An industry application intelligent agent platform based on a large model

Abstract

This invention discloses an industry application intelligent agent platform based on a large model, relating to the field of computer application technology. The invention generates resource data fragments based on historical industry data, inputs these fragments into an industry resource pool, obtains initial user demand tasks, establishes multiple demand elements based on these tasks, and constructs a demand association logic tree based on the relationship logic between these elements. Resource data blocks are then retrieved from the industry resource pool through this tree. These resource data blocks are categorized and combined according to the connection relationships between demand elements to generate multiple application intelligent agents. Furthermore, resource data blocks are retrieved again from the industry resource pool based on subsequent user demand tasks to optimize each application intelligent agent.

Description

Technical Field

[0001] This invention relates to the field of computer application technology, and more specifically to an industry application intelligent agent platform based on a large model. Background Technology

[0002] With the accelerated advancement of digital transformation, various industries have an increasingly urgent need for intelligent applications. However, traditional industry application development and implementation models face numerous bottlenecks. On the one hand, industry data is characterized by fragmentation and decentralization, lacking a unified resource integration and management mechanism. A large amount of historical industry data has not been effectively mined and utilized, resulting in a disconnect between resource supply and actual demand, making it difficult to form large-scale and systematic resource support.

[0003] On the other hand, user needs are often personalized and scenario-based. Initial requirements are often vague and abstract, lacking detailed breakdown and logical organization, which can easily lead to discrepancies between application development and actual needs. In addition, traditional application development is mostly a fixed process, and the resulting applications cannot be dynamically adjusted and optimized according to subsequent user needs. They lack adaptability and flexibility, and have long development cycles and high costs, making it difficult to meet the industry's need for rapid iteration.

[0004] Against this backdrop, there is an urgent need to build an efficient, collaborative, and intelligently adaptable industry application solution, and to this end, we provide an industry application intelligent agent platform based on a large model. Summary of the Invention

[0005] The purpose of this invention is to provide an industry application intelligent agent platform based on a large model to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An industry application intelligent agent platform based on a large model includes a cloud resource management module, an industry demand analysis module, and a decision support module; The cloud resource management module is used to acquire historical industry data and has a built-in industry resource pool. It generates resource data fragments based on the historical industry data and inputs the resource data fragments into the industry resource pool. The industry demand analysis module is used to obtain initial user demand tasks, establish multiple demand elements based on the initial user demand tasks, and establish a demand association logic tree based on the association logic between each demand element. The decision support module is used to obtain resource data blocks from the industry resource pool through the demand association logic tree, classify and combine the resource data blocks according to the connection relationship between demand elements to generate multiple application agents, and obtain resource data blocks from the industry resource pool again according to subsequent user demand tasks to optimize each application agent.

[0007] Furthermore, the industry resource pool includes a main pool, sub-pools, and sub-pools. The main pool is responsible for coordinating the metadata management and cross-industry resource scheduling of resources across the entire industry. The sub-pools include multiple first-level industry sub-pools, and each first-level industry sub-pool is further divided into sub-pools according to core business scenarios. Data synchronization and isolation between resource pools at all levels are achieved through a distributed database cluster.

[0008] Furthermore, the process of generating the resource data fragment includes: Each resource data block consists of several resource data fragments. Historical industry data is acquired, and resource data fragments are extracted from the historical industry data. A unique data feature is assigned to each resource data fragment, and the data feature is used for the retrieval and matching of resource data fragments; Set the resource data fragment update cycle, which is generally 1 hour. After each resource data fragment is updated, re-establish the association relationship between each resource data fragment. The association relationship includes association type and association scenario.

[0009] Furthermore, the process of extracting requirement elements from the initial user requirement task includes: The requirement element extraction process includes keyword extraction, entity recognition, and intent parsing. First, the core keywords in the requirements are extracted from the initial user requirement task, and then the corresponding requirement elements are extracted from the core keywords. The extracted requirement elements are formatted in a unified manner according to the preset industry requirement element standard specifications. After the requirement elements are extracted, a requirement association logic tree is established based on the association logic between each requirement element.

[0010] Furthermore, the association logic includes serial logic, parallel logic, inclusion logic, and constraint logic; The sequential logic states that there is a sequential dependency relationship between the requirement elements, that is, the implementation of the previous requirement element is a prerequisite for the implementation of the next requirement element. Parallel logic: The demand elements are independent of each other, can be promoted simultaneously, and have no obvious dependency relationship; The inclusion logic is as follows: one requirement element contains multiple sub-requirement elements; The constraint logic states that one requirement element restricts the scope, method, or goal of implementing another requirement element.

[0011] Furthermore, the process of establishing the requirement association logic tree includes: The user's core needs and intentions are used as the root node of the requirement association logic tree. The requirement elements directly associated with the root node are used as first-level branches. The association logic type is marked on the connection line of each branch, and the related association scenarios of the requirement elements are also marked.

[0012] Furthermore, the decision support module performs semantic mapping between the core keywords of each demand element in the demand association logic tree and the data features of the resource data blocks in the cloud resource management module, and establishes a demand-resource feature mapping table, which contains the demand feature vector of each demand element and the resource feature vector of the resource data block. The cosine similarity algorithm is used to calculate the matching degree between each demand element and each resource data block. A resource matching degree threshold is set. For each demand element, all resource data blocks with a matching degree higher than the resource matching degree threshold are retrieved. If there is a demand element without a resource data block with a matching degree higher than the resource matching degree threshold, a resource missing warning is automatically triggered, and the user is prompted to adjust the demand constraints or wait for the industry resource pool to be updated.

[0013] Furthermore, the process of constructing the application agent includes: For demand elements that have a serial or inclusive relationship in the demand association logic tree, the corresponding resource data blocks are combined in the serial order to form the process components of the intelligent agent, forming an input-processing-output process link; For demand elements that have parallel relationships, the corresponding resource data blocks are combined in parallel to form parallelized components of the agent. The resource data blocks corresponding to the constraints are used as adapting components and embedded in various levels of the agent to constrain the operating range and mode of the core components and extended components. Then, based on the resource data block classification results, an application intelligent agent is constructed. The application intelligent agent consists of a core component, an extended component, and an adaptation component. The core component is composed of core resource data blocks that directly match the requirements of the root node and is responsible for realizing the user's core requirements. The extended component consists of resource data blocks that match the requirements of the first-level branches and is responsible for achieving the partitioning objectives under the core requirements; The adaptation component consists of resource data blocks that match the constraint requirements.

[0014] Furthermore, based on changes in user needs and tasks, resource data blocks are retrieved from the industry resource pool and the various components of the application agent are adjusted. This includes matching the corresponding resource data blocks as new components to the application agent when users add new requirement elements. When a user modifies a requirement element, the original resource data block is replaced with a resource data block that matches the new requirement. When a user deletes a requirement element, the corresponding resource data block component in the application agent is removed. When all components of any application agent are deleted, the corresponding application agent is automatically deleted.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention integrates and structures historical industry data to form standardized resource data fragments and industry resource pools, enabling centralized management and efficient scheduling of resources. It fully leverages the value of historical data, avoids resource waste, and provides comprehensive and accurate resource support for application development. At the same time, it breaks down abstract initial requirements into specific requirement elements and constructs a requirement association logic tree to clearly sort out the logical relationships of requirements. This ensures a deep understanding of user needs to a certain extent, making subsequent resource matching and intelligent agent generation more in line with actual needs and reducing the risk of requirement deviation.

[0016] 2. This invention can quickly match resources to generate application agents based on demand logic, and can respond to subsequent user needs by dynamically retrieving resources to optimize and iterate the agents. This breaks the fixed limitations of traditional applications, effectively improves the application's adaptability to scene changes, and meets users' personalized and dynamic needs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of an industry application intelligent agent platform based on a large model, as described in this invention.

[0019] Figure 2 This is a system architecture block diagram of an industry application intelligent agent platform based on a large model, as described in this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 and Figure 2 As shown, an industry application intelligent agent platform based on a large model includes a cloud resource management module, an industry demand analysis module, and a decision support module. The cloud resource management module is used to acquire historical industry data and has a built-in industry resource pool. It generates resource data fragments based on the historical industry data and inputs the resource data fragments into the industry resource pool. The industry demand analysis module is used to obtain initial user demand tasks, establish multiple demand elements based on the initial user demand tasks, and establish a demand association logic tree based on the association logic between each demand element. The decision support module is used to obtain resource data blocks from the industry resource pool through the demand association logic tree, classify and combine the resource data blocks according to the connection relationship between demand elements to generate multiple application agents, and obtain resource data blocks from the industry resource pool again according to subsequent user demand tasks to optimize each application agent.

[0022] Furthermore, the working principle of the present invention will be illustrated below through embodiments: The cloud resource management module has a built-in industry resource pool, which includes a main pool, sub-pools, and sub-pools. The main pool is responsible for coordinating the metadata management and cross-industry resource scheduling of resources across the entire industry. The sub-pools are divided into multiple first-level industry sub-pools according to the national economic industry classification standards, such as the manufacturing industry resource pool, the financial industry resource pool, the healthcare industry resource pool, and the smart city resource pool. Each first-level industry sub-pool is further divided into sub-pools based on core business scenarios (e.g., the manufacturing sub-pool includes production scheduling sub-pools, quality inspection sub-pools, supply chain management sub-pools, equipment operation and maintenance sub-pools, etc.). Data synchronization and isolation between resource pools at all levels are achieved through a distributed database cluster to ensure that resource calls from different industries and scenarios do not interfere with each other, while also supporting the linkage of cross-industry resources (e.g., the integration of resources from multiple industries such as transportation, security, and energy is required in smart city construction). Each sub-pool contains several resource data blocks. The resource data block is the basic unit of resource organization. Its division follows the principle of common business scenario and data feature association. That is, each resource data block corresponds to a core business requirement under a certain business scenario. For example, in the manufacturing production scheduling sub-pool, there are data blocks such as production line capacity allocation, production plan adjustment, and material supply and demand balance; in the financial industry resource pool, there are data blocks such as customer credit assessment, credit risk control, and loan amount calculation in the credit business sub-pool. Each resource data block consists of several resource data fragments. The resource data fragment is the smallest data unit that makes up the resource data block. It contains a variety of structured and unstructured data, such as industry information, industry decision-making cases, business process specifications, technical parameter standards, market dynamic data, and policy and regulatory provisions. The process of generating the resource data fragments includes: collecting historical industry business data, classic decision-making cases, and past project implementation plans from the past 5-10 years by connecting with the business systems, industry databases, and public data platforms of leading companies in the industry; performing semantic parsing, entity recognition, and relation extraction on all historical industry data; and then extracting the core information from the historical industry data to generate structured resource data fragments. Each resource data segment is assigned a unique data feature, which includes core attributes such as data type (e.g., text, numerical, image, audio), business tags (e.g., production scheduling - capacity optimization, credit approval - risk assessment), data source (e.g., an ERP system of an automobile manufacturing company), data validity period (e.g., long-term validity until the end of 2024), and data credibility level, for the retrieval and matching of resource data segments. Set the resource data fragment update cycle. The duration of the resource data fragment update cycle is generally 1 hour. After each resource data fragment is updated, the association relationship between each resource data fragment is re-established. The association relationship includes the association type and association scenario. The association types include business process associations (such as production planning data blocks and material procurement data blocks), data dependency associations (such as customer credit assessment data blocks and customer asset verification data blocks), and goal collaboration associations (such as energy consumption reduction data blocks and production efficiency improvement data blocks). The associated scenarios are used to clarify the specific business scenarios for associated calls to resource data blocks, such as production line changeover scenarios in manufacturing workshops and personal housing loan approval scenarios in the financial industry.

[0023] Furthermore, users can upload initial user requirements tasks through the industry requirements analysis module. These initial user requirements tasks can be input in the form of natural language text, voice, mind maps, etc. For example, developing a production scheduling optimization system suitable for small and medium-sized electronics factories, requiring a 10% reduction in production costs and a 15% increase in production efficiency. Then, the industry demand analysis module extracts the demand elements from the initial user demand task: The requirement element extraction process includes keyword extraction, entity recognition, and intent parsing. First, core keywords (such as production scheduling, cost reduction, 10%, 15%) are extracted from the initial user requirement task. Then, the entity types corresponding to the core keywords are identified (such as production scheduling for business scenarios, cost reduction for target indicators, and 10% for numerical constraints). This clarifies the user's core requirement intent (such as building a production scheduling optimization system to achieve cost reduction and efficiency improvement). Finally, requirement elements are extracted from the core requirement intent. These requirement elements include business scenario elements, target indicator elements, constraint elements, technical requirement elements, and time cycle elements. For the extracted demand elements, they are uniformly formatted according to the preset industry demand element standards and specifications. For example, target indicator elements are uniformly standardized into the format of indicator type-target value-unit of measurement (such as cost reduction-10%-percentage, efficiency improvement-15%-percentage), and constraint elements are standardized into the format of constraint type-constraint content-constraint strength (such as technical constraint-compatible with existing ERP system-strong constraint, budget constraint-not exceeding 500,000 yuan-strong constraint). This ensures that demand elements from different sources and in different formats have uniform identifiability and comparability. Once the requirement elements are extracted, a requirement association logic tree is established based on the association logic between each requirement element. The association logic includes serial logic, parallel logic, inclusion logic, and constraint logic. The sequential logic is as follows: there is a sequential dependency relationship between the demand elements, that is, the implementation of the previous demand element is a prerequisite for the implementation of the next demand element, such as the sequential relationship of equipment data acquisition → data anomaly analysis → fault early warning generation. Parallel logic: Demand elements are independent of each other and can be promoted simultaneously without obvious dependencies, such as the parallel relationship between production planning optimization and material inventory optimization; The included logic is as follows: a requirement element contains multiple sub-requirement elements, such as customer management containing sub-elements such as customer information entry, customer classification tags, and customer behavior analysis; The constraint logic refers to the restriction imposed by one requirement element on the scope, method, or goal of another requirement element, such as the constraint relationship between budget constraints and technology selection. The process of establishing the demand association logic tree includes: taking the user's core demand intent as the root node of the demand association logic tree, such as building a production scheduling optimization system for small and medium-sized electronics factories, and then taking the demand elements directly associated with the root node as first-level branches, such as cost reduction, efficiency improvement, and compatibility with existing system budget control. Label the associated logic type (serial, parallel, containment, and constraint) on the connection lines of each branch, and also label the associated scenarios related to the requirement elements.

[0024] Furthermore, once the demand association logic tree is established, the industry demand analysis module will synchronize the demand association logic tree with the decision support module. Then, the decision support module performs semantic mapping between the core keywords (such as business tags, target indicators, and constraints) of each requirement element in the requirement association logic tree and the data characteristics (such as business tags, data usage, and adaptation scenarios) of the resource data blocks in the cloud resource management module, and establishes a requirement-resource feature mapping table. The requirement-resource feature mapping table contains the requirement feature vector of each requirement element and the resource feature vector of the resource data block. The cosine similarity algorithm is used to calculate the matching degree between each requirement element and each resource data block. The matching degree calculation formula is as follows: Matching degree = (demand feature vector * resource feature vector) / (||demand feature vector|| × ||resource feature vector||), where the demand feature vector and resource feature vector are generated by the large model based on the feature extraction algorithm, and the matching degree ranges from 0 to 1, with the closer to 1 indicating a higher matching degree. Set the resource matching threshold (default value is 0.7, which can be manually adjusted by the user). For each requirement element, retrieve all resource data blocks with a matching degree higher than the resource matching threshold. If there is a requirement element without a resource data block with a matching degree higher than the resource matching threshold, a resource missing warning will be automatically triggered, and the user will be prompted to adjust the requirement constraints or wait for the industry resource pool to be updated. Based on the association information between resource data blocks marked in the cloud resource management module, when a core resource data block is retrieved, other related resource data blocks are automatically retrieved to ensure the integrity and relevance of resources. For example, when retrieving the production plan optimization data block, related resource data blocks such as material supply and demand balance data block and equipment capacity data block are automatically retrieved.

[0025] Furthermore, after the resource data block scheduling is completed, the resource data blocks are classified and combined according to the connection relationships between the corresponding demand elements to generate multiple application intelligent agents. The construction process of the application intelligent agents includes: For demand elements that have a serial or inclusive relationship in the demand association logic tree, the corresponding resource data blocks are combined in the serial order to form the process components of the intelligent agent, forming a process link of input, processing and output, such as the serial combination of equipment data acquisition data block → data anomaly analysis data block → fault early warning generation data block. For demand elements that are connected in parallel, the corresponding resource data blocks are combined in parallel to form parallel components of the agent, supporting the simultaneous processing of multiple tasks, such as the parallel combination of production planning optimization data blocks and material inventory optimization data blocks. The resource data blocks corresponding to the constraints are used as adaptation components and embedded in various levels of the intelligent agent to constrain the operating scope and methods of the core components and extended components. For example, the budget control data blocks are embedded in the intelligent agent to limit the cost of technology selection for the core components. Then, based on the resource data block classification results, an application intelligent agent is constructed. The application intelligent agent consists of a core component, an extended component, and an adaptation component. The core component is composed of core resource data blocks that directly match the requirements of the root node and is responsible for realizing the user's core requirements. The extended component consists of resource data blocks that match the requirements of the first-level branches and is responsible for achieving the partitioning objectives under the core requirements; The adaptation component consists of resource data blocks that match the constraint requirements and is responsible for ensuring the compatibility, security, and compliance of the intelligent agent. Assign a unique agent ID to each generated application agent, set initial running parameters (such as data processing priority and resource quota), and generate an agent description document containing information such as the agent's functional modules, resource composition, suitable scenarios, and expected goals, so that users can view and compare it. Once the application agents are built, the decision support module continuously optimizes and iterates each application agent based on user-uploaded task requirements, ensuring dynamic adaptation between the application agents and user needs. Users can upload new user requirement tasks through real-time communication with the industry requirement analysis module. This includes requirement adjustments (such as modifying target indicators and adding constraints), requirement refinement (such as supplementing specific business process requirements), and requirement expansion (such as adding new functional module requirements). Based on the changes in user requirement tasks, resource data blocks are retrieved from the industry resource pool to adjust the various components of the application agent. This includes matching the corresponding resource data blocks as new components to the application agent when users add new requirement elements. When a user modifies a requirement element (such as adjusting target metrics), the original resource data block is replaced with a resource data block that matches the new requirement. When a user deletes a requirement element, the corresponding resource data block component in the application agent is removed. When all components of any application agent are deleted, the corresponding application agent is automatically deleted. The process of continuously optimizing and iterating the intelligent agent is repeated until the user stops uploading user-required tasks.

[0026] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A large model-based industry application intelligent agent platform, characterized in that, It includes a cloud resource management module, an industry demand analysis module, and a decision support module; The cloud resource management module is used to acquire historical industry data and has a built-in industry resource pool. It generates resource data fragments based on the historical industry data and inputs the resource data fragments into the industry resource pool. The industry demand analysis module is used to obtain initial user demand tasks, establish demand elements based on the initial user demand tasks, and establish a demand association logic tree based on the association logic between each demand element. The decision support module is used to obtain resource data blocks from the industry resource pool through the demand association logic tree, and classify and combine the resource data blocks according to the connection relationship between demand elements to generate multiple application intelligent agents. Then, based on subsequent user demand tasks, it obtains resource data blocks from the industry resource pool again to optimize each application intelligent agent.

2. The industry application intelligent agent platform based on a large model according to claim 1, characterized in that, The industry resource pool includes a main pool, sub-pools, and sub-pools. The main pool is responsible for coordinating the metadata management and cross-industry resource scheduling of resources across the entire industry. The sub-pools include multiple first-level industry sub-pools, and each first-level industry sub-pool is further divided into sub-pools based on core business scenarios.

3. The large model-based industry application agent platform of claim 2, wherein, The process of generating the resource data fragment includes: Each resource data block consists of several resource data fragments. Historical industry data is acquired, and resource data fragments are extracted from the historical industry data. A unique data feature is assigned to each resource data fragment, and the data feature is used for the retrieval and matching of resource data fragments; Set the resource data fragment update cycle, and re-establish the association relationship between each resource data fragment after each resource data fragment is updated. The association relationship includes association type and association scenario.

4. The large model-based industry application agent platform of claim 3, wherein, The process of extracting requirement elements from the initial user requirement task includes: The requirement element extraction process includes keyword extraction, entity recognition, and intent parsing. First, the core keywords in the requirements are extracted from the initial user requirement task, and then the corresponding requirement elements are extracted from the core keywords. The extracted requirement elements are formatted in a unified manner according to the preset industry requirement element standard specifications. After the requirement elements are extracted, a requirement association logic tree is established based on the association logic between each requirement element.

5. The large model based industry application agent platform of claim 4, wherein, The associated logic includes serial logic, parallel logic, inclusion logic, and constraint logic; The sequential logic states that there is a sequential dependency relationship between the requirement elements, that is, the implementation of the previous requirement element is a prerequisite for the implementation of the next requirement element. Parallel logic: The demand elements are independent of each other, can be promoted simultaneously, and have no obvious dependency relationship; The inclusion logic is as follows: one requirement element contains multiple sub-requirement elements; The constraint logic states that one requirement element restricts the scope, method, or goal of implementing another requirement element.

6. The large model-based industry application agent platform of claim 5, wherein, The process of establishing the demand association logic tree includes: The user's core needs and intentions are used as the root node of the requirement association logic tree. The requirement elements directly associated with the root node are used as first-level branches. The association logic type is marked on the connection line of each branch, and the related association scenarios of the requirement elements are also marked.

7. The large model based industry application agent platform according to claim 6, wherein, The process of retrieving resource data blocks from the industry resource pool through the demand association logic tree includes: The decision support module performs semantic mapping between the core keywords of each demand element in the demand association logic tree and the data features of the resource data blocks in the cloud resource management module, and establishes a demand-resource feature mapping table, which contains the demand feature vector of each demand element and the resource feature vector of the resource data block. The cosine similarity algorithm is used to calculate the matching degree between each demand element and each resource data block. A resource matching degree threshold is set. For each demand element, all resource data blocks with a matching degree higher than the resource matching degree threshold are retrieved. If there is a demand element without a resource data block with a matching degree higher than the resource matching degree threshold, a resource missing warning is automatically triggered.

8. The large model-based industry application agent platform of claim 7, wherein, The process of constructing the application intelligent agent includes: For demand elements that have a serial or inclusive relationship in the demand association logic tree, the corresponding resource data blocks are combined in the serial order to form the process components of the intelligent agent, forming an input-processing-output process link; For demand elements that have parallel relationships, the corresponding resource data blocks are combined in parallel to form parallelized components of the agent. The resource data blocks corresponding to the constraints are used as adapting components and embedded in various levels of the agent to constrain the operating range and mode of the core components and extended components. Then, based on the resource data block classification results, an application intelligent agent is constructed. The application intelligent agent consists of a core component, an extended component, and an adaptation component. The core component is composed of core resource data blocks that directly match the requirements of the root node and is responsible for realizing the user's core requirements. The extended component consists of resource data blocks that match the requirements of the first-level branches and is responsible for achieving the partitioning objectives under the core requirements; The adaptation component consists of resource data blocks that match the constraint requirements.

9. The large model based industry application agent platform of claim 8, wherein, The process of re-obtaining resource data blocks from the industry resource pool and optimizing each application agent based on subsequent user needs includes: Based on changes in user needs and tasks, resource data blocks are retrieved from the industry resource pool and the various components of the application agent are adjusted. This includes matching the corresponding resource data blocks as new components to the application agent when users add new requirement elements. When a user modifies a requirement element, the original resource data block is replaced with a resource data block that matches the new requirement. When a user deletes a requirement element, the corresponding resource data block component in the application agent is removed. When all components of any application agent are deleted, the corresponding application agent is automatically deleted.

Images