine an updated task flow chart based on the updated visual view, and generate an action sequence based on the updated task flow chart.

[0136] In some optional implementations of the present embodiment, the second action sequence generation unit includes: a data processing action determination subunit, configured to determine the data processing action corresponding to the task node based on the node data of the task node in the task flow chart and the node data of the next task node adjacent to the task node, wherein the data processing action is used to process the node data of the task node into the node data of the next task node adjacent to the task node; an agent determination subunit, configured to determine an agent that can complete the data processing action from the agent library; an action sequence generation subunit, configured to use the task node as the action node in the action sequence, and generate an action sequence with the agent calling intention of calling the action agent to complete the data processing action as the action node.

[0137] In some optional implementations of the present embodiment, the update agent query unit 601 includes: an agent library query sub-unit, configured to query whether there is an update agent in the agent library in response to the first confidence of the agent processing result of the action node in the action sequence being less than the confidence threshold; an update agent request sub-unit, configured to query the second device whether the update agent can be provided by the second device based on the second communication path in response to the absence of the update agent in the agent library; and an update result determination unit 602, including: an update agent call control sub-unit, configured to control the second device to call the update agent in response to the second device being able to provide the update agent, and determine the update agent processing result corresponding to the action node; an update result return control sub-unit, configured to control the second device to return the update agent processing result.

[0138] In some optional implementations of this embodiment, the device 600 also includes: an action sequence storage information generation unit, which is configured to associate the intelligent agent used in the action sequence with the action node in the action sequence in response to the task processing result corresponding to the action sequence meeting the requirements, and generate action sequence storage information.

[0139] In some optional implementations of this embodiment, the device 600 also includes: a call intention update unit, which is configured to update the agent call intention of the indicated action node based on the update agent call intention included in the node update instruction in response to receiving a node update instruction for storing information on the action sequence.

[0140] In some optional implementations of the present embodiment, the apparatus 600 further includes: a correction processing result requesting unit, configured to, in response to the absence of an updated intelligent agent, request a third device, based on a third communication path, to generate a correction processing result of an action node from the third device; and a correction result replacement unit, configured to replace the intelligent agent processing result with the correction processing result.

[0141] This embodiment exists as an apparatus embodiment corresponding to the above-mentioned method embodiment. The agent-based action sequence execution device provided in this embodiment can monitor the calling and use results of the agent in the process of calling and using the agent based on the action sequence. Then, when it is monitored that the agent calling and use results of the action node do not meet the requirements, other updated agents are used to reprocess the agent calling intention indicated by the action node to achieve timely update of the results that do not meet the requirements. In this way, it is possible to avoid "errors" and "results that do not meet the requirements" from erroneously flowing into subsequent action nodes, while reducing the waste of computing resources, ensuring and improving the quality of the execution results obtained by executing the agent-based action sequence.

[0142] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0143] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0144] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0145] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0146] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as an agent-based action sequence execution method. For example, in some embodiments, an agent-based action sequence execution method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the agent-based action sequence execution method described above may be executed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the agent-based action sequence execution method in any other appropriate manner (eg, by means of firmware).

[0147] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0148] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0149] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, 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 foregoing.

[0150] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0151] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0152] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services. The server can also be divided into a server of a distributed system, or a server combined with a blockchain.

[0153] According to the technical solution of the embodiment of the present disclosure, the calling and using results of the agent can be monitored in the process of calling and using the agent based on the action sequence. Then, when it is detected that the calling and using results of the agent at the action node do not meet the requirements, other updated agents are used to reprocess the agent calling intention indicated by the action node to achieve timely update of the results that do not meet the requirements. In this way, it is possible to avoid "errors" and "results that do not meet the requirements" from erroneously flowing into subsequent action nodes, while reducing the waste of computing resources, ensuring and improving the quality of the execution results obtained by executing the action sequence based on the agent.

[0154] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions provided by this disclosure can be achieved, and this document does not limit this.

[0155] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An agent-based action sequence execution method, comprising: In response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence level threshold, querying whether there is an update agent, wherein the update agent can be used to implement the agent calling intention indicated by the action node; In response to the existence of the update agent, calling the update agent and determining the update agent processing result corresponding to the action node; In response to the second confidence of the updated agent processing result being greater than or equal to the confidence threshold, the updated agent processing result is used to replace the agent processing result.

2. The method according to claim 1, further comprising: Determine a sub-flowchart based on the task dependency relationship in the task flow chart of the task to be executed, wherein the tasks included in the sub-flowchart belong to the same task chain, and the task at the first position in the sub-flowchart is a task without a prior dependency; Generate an action sequence corresponding to the sub-flowchart; In response to the existence of at least two of the action sequences, the at least two action sequences are executed concurrently.

3. The method according to claim 2, in response to the existence of at least two conflicting action sequences, at least two conflicting action sequences simultaneously relying on the same target agent to complete the first action node, further comprising: Based on the descending order arrangement result of the estimated execution time of the conflicting action sequence, the utilization order of the conflicting action sequence for the target intelligent agent is determined, wherein the utilization order corresponds to the arrangement order indicated by the descending order arrangement result.

4. The method according to claim 1, further comprising: Get the task description text of the task to be executed; Based on the task entity recognition result of the task description text, generate a task flow chart of the task to be performed; The action sequence is generated based on the task flow chart.

5. The method according to claim 4, further comprising: In response to a task correlation between a first task and a second task in the task flow chart being less than a correlation threshold, determining a third task for connecting the first task and the second task; Updating the third task to the task flow chart to obtain an updated task flow chart; as well as The step of generating the action sequence based on the task flow chart includes: The action sequence is generated based on the updated task flow chart.

6. The method according to claim 4, further comprising: Based on the first communication path, using the first device to present a visual view of the task flow chart; In response to receiving an updated visualization view for the task flow diagram, replacing the visualization view with the updated visualization view; as well as The step of generating the action sequence based on the task flow chart includes: An updated task flow chart is determined based on the updated visualization view, and the action sequence is generated based on the updated task flow chart.

7. The method according to claim 4, wherein: The step of generating the action sequence based on the task flow chart includes: Based on the node data of the task node in the task flow chart and the node data of the next task node adjacent to the task node, determine the data processing action corresponding to the task node, wherein the data processing action is used to process the node data of the task node into the node data of the next task node adjacent to the task node; Determining an agent capable of completing the data processing action from an agent library; The task node is used as the action node in the action sequence, and the action sequence is generated by taking calling the action agent to complete the data processing action as the agent calling intention of the action node.

8. The method according to claim 7, wherein: In response to the first confidence level of the agent processing result of the action node in the action sequence being less than the confidence level threshold, querying whether there is an update agent comprises: In response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence level threshold, querying whether the update agent exists in the agent library; In response to the update agent not existing in the agent library, querying the second device based on the second communication path whether the update agent can be provided by the second device; and In response to the existence of the update agent, calling the update agent and determining the update agent processing result corresponding to the action node include: In response to the second device being able to provide the update agent, controlling the second device to call the update agent and determining the update agent processing result corresponding to the action node; Control the second device to return the updated agent processing result.

9. The method according to claim 1, further comprising: In response to the task processing result corresponding to the action sequence meeting the requirement, the intelligent agent used in the action sequence is associated with the action node in the action sequence to generate action sequence storage information.

10. The method according to claim 9, further comprising: In response to receiving a node update instruction for storing information of the action sequence, the agent call intent of the indicated action node is updated based on the update agent call intent included in the node update instruction.

11. The method according to any one of claims 1 to 10, further comprising: In response to the absence of the update agent, requesting a third device to generate a revised processing result of the action node based on a third communication path; The agent processing result is replaced by the modified processing result.

12. An agent-based action sequence execution device, comprising: An update agent query unit is configured to query whether there is an update agent in response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence level threshold, wherein the update agent can be used to implement the agent calling intention indicated by the action node; an update result determination unit, configured to call the update agent in response to the existence of the update agent, and determine the update agent processing result corresponding to the action node; The update result replacement unit is configured to replace the agent processing result with the update agent processing result in response to the second confidence of the update agent processing result being greater than or equal to the confidence threshold.

13. The apparatus according to claim 12, further comprising: A sub-flowchart determining unit is configured to determine a sub-flowchart based on the task dependency relationship in the task flow chart of the task to be executed, wherein the tasks included in the sub-flowchart belong to the same task chain, and the task at the first position in the sub-flowchart is a task without a prior dependency; A first action sequence generating unit is configured to generate an action sequence corresponding to the sub-flowchart; The action sequence execution unit is configured to execute at least two action sequences concurrently in response to the existence of at least two action sequences.

14. The apparatus according to claim 13, further comprising: The utilization order determination unit is configured to determine the utilization order of the conflicting action sequences for the target intelligent agent in response to the existence of at least two conflicting action sequences, using the length descending order arrangement results based on the estimated execution time of the conflicting action sequences, wherein at least two conflicting action sequences simultaneously rely on the same target intelligent agent to complete the action node located at the first position, and the utilization order corresponds to the arrangement order indicated by the length descending order arrangement results.

15. The apparatus according to claim 12, further comprising: A description text acquisition unit, configured to acquire a task description text of the task to be executed; A task flow chart generating unit is configured to generate a task flow chart of the to-be-performed task based on the task entity recognition result of the task description text; The second action sequence generating unit is configured to generate the action sequence based on the task flow chart.

16. The apparatus according to claim 15, further comprising: a connecting task determining unit, configured to determine a third task for connecting the first task and the second task in response to a task correlation between the first task and the second task in the task flow chart being less than a correlation threshold; a task flow chart updating unit, configured to update the third task to the task flow chart to obtain an updated task flow chart; as well as The second action sequence generating unit is further configured to generate the action sequence based on the updated task flow chart.

17. The apparatus according to claim 15, further comprising: a task flow chart presenting unit, configured to present a visual view of the task flow chart using the first device based on a first communication path; a visualization view replacement unit, configured to, in response to receiving an updated visualization view for the task flow chart, replace the visualization view with the updated visualization view; as well as The second action sequence generating unit is further configured to determine an updated task flow chart based on the updated visualization view, and generate the action sequence based on the updated task flow chart.

18. The device according to claim 15, wherein: The second action sequence generating unit includes: A data processing action determination subunit is configured to determine a data processing action corresponding to the task node based on the node data of the task node in the task flow chart and the node data of the next task node adjacent to the task node, wherein the data processing action is used to process the node data of the task node into the node data of the next task node adjacent to the task node; An agent determination subunit, configured to determine an agent capable of completing the data processing action from an agent library; The action sequence generation subunit is configured to use the task node as the action node in the action sequence, and to generate an action sequence with the agent calling intention of calling the action agent to complete the data processing action as the action node.

19. The device according to claim 18, wherein: The updating agent query unit comprises: an agent library query subunit, configured to query whether the updated agent exists in the agent library in response to a first confidence level of an agent processing result of an action node in an action sequence being less than a confidence level threshold; an update agent request subunit, configured to query a second device whether the update agent can be provided by the second device based on a second communication path in response to the update agent not existing in the agent library; and The update result determination unit comprises: an update agent calling control subunit, configured to control the second device to call the update agent in response to the second device being able to provide the update agent, and determine the update agent processing result corresponding to the action node; The update result return control subunit is configured to control the second device to return the update agent processing result.

20. The apparatus according to claim 12, further comprising: The action sequence storage information generating unit is configured to associate the intelligent agent used in the action sequence with the action node in the action sequence in response to the task processing result corresponding to the action sequence meeting the requirement, and generate the action sequence storage information.

21. The apparatus according to claim 20, further comprising: The calling intention updating unit is configured to update the agent calling intention of the indicated action node in response to receiving a node updating instruction storing information for the action sequence, based on the updating agent calling intention included in the node updating instruction.

22. The apparatus according to any one of claims 12 to 21, further comprising: a correction processing result requesting unit configured to, in response to the absence of the update agent, request the third device to generate a correction processing result of the action node based on the third communication path from the third device; The correction result replacement unit is configured to replace the agent processing result with the correction processing result.

23. An electronic device, comprising: at least one processor; as well as 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 to enable the at least one processor to execute the agent-based action sequence execution method described in any one of claims 1-11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the agent-based action sequence execution method according to any one of claims 1-11.

25. A computer program product, comprising a computer program, which, when executed by a processor, implements the agent-based action sequence execution method according to any one of claims 1-11.

Images