Application publishing method, system and computer storage medium

By determining the functional requirements of the target application, selecting and configuring the target intelligent agent and tool body, and encapsulating them into the target application, the problems of flexibility and customizability in application design in the prior art are solved, realizing flexible design and customizability of the application, and improving development efficiency and scalability.

CN118567712BActive Publication Date: 2024-11-08LENOVO NEW VISION BEIJING TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411046836.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-08
Estimated Expiration
2044-08-01

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve flexible and customizable application design, and cannot define and execute sets of tasks and activities based on user needs and scenarios.

Method used

By determining the functional requirements of the target application, selecting and configuring the target intelligent agent and tool body, and encapsulating them into the target application, the application can be designed flexibly and customizablely.

Benefits of technology

It enables flexible and customizable design of the target application, meets users' personalized needs, and improves the efficiency and scalability of application development.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118567712B_ABST
    Figure CN118567712B_ABST
Patent Text Reader

Abstract

The present application provides an application publishing method, system and computer storage medium, wherein the publishing method includes determining a target application in response to an application creation request; determining initial parameters of the target application based on the functional requirements of the determined target application, and determining a target intelligent body and a target tool body, wherein the target intelligent body is configured to be able to realize the functions required by the target application alone or in collaboration, and the target tool body is configured to be able to assist the operation of the target intelligent body alone or in collaboration; configuring the parameters of the determined target intelligent body and / or the target tool body based on the initial parameters of the target application to obtain a first intelligent body and a first tool body adapted to the target application; encapsulating the first intelligent body and the first tool body as a target application to perform a publishing operation on the target application. Flexible design and customizable design of the target application are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of application technology, and in particular to an application publishing method, system and computer storage medium. Background Art

[0002] An autonomous, interactive and adaptive artificial intelligence entity or degree can perceive the environment and make decisions to complete specific tasks. A series of interrelated tasks and activities can achieve specific target functions by executing them in a certain order and rules. For the design of an application, it may require a more flexible or customizable design, so that users can define and execute the above-mentioned interrelated tasks and activities according to their own needs and scenarios. How to achieve a more flexible or customizable application design is a problem that needs to be solved. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide an application publishing method, system and computer storage medium.

[0004] The embodiment of the present application adopts the following technical solution: an application publishing method, comprising:

[0005] In response to the application creation request, determining a target application;

[0006] Based on the determined functional requirements of the target application, initial parameters of the target application are determined, and a target intelligent agent and a target tool agent are determined, wherein the target intelligent agent is configured to be able to realize the functions required by the target application alone or in collaboration, and the target tool agent is configured to be able to assist the operation of the target intelligent agent alone or in collaboration;

[0007] Based on the initial parameters of the target application, the parameters of the target intelligent agent and / or the target tool body are configured to obtain a first intelligent agent and a first tool body adapted to the target application;

[0008] The first intelligent body and the first tool body are encapsulated as a target application to perform a publishing operation on the target application.

[0009] In some embodiments, determining the target application in response to the application creation request includes:

[0010] In response to the application creation request, generating a first application, determining the first application as the target application, or,

[0011] A second application is selected from the application template, and the second application is determined as the target application.

[0012] In some embodiments, based on the determined functional requirements of the target application, determining a target intelligent agent and a target tool agent includes:

[0013] Dividing the functional requirements of the target application into multiple tasks or operations;

[0014] Based on the determined task requirements or operation requirements of the target application, at least one target agent capable of performing the task or the operation is retrieved and selected from an agent library, wherein the target agent is configured to be capable of realizing the functions of one or more of the tasks or the operation individually or collaboratively, and at least one target tool body capable of assisting the operation of the target agent is retrieved and selected from a tool library;

[0015] Based on the compatibility requirements of the target application, at least one target intelligent entity that meets the compatibility requirements of the target application is determined, and at least one target tool entity that meets the compatibility requirements of the target application is determined.

[0016] In some embodiments, the parameters of the target intelligent agent and / or the target tool agent that are determined are configured to include at least one of the following configurations:

[0017] Setting parameters of the target intelligent entity and / or the target tool entity;

[0018] Selecting a model that is compatible with the target intelligent entity and / or the target tool entity;

[0019] Adjust the behavior pattern or strategy of the target agent.

[0020] In some embodiments, before determining the target agent, the method further includes:

[0021] Based on the requirements of the composite agent workflow, the steps of the composite agent workflow are set to obtain a plurality of second agents required to complete the composite agent workflow;

[0022] Configuring the obtained parameters of the plurality of second agents to obtain a plurality of third agents adapted to complete the composite agent workflow;

[0023] Determining the relationship between the plurality of said third agents;

[0024] Based on the steps of the composite agent workflow and the relationship between the plurality of third agents, execute the composite agent workflow and output the operation result;

[0025] The plurality of third agents corresponding to the operation results that meet the expected result requirements are encapsulated to generate a composite agent.

[0026] In some embodiments, the encapsulating of the plurality of third agents corresponding to the operation results satisfying the expected results to generate a composite agent comprises:

[0027] Encapsulate the plurality of third agents corresponding to the operation results that meet the expected results, generate a composite agent in the first format, and save it to the agent library; or,

[0028] The plurality of third agents corresponding to the operation results that meet the expected results are encapsulated to generate a composite agent having a second format, and the composite agent is saved to the tool library.

[0029] In some embodiments, obtaining a plurality of second agents required to complete the agent workflow comprises:

[0030] Based on the needs of the individual intelligent agent to be generated, determine the basic intelligent agent, the tool agent and the model, wherein the basic intelligent agent, the tool agent and the model are adapted to the needs of the individual intelligent agent to be generated;

[0031] Adding or modifying parameters for the basic agent, and / or modifying parameters of the model, to obtain a fourth agent;

[0032] Based on the application scenario of the personalized agent to be generated, selecting a prompt template to guide the operation of the fourth agent;

[0033] Based on the needs of the personalized intelligent agent to be generated, determining a knowledge base that can provide adaptation information for the personalized intelligent agent to be generated;

[0034] In case the operating result of the fourth agent meets the expected result, the fourth agent is saved as the individual agent.

[0035] In some embodiments, performing a publishing operation on the target application includes:

[0036] The packaged target application, configuration file and user manual are published.

[0037] The present application also provides an application publishing system, including:

[0038] A first determination module configured to determine a target application in response to an application creation request;

[0039] A second determination module, configured to determine initial parameters of the target application, and determine a target intelligent agent and a target tool agent based on the determined functional requirements of the target application;

[0040] A configuration module configured to configure the determined parameters of the target intelligent agent and / or the target tool body based on the initial parameters of the target application to obtain a first intelligent agent and a first tool body adapted to the target application;

[0041] The packaging and publishing module is configured to package the first intelligent body and the first tool body into a target application so as to perform a publishing operation on the target application.

[0042] An embodiment of the present application also provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0043] The beneficial effects of the embodiments of the present application are:

[0044] According to the requirements of application creation, the target application is determined, and then according to the requirements of the target application, the corresponding target intelligent body and target tool body are determined, and the first intelligent body and the first target tool body obtained after the parameter configuration can be adapted to the target application by configuring the parameters of the determined target intelligent body and the target tool body. Then the first intelligent body and the first target tool body are encapsulated as a target application and released, thereby realizing the flexible design and customizable design of the target application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0046] Figure 1 A flowchart of the publishing method for this application application;

[0047] Figure 2 For this application Figure 1 A flowchart of one embodiment of step S20;

[0048] Figure 3 For this application Figure 2 A flowchart of one embodiment of step S22;

[0049] Figure 4 A flowchart for creating a composite agent for this application;

[0050] Figure 5 For this application Figure 3 A flowchart of one embodiment of step S221;

[0051] Figure 6 Flowchart when generating a personality agent for this application;

[0052] Figure 7 Another flow chart of the publishing method for the present application;

[0053] Figure 8 This is a structural block diagram of the application publishing system for this application. DETAILED DESCRIPTION

[0054] Various aspects and features of the present application are described herein with reference to the accompanying drawings.

[0055] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but only as an example of an embodiment. Other modifications within the scope and spirit of the present application will occur to those skilled in the art.

[0056] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0057] These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0058] It should also be understood that although the present application has been described with reference to some specific examples, those skilled in the art will be able to readily implement many other equivalent forms of the present application.

[0059] The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.

[0060] Specific embodiments of the present application are described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments applied for are merely examples of the present application, which may be implemented in a variety of ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that obscure the present application. Therefore, the specific structural and functional details applied for herein are not intended to be limiting, but merely serve as a basis and representative basis for the claims to teach those skilled in the art to use the present application in a variety of ways with substantially any suitable detailed structure.

[0061] This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," all of which may refer to one or more of the same or different embodiments according to the present application.

[0062] The embodiments of the present application first introduce and illustrate the basic concepts that may be involved in the present application. The basic library includes: a model library, a tool library, a prompt template library, and a knowledge base.

[0063] Algorithm library:

[0064] The model library is a collection or system that centrally stores and manages various algorithms (such as machine learning models, simulation models, image processing models, etc.).

[0065] LLM (Large Language Models) library:

[0066] LLM libraries refer to a collection of large language models that can be used in various applications. These libraries are designed to provide developers with access to pre-trained LLMs that can be integrated into tools or systems to perform natural language processing tasks such as text generation, translation, summarization, etc. LLM libraries include local LLM libraries and online LLM libraries.

[0067] Local LLM Libraries: Local LLM libraries consist of LLMs that are stored and run on a local computer or server. This allows developers to deploy and use LLMs directly on their own systems without requiring an internet connection to access remote services. Local LLMs facilitate privacy, control, and reduce potential latency because the models are processed directly on the user's hardware.

[0068] Online LLM Libraries: Online LLM libraries are collections of LLMs hosted on remote servers and accessed over the internet. Users interact with these models through APIs or web services. The advantage of online LLM libraries is that they can provide more powerful computing resources and do not have to manage the infrastructure required to run large models. In addition, online libraries can provide access to the latest models and updates without the need for local installation.

[0069] Tool Library:

[0070] The tool library can be divided into service tools, library tools, and script tools based on tool types.

[0071] Service tools: Service tools are software tools specifically used to interact and manage external services. Such tools usually allow users to call services by configuring specific hosts and API addresses to perform tasks such as data acquisition, function calls, and status monitoring.

[0072] Library tools: Library tools usually refer to library files that are mainly implemented in Python. They provide a set of predefined functions, classes, interfaces, etc. to support the development of specific functions or services.

[0073] Scripting tools: Scripting tools are programs or tools written in a scripting language that are used to automate a series of tasks or operations. Scripting tools can handle a variety of complex tasks, such as license agreement management, network speed measurement, etc.

[0074] The tool library can be divided into pre-processing tools, post-processing tools, functional tools and intermediate tools according to their functions.

[0075] Pre-processing tools: Pre-processing tools are a series of tools used to pre-process input data before the agent receives external input. These tools can perform operations such as data cleaning, format conversion, feature extraction, standardization, and normalization to ensure that the input data meets the requirements of the agent's internal processing flow. For example: speech-to-text, text cleaning, and image pre-processing.

[0076] Post-processing tools: Post-processing tools are a series of tools used to post-process the internal processing results before the agent generates output. These tools can optimize, transform or interpret the output results of the agent to meet user needs or adapt to specific application scenarios. For example: text-to-speech, sentiment analysis, and visualization.

[0077] Functional tools: Functional tools are a collection of tools used to achieve specific functions in the internal processing flow of the agent. These tools are usually designed according to the business needs or task requirements of the agent and are used to complete tasks such as data processing, decision making, action planning, etc. within the agent. For example: natural language understanding, knowledge base query, recommendation algorithm

[0078] Centralized tools: Centralized tools are a series of tools that provide the required environment for the engine running the intelligent agent. These tools usually include runtime engines, simulation engines, middleware, etc., which are used to support the operation, testing, deployment and expansion of intelligent agents.

[0079] Prompt Template Library: It is a resource library that contains prompt templates required by various large language models. These prompt templates are designed to guide LLMs to generate specific types of responses or perform specific tasks. The prompt template library is crucial to improving the performance of LLMs and adapting to different application scenarios.

[0080] Knowledge base: It is a complex information collection consisting of files, relational databases, graph databases, and vector databases. It is generated with the assistance of LLM and is used to store, manage, and retrieve knowledge and data in professional fields. It provides professional field context to LLM, enabling LLM to handle professional field problems.

[0081] Agent Library:

[0082] The agent library is divided into basic agents, personalized agents, and composite agents.

[0083] Basic agent: The basic agent is a single agent that introduces new mechanisms and capabilities to the existing LLM. It is an agent with intelligent characteristics and functions. It integrates advanced algorithms and processes to cope with various complex tasks more efficiently and intelligently.

[0084] Personalized agent: Personalized agent is an agent with unique personality and functions, which is the inheritance and extension of basic agent or compound agent. It provides personalized services for specific users or specific scenarios by binding specific tools, models, parameters, knowledge bases and templates.

[0085] Composite Agent: Composite Agent is an encapsulation of agent workflow, combining multiple individual agents or composite agents into a whole to complete more complex tasks. It can be encapsulated as a service or used directly in chatBot as a module. Encapsulated as a logical single agent

[0086] When creating an intelligent complex, an agent workflow is required: a method of combining multiple individual agents or composite agents according to a specific process to complete complex tasks. This method ensures that the task can be completed efficiently and accurately by defining the various stages and steps of the task and the individual agents or composite agents required in each stage and step.

[0087] The application library is a collection of applications that centrally stores and manages different types of applications. In our solution, a specific application can be generated by binding and setting the parameters of tools and agents and configuring related runtime files.

[0088] After introducing the relevant concepts involved in this application, in order to solve the problems in the background technology, the embodiment of this application provides an application publishing method. The application can be a web application installed on a computer terminal or an APP or other application on other smart devices such as a mobile phone.

[0089] Combination Figure 1 , the application publishing method includes the following steps:

[0090] S10, in response to the application creation request, determining a target application.

[0091] For example, the user can send an application creation request as needed, and in response to the application creation request, the user's functional requirements for the application are clarified to determine the target application. For example, if the user wants to create an animation APP (application program) that outputs an animation based on the input image, the target application can be an animation production APP.

[0092] S20, based on the determined functional requirements of the target application, determine the initial parameters of the target application, and determine the target intelligent agent and the target tool body, wherein the target intelligent agent is configured to be able to realize the functions required by the target application alone or in collaboration, and the target tool body is configured to be able to assist the operation of the target intelligent agent alone or in collaboration.

[0093] Exemplarily, in combination with step S10, based on the determined functional requirements of the target application, the initial parameters of the target application are determined, where the initial parameters may include, but are not limited to, the name, version, and operating environment of the target application. For example, in combination with the above embodiment, the initial parameters of the target application include: animation production, version 1.0, IOS system or Android system.

[0094] In addition, according to the functional requirements of the target application, a suitable tool is determined, which can be selected from the tool library. It should be noted that the tool library here can include both agents and tool bodies. That is, the agent can also be used as a tool body. Of course, suitable agents and tool bodies can also be selected from the agent library and the tool library respectively. Specifically, according to the functional requirements of the target application, a suitable target agent is determined, where the target agent can be one or more. These target agents will be responsible for executing key functions in the target application, etc. In the case where there is one target agent, the target agent can realize the functions required by the target application alone; in the case where there are multiple target agents, multiple target agents can cooperate to realize the functions required by the target application. Continuing with the above embodiment, according to the functional requirements of the animation production APP, a suitable target agent can be selected from the agent library, and the target agent can have both the function of image processing and the function of converting images into animations. Two target agents can also be selected, one of which has the function of image processing and the other has the function of converting processed images into animations.

[0095] Similarly, the appropriate target tool body can be determined according to the functional requirements of the target application. There are various types of target tool bodies, such as service tools, library tools, script tools, etc. The target tool body can also be one or multiple. In the case of one target tool body, it can independently assist the operation of the target intelligent body. For example, the target tool body has functions such as cropping, scaling, rotating, and adjusting image color, which can assist the target intelligent body of image processing to complete the processing of the input image. In the case of multiple target tool bodies, multiple target tool bodies can cooperate to assist the operation of the target intelligent body. For example, one of the target tool bodies has the function of cropping, scaling, and rotating the image, and the other target tool body can adjust the color of the image to increase the ability of the target intelligent body and help the target intelligent body better perform and complete its functions. The two target tool body systems cooperate with the target intelligent body of auxiliary image processing to perform image processing.

[0096] S30, based on the initial parameters of the target application, configure the determined parameters of the target intelligent agent and / or the target tool body to obtain a first intelligent agent and a first tool body adapted to the target application.

[0097] Exemplarily, according to the initial parameters of the target application, the parameters of the target agent and / or the target tool body are configured and edited. The configuration and editing of the parameters of the target agent and / or the target tool body may include, but is not limited to, setting corresponding parameters, selecting an adapted model, adjusting behavior, etc., to ensure that they can work as expected. For example, if the target agent needs to be modified, the target agent is entered to modify it.

[0098] After modification, a first agent adapted to the target application is obtained, and the first agent may include one or more agents. After modification, a first tool body adapted to the target application is obtained, and the first tool body may include one or more agents.

[0099] S40, encapsulating the first intelligent body and the first tool body into a target application to perform a publishing operation on the target application.

[0100] Exemplarily, the first intelligent agent and the first tool agent are encapsulated to obtain an independent target application, for example, the target application is an animation production APP. The target application is published, for example, according to user needs, the target application is published to one or more corresponding platforms.

[0101] The embodiment of the present application can determine the target application according to the requirements of application creation, and then determine the corresponding target intelligent body and target tool body according to the requirements of the target application, and configure the parameters of the determined target intelligent body and target tool body so that the first intelligent body and the first target tool body obtained after the parameter configuration can adapt to the target application. Then the first intelligent body and the first target tool body are encapsulated as a target application and released, thereby realizing the flexible design and customizable design of the target application.

[0102] In some embodiments, determining the target application in response to the application creation request includes:

[0103] In response to the application creation request, generating a first application, determining the first application as the target application, or,

[0104] A second application is selected from the application template, and the second application is determined as the target application.

[0105] For example, after receiving a user's application creation request, a first application can be newly created according to the specific situation, and the first application can be determined as the target application. Alternatively, an application template can be selected from the application template library, and the application template can be a second application, and the second application can be determined as the target application. In specific operations, you can first select from the application template library. If there is no second application that meets the requirements, you can determine the first application as the target application by creating a new first application. Then set the initial parameters for the target application.

[0106] In some embodiments, in combination Figure 2 , based on the determined functional requirements of the target application, determining a target intelligent agent and a target tool agent, including:

[0107] S21, dividing the functional requirements of the target application into multiple tasks or operations.

[0108] For example, the functions that the target application wants to implement may have several implementation stages when being implemented. According to the implementation stages, the functional requirements of the target application can be divided into multiple tasks or operations corresponding to the implementation stages through the functional analysis module.

[0109] Continuing with the above embodiment, if the animation production APP wants to realize the conversion of input images into animation output, it needs to go through the first stage, that is, the processing stage of the input images; then the second stage, that is, converting the processed images into animations; and then the third stage, that is, the animation output stage. The functional requirements of the animation production APP can be divided into three tasks. The first task is to process the input images, the second task is to convert the processed images into animations, and the third task is to output the animations.

[0110] S22, based on the determined task requirements or operation requirements of the target application, retrieve and select from the agent library at least one target agent capable of performing the task or the operation, wherein the target agent is configured to be able to implement the functions of one or more of the tasks or the operations individually or collaboratively, and retrieve and select from the tool library at least one target tool body capable of assisting the operation of the target agent.

[0111] Exemplarily, a plurality of agents are stored in the agent library, and each agent has the ability to perform a specific task or operation. A retrieval and selection module can be connected to the functional requirement analysis module and the agent library to retrieve and select a target agent that can realize the functional requirements of the target application from the agent library. After the functional requirements of the target application are divided into a plurality of tasks or operations, at least one target agent that can perform the task or operation is retrieved and selected from the agent library based on the divided plurality of tasks or operations. The target agent is configured to realize the functions of one or more of the tasks or operations individually or collaboratively, and at least one target tool body that can assist the operation of the target agent is retrieved and selected from the tool library.

[0112] For example, in combination with the above embodiment, the search keyword can be set as "image processing" when searching from the agent library, and then at least one target agent that can perform the first task can be selected based on the search results; then the search can be performed by the search word "animation", and at least one target agent that can perform the second task can be selected based on the search results; and the search can be performed by the search word "output", and at least one target agent that can perform the third task can be selected based on the search results. Similarly, at least one target tool body that can assist the operation of the target agent can be retrieved and selected from the tool library by inputting the corresponding search word.

[0113] S23, based on the compatibility requirements of the target application, determine at least one target intelligent entity that meets the compatibility requirements of the target application, and determine at least one target tool entity that meets the compatibility requirements of the target application.

[0114] Exemplarily, the corresponding search results in step S23 may be pre-screened to exclude agents that do not meet the basic compatibility requirements of the target application. Of course, it is understandable that the target agent may also be determined from the retrieved agents based on other factors such as execution efficiency and resource consumption.

[0115] The intelligent agents in the intelligent agent library and the tool bodies in the tool library have all been created. Users can directly drag the intelligent agents into the workspace by dragging and dropping, connect them, set the workflow direction, and add judgment statements when necessary, so that this application can create the target application in a low-code form.

[0116] In some embodiments, the parameters of the target intelligent agent and / or the target tool agent that are determined are configured to include at least one of the following configurations:

[0117] One of them is to set the parameters of the target intelligent agent and / or the target tool body. For example, in combination with the above embodiment, when adjusting the color of an image, the saturation, brightness and other parameters of the target tool for image adjustment can be set.

[0118] Another is to select a model that is compatible with the target agent and / or the target tool body. The choice of model is directly related to the performance of the target agent. When selecting a model, it is necessary to clarify the application scenario, goals and specific problems to be solved of the target agent. This helps to determine what type of model is needed to support the decision-making and behavior of the target agent. Evaluate the applicability of different models according to the needs. This may include machine learning models (such as decision trees, random forests, neural networks, deep learning models, etc.), reinforcement learning models (such as policy gradients, etc.) or other types of agent models (such as rule-based models, fuzzy logic models, etc.). Based on the evaluation results, select the model that best suits the current application scenario and needs. Factors such as model accuracy, efficiency, interpretability, and availability of training data may be considered when selecting. In some cases, it may be necessary to integrate multiple models into an agent system to take advantage of the advantages of different models. This can be achieved through model fusion (such as weighted voting, stacked generalization, etc.) or hybrid intelligence.

[0119] The third is to adjust the behavior pattern or strategy of the target agent. The adjustment of the behavior pattern or strategy is directly related to the behavior performance of the target agent. Clarify the behavior that the target agent should exhibit in different situations. For parameter-based models (such as weights and biases in neural networks, learning rates and discount factors in reinforcement learning, etc.), optimize the behavior of the target agent by adjusting these parameters. This can be achieved through manual adjustment, grid search, random search, Bayesian optimization, gradient descent and other algorithms. In reinforcement learning, the behavior of the target agent is usually determined by its strategy. Policy optimization involves finding the optimal strategy through trial and error learning (such as training through simulated environments) so that the target agent can make the best decision in a given state. By allowing the target agent to interact in a simulated or real environment, observing its behavior, and adjusting it as needed, it helps the target agent gradually adapt to the environment and optimize its behavior.

[0120] The selected target agent has its own specific parameters. Some parameters of the target agent can be adjusted directly in the workspace, which also enables this application to create the target application in the form of low code, accelerate the development cycle of the target application, reduce the technical threshold of development, and make it easier for users to develop applications. In addition, it also enhances the scalability and maintainability of the target application and improves user satisfaction.

[0121] In some embodiments, in combination Figure 3 and Figure 4 Before determining the target agent, the method further includes:

[0122] S221, based on the requirements of the composite agent workflow, setting the steps of the composite agent workflow, and obtaining a plurality of second agents required to complete the composite agent workflow.

[0123] Exemplarily, the agent may include a composite agent, which may be a combination of multiple agents encapsulated in. The composite agent may be stored in an agent library and / or tool library and called according to user needs.

[0124] Specifically, the purpose and requirements of the composite agent workflow are first clarified, and based on the requirements of the composite agent work, which agents are used to complete the workflow are determined. According to the requirements of the composite agent workflow, the various steps of the composite agent workflow are planned and set, and the role of each step and the logical relationship between them are clarified. According to the requirements of the composite agent workflow and the agents used to complete the workflow, multiple second agents that need to be used are obtained. That is, the second agent is the agent required to complete the composite agent workflow.

[0125] S222, configuring the obtained parameters of the plurality of second agents to obtain a plurality of third agents adapted to complete the composite agent workflow.

[0126] Exemplarily, an agent library is maintained, which contains a variety of agents and their characteristics. These agents can be pre-defined or user-defined. According to the needs of the workflow, a suitable second agent is selected from the agent library. These second agents will serve as execution units in the composite agent workflow. For the selected second agent, the user can perform detailed editing, including selecting appropriate tools, models, and setting various parameters. This ensures that the third agent obtained after configuring the parameters can perform the task as expected.

[0127] S223, determining the relationship between the plurality of third agents.

[0128] Exemplarily, the relationship between multiple agents is determined, including the serial or parallel operation mode between them. The serial relationship means that multiple third agents are executed in a specific order, while the parallel relationship allows the third agents to execute at the same time. The user also needs to set the output results of the third agents and the interaction mode between them.

[0129] S224, based on the steps of the composite agent workflow and the relationship between the plurality of third agents, execute the composite agent workflow and output the operation result.

[0130] Exemplarily, the composite agent workflow is executed according to the set steps of the composite agent workflow and the relationship between the plurality of third agents. During the execution process, the third agents perform tasks according to the preset parameters and logic and generate corresponding outputs.

[0131] S225, encapsulating the plurality of third agents corresponding to the operation results that meet the expected result requirements to generate a composite agent.

[0132] Exemplarily, when all steps in the composite agent workflow are executed, the process ends. The user can check and evaluate the output operation results to ensure that the composite agent workflow has achieved the expected effect. For the third agent with a relatively good output operation result, the user can choose to encapsulate it for reuse in other workflows.

[0133] In some embodiments, the encapsulating of the plurality of third agents corresponding to the operation results satisfying the expected results to generate a composite agent comprises:

[0134] Encapsulate the plurality of third agents corresponding to the operation results that meet the expected results, generate a composite agent in the first format, and save it to the agent library; or,

[0135] The plurality of third agents corresponding to the operation results that meet the expected results are encapsulated to generate a composite agent having a second format, and the composite agent is saved to the tool library.

[0136] Exemplarily, the composite agent obtained after encapsulation can be saved as a Chat Bot, API (application programming interface) or other forms, which is convenient for users to manage and call. Specifically, multiple third agents corresponding to the operation results that meet the expected results are encapsulated to generate a composite agent with a first format, which is saved to the agent library. It is also possible to encapsulate multiple third agents corresponding to the operation results that meet the expected results to generate a composite agent with a second format. For example, it can be saved as a Chat Bot, and the Chat Bot is saved in a Bot library. Of course, the Bot library can also be one of the tool libraries. For another example, it is also possible to encapsulate multiple third agents corresponding to the operation results that meet the expected results to generate a composite agent with a second format, such as an API, and save it to a tool library.

[0137] The composite agent workflow is created in a low-code way, and the running results of the composite agent workflow are stored, which can be quickly packaged and re-edited. In addition, the created composite agent workflow can also be used to create other agents or applications, which is convenient for use.

[0138] In some embodiments, in combination Figure 5 and Figure 6 , obtaining a plurality of second agents required to complete the agent workflow, the method comprising:

[0139] S2211, based on the needs of the individual intelligent agent to be generated, determine the basic intelligent agent, tool body and model, wherein the basic intelligent agent, tool body and model are adapted to the needs of the individual intelligent agent to be generated.

[0140] Exemplarily, taking the LLM type of personalized agent as an example, multiple types of basic agents are preset, and each basic agent has a corresponding parameter template. When creating a personalized agent, the corresponding template will be generated according to the selected basic agent.

[0141] Based on the needs of the personalized agent to be generated, select the basic agent, tool agent and model. This is the basis for the construction of the personalized agent, among which the basic agent provides the core functions and architecture of the agent. According to the needs of the personalized agent, select appropriate pre-processing tools, intermediate tools, functional tools and post-processing tools. These tools will be used for data preprocessing, model training and result post-processing. You can also select appropriate models from the model library, which will be used for the actual operation and decision-making of the personalized agent.

[0142] S2212, adding or modifying parameters for the basic agent, and / or modifying the parameters of the model to obtain a fourth agent.

[0143] Exemplarily, according to the specific needs of the individual intelligent agent, parameters can be added or modified for the individual intelligent agent itself, or model-related parameters can be modified to obtain a fourth intelligent agent to optimize the performance and accuracy of the individual intelligent agent.

[0144] S2213, based on the application scenario of the personalized agent to be generated, select a prompt template to guide the operation of the fourth agent.

[0145] Exemplarily, according to the application scenario of the individual agent, an existing prompt template or a newly created template is selected to guide the behavior of the fourth agent. In the LLM type individual agent, the parameters added in step S2203 can also be associated with the content in the prompt template.

[0146] S2214: Based on the needs of the personalized intelligent agent to be generated, determine a knowledge base that can provide adaptation information for the personalized intelligent agent to be generated.

[0147] Exemplarily, based on the needs of the individual intelligent agent to be generated, an appropriate knowledge base is selected to provide the agent with necessary knowledge and information.

[0148] S2215: When the operation result of the fourth agent meets the expected result, the fourth agent is saved as the personalized agent.

[0149] Exemplarily, the constructed agent can be tested to ensure that it can run as expected and produce accurate results. That is, when the operating results of the fourth agent meet the expected results, the fourth agent is saved as a personalized agent. The personalized agents that pass the test are stored for subsequent use and application. The newly stored personalized agents can be updated to the agent library or to the tool library.

[0150] Combined with the above steps, take the establishment of a question answering personality agent as an example. First, select the react basic agent, and then fill in the basic information. The selected pre-processing tool is: text input tool. The intermediate tool is: run engine. The functional tool is: Wikipedia search tool, Google search tool. The post-processing tool is: text output. Select the LLM model: qwen-7B. Set parameters: context, question and answer word count. Prompt template: question answering prompt template. Knowledge base: Wikipedia knowledge base.

[0151] The main generation process of ML-type personalized agents is the same as that of LLM-type personalized agents, but there are no functional tools in the tool, and no prompt templates and knowledge bases are required, and the selection of this type of agent is mainly based on the selection of the algorithm library. By editing and modifying the basic intelligent templates different from the LLM class, it can also immediately become an agent with specific functions. Through the preset basic agents, various types of agents can be realized.

[0152] Take the establishment of a target detection agent as an example. First, select the ML basic agent and then fill in the basic information. The pre-processing tool selected is: image input tool. The intermediate tool is: detection engine. The post-processing tools are: text output, image output. Select the algorithm: target detection. Set the parameters: detection type, detection threshold.

[0153] Based on the original basic agent structure, a new personalized agent architecture is proposed. On the basis of the basic agent, relevant parameter settings are performed to make the personalized agent formed by the basic agent have better operation effects. In addition, the agent is not limited to the LLM class, so that the obtained personalized agent can handle more tasks.

[0154] In some embodiments, in combination Figure 7 , to publish the target application, including:

[0155] The packaged target application, configuration file and user manual are published.

[0156] For example, when packaging, the agent, tools and related resources are packaged into a complete application package. After packaging, the application package and related files are uploaded to the publishing platform. These files may include application packages, configuration files, user manuals, etc. Finally, the target application can also be updated to the application list. This may involve steps such as review, testing, and distribution to ensure the quality and security of the application.

[0157] Combination Figure 8 The present application also provides an application publishing system, including:

[0158] The first determination module is configured to determine a target application in response to an application creation request.

[0159] For example, the user can send an application creation request as needed, and in response to the application creation request, the user's functional requirements for the application are clarified to determine the target application. For example, if the user wants to create an animation APP (application program) that outputs an animation based on the input image, the target application can be an animation production APP.

[0160] The second determination module is configured to determine the initial parameters of the target application, and determine the target intelligent body and the target tool body based on the determined functional requirements of the target application.

[0161] Exemplarily, in combination with the first determination module, based on the determined functional requirements of the target application, the initial parameters of the target application are determined, where the initial parameters may include, but are not limited to, the name, version, and operating environment of the target application. For example, in combination with the above embodiment, the initial parameters of the target application include: animation production, version 1.0, IOS system or Android system.

[0162] In addition, according to the functional requirements of the target application, a suitable tool is determined, which can be selected from the tool library. It should be noted that the tool library here can include both agents and tool bodies. That is, the agent can also be used as a tool body. Of course, suitable agents and tool bodies can also be selected from the agent library and the tool library respectively. Specifically, according to the functional requirements of the target application, a suitable target agent is determined, where the target agent can be one or more. These target agents will be responsible for executing key functions in the target application, etc. In the case where there is one target agent, the target agent can realize the functions required by the target application alone; in the case where there are multiple target agents, multiple target agents can cooperate to realize the functions required by the target application. Continuing with the above embodiment, according to the functional requirements of the animation production APP, a suitable target agent can be selected from the agent library, and the target agent can have both the function of image processing and the function of converting images into animations. Two target agents can also be selected, one of which has the function of image processing and the other has the function of converting processed images into animations.

[0163] Similarly, the appropriate target tool body can be determined according to the functional requirements of the target application. There are various types of target tool bodies, such as service tools, library tools, script tools, etc. The target tool body can also be one or multiple. In the case of one target tool body, it can independently assist the operation of the target intelligent body. For example, the target tool body has functions such as cropping, scaling, rotating, and adjusting image color, which can assist the target intelligent body of image processing to complete the processing of the input image. In the case of multiple target tool bodies, multiple target tool bodies can cooperate to assist the operation of the target intelligent body. For example, one of the target tool bodies has the function of cropping, scaling, and rotating the image, and the other target tool body can adjust the color of the image to increase the ability of the target intelligent body and help the target intelligent body better perform and complete its functions. The two target tool body systems cooperate with the target intelligent body of auxiliary image processing to perform image processing.

[0164] A configuration module is configured to configure the determined parameters of the target intelligent agent and / or the target tool body based on the initial parameters of the target application to obtain a first intelligent agent and a first tool body adapted to the target application.

[0165] Exemplarily, according to the initial parameters of the target application, the parameters of the target agent and / or the target tool body are configured and edited. The configuration and editing of the parameters of the target agent and / or the target tool body may include, but is not limited to, setting corresponding parameters, selecting an adapted model, adjusting behavior, etc., to ensure that they can work as expected. For example, if the target agent needs to be modified, the target agent is entered to modify it.

[0166] After modification, a first agent adapted to the target application is obtained, and the first agent may include one or more agents. After modification, a first tool body adapted to the target application is obtained, and the first tool body may include one or more agents.

[0167] The packaging and publishing module is configured to package the first intelligent body and the first tool body into a target application so as to perform a publishing operation on the target application.

[0168] Exemplarily, the first intelligent agent and the first tool agent are encapsulated to obtain an independent target application, for example, the target application is an animation production APP. The target application is published, for example, according to user needs, the target application is published to one or more corresponding platforms.

[0169] An embodiment of the present application also provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0170] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, electronic equipment, computer-readable storage media or computer program products. Therefore, the present application may adopt the form of complete hardware embodiments, complete software embodiments or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media that include computer-usable program code. When implemented by software, these functions may be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0171] The processor may be a general purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof. The general purpose processor may be a microprocessor or any conventional processor.

[0172] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0173] The above-mentioned readable storage medium can be a disk, an optical disk, a DVD, a USB, a read-only memory (ROM) or a random access memory (RAM), etc. This application does not limit the specific form of the storage medium.

[0174] Multiple embodiments of the present application are described in detail above, but the present application is not limited to these specific embodiments. Those skilled in the art can make various variations and modifications to the embodiments based on the concept of the present application, and these variations and modifications should all fall within the scope of protection required by the present application.

Claims

1. An application publishing method, characterized in that: include: In response to the application creation request, determining a target application; Based on the determined functional requirements of the target application, initial parameters of the target application are determined, and a target agent is determined from an agent library and a target tool body is determined from a tool library, wherein the target agent is configured to be able to realize the functions required by the target application alone or in collaboration, and the target tool body is configured to be able to assist the operation of the target agent alone or in collaboration; the tool library can include agents and tool bodies at the same time; Before determining the target agent, the method further includes: Based on the requirements of the composite agent workflow, the steps of the composite agent workflow are set to obtain a plurality of second agents required to complete the composite agent workflow; Wherein, obtaining a plurality of second agents required to complete the agent workflow, the method comprises: Among them, the basic agent is a single agent with intelligent characteristics and functions, and the personalized agent is an agent with unique personality and functions, which is the inheritance and extension of the basic agent or the composite agent. The basic agent, tool agent and model are adapted to the needs of the personalized agent to be generated; Adding or modifying parameters for the basic agent, and / or modifying parameters of the model, to obtain a fourth agent; Based on the application scenario of the personalized agent to be generated, selecting a prompt template to guide the operation of the fourth agent; Based on the needs of the personalized intelligent agent to be generated, determining a knowledge base that can provide adaptation information for the personalized intelligent agent to be generated; When the operation result of the fourth agent meets the expected result, the fourth agent is saved as the personalized agent; Configuring the obtained parameters of the plurality of second agents to obtain a plurality of third agents adapted to complete the composite agent workflow; Determining the relationship between the plurality of said third agents; Based on the steps of the composite agent workflow and the relationship between the plurality of third agents, execute the composite agent workflow and output the operation result; Encapsulating a plurality of the third agents corresponding to the operation results that meet the expected result requirements to generate a composite agent; wherein the composite agent includes a chatbot and an application programming interface; Based on the initial parameters of the target application, the parameters of the target intelligent agent and / or the target tool body are configured to obtain a first intelligent agent and a first tool body adapted to the target application; The first intelligent agent and the first tool agent are encapsulated as a target application or a chat robot to perform a publishing operation on the target application.

2. The application publishing method according to claim 1, characterized in that: The step of determining the target application in response to the application creation request includes: In response to the application creation request, generating a first application, determining the first application as the target application, or, A second application is selected from the application template, and the second application is determined as the target application.

3. The application publishing method according to claim 1, characterized in that: Based on the determined functional requirements of the target application, a target intelligent agent and a target tool agent are determined, including: Dividing the functional requirements of the target application into multiple tasks or operations; Based on the determined task requirements or operation requirements of the target application, at least one target agent capable of performing the task or the operation is retrieved and selected from an agent library, wherein the target agent is configured to be capable of realizing the functions of one or more of the tasks or the operation individually or collaboratively, and at least one target tool body capable of assisting the operation of the target agent is retrieved and selected from a tool library; Based on the compatibility requirements of the target application, at least one target intelligent entity that meets the compatibility requirements of the target application is determined, and at least one target tool entity that meets the compatibility requirements of the target application is determined.

4. The application publishing method according to claim 1, characterized in that: The parameters of the target intelligent entity and / or the target tool entity determined by configuration include at least one of the following configuration modes: Setting parameters of the target intelligent entity and / or the target tool entity; Selecting a model that is compatible with the target intelligent entity and / or the target tool entity; Adjust the behavior pattern or strategy of the target agent.

5. The application publishing method according to claim 1, characterized in that: The step of encapsulating a plurality of third agents corresponding to the operation results that meet the expected results to generate a composite agent includes: Encapsulate the plurality of third agents corresponding to the operation results that meet the expected results, generate a composite agent in the first format, and save it to the agent library; or, The plurality of third agents corresponding to the operation results that meet the expected results are encapsulated to generate a composite agent having a second format, and the composite agent is saved to the tool library.

6. The application publishing method according to claim 1, characterized in that: The target application is published, including: The packaged target application, configuration file and user manual are published.

7. An application publishing system, characterized in that: include: A first determination module configured to determine a target application in response to an application creation request; A second determination module is configured to determine initial parameters of the target application based on the determined functional requirements of the target application, and to determine a target agent from an agent library and a target tool body from a tool library, wherein the tool library can include both agents and tool bodies; Before determining the target intelligent agent, the system further includes: Based on the requirements of the composite agent workflow, the steps of the composite agent workflow are set to obtain a plurality of second agents required to complete the composite agent workflow; Wherein, a plurality of second agents required to complete the agent workflow are obtained, and the system comprises: Based on the needs of the individual intelligent agent to be generated, determine the basic intelligent agent, the tool agent and the model, wherein the basic intelligent agent, the tool agent and the model are adapted to the needs of the individual intelligent agent to be generated; Adding or modifying parameters for the basic agent, and / or modifying parameters of the model, to obtain a fourth agent; Based on the application scenario of the personalized agent to be generated, selecting a prompt template to guide the operation of the fourth agent; Based on the needs of the personalized intelligent agent to be generated, determining a knowledge base that can provide adaptation information for the personalized intelligent agent to be generated; When the operation result of the fourth agent meets the expected result, the fourth agent is saved as the personalized agent; Configuring the obtained parameters of the plurality of second agents to obtain a plurality of third agents adapted to complete the composite agent workflow; Determining the relationship between the plurality of said third agents; Based on the steps of the composite agent workflow and the relationship between the plurality of third agents, execute the composite agent workflow and output the operation result; Encapsulating a plurality of the third agents corresponding to the operation results that meet the expected result requirements to generate a composite agent; wherein the composite agent includes a chatbot and an application programming interface; A configuration module configured to configure the determined parameters of the target intelligent agent and / or the target tool body based on the initial parameters of the target application to obtain a first intelligent agent and a first tool body adapted to the target application; A packaging and publishing module is configured to package the first intelligent agent and the first tool agent into a target application or a chat robot so as to perform a publishing operation on the target application.

8. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method as claimed in any one of claims 1 to 6 are implemented.