Application management method and related equipment

Abstract

The application management method is executed by an application management platform, the application management platform comprises an interaction device, an agent expansion device and a service agent of at least one service, and the method comprises the following steps: the interaction device obtains a query input by a user, obtains an agent identifier and a condition, and collects a context of the query; the agent expansion device determines strategy configuration information of a target agent corresponding to the agent identifier according to the agent identifier and the condition; and the target agent expands the context of the query according to the strategy configuration information, assembles the context and the expanded context according to the strategy configuration information to obtain a prompt, inputs the prompt into the generation model for reasoning, and obtains an answer. According to the method, through a pluggable configuration mechanism, strategy configuration is dynamically selected and adjusted, the problem of insufficient flexibility caused by fixed configuration is avoided, and a large number of strategy combinations generated under different conditions are effectively managed and optimized. And the routing strategy can be quickly adjusted according to the customization requirement, so that the problem of combinatorial explosion is effectively solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence (AI), and more particularly to an application management method, an application management platform, a computing device cluster, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous development of artificial intelligence (AI) technology, especially in the field of natural language processing (NLP), language models (LM) are widely used in many scenarios, such as machine translation and text generation.

[0003] In recent years, with the emergence of pre-trained transformer architectures, such as Bidirectional Encoder Representations from Transformers (BERT) and Generative Pre-trained Transformer (GPT), large language models (LLMs) have made significant progress. As the parameter scale of large language models expands to tens of billions and begins to exhibit early signs of artificial general intelligence, LLMs can also be applied to domains beyond text processing. For example, in application development, LLMs can assist developers by generating code.

[0004] LLMs can generate answers to user-input queries. Currently, many LLM responses have low acceptance rates. Taking LLM application in code generation as an example, although LLM has shown great potential in code generation, code acceptance in real-world projects still faces many challenges. To address this, the industry has introduced agents. These agents enhance LLM capabilities through techniques such as Supervised Fine-Tuning (SFT), Retrieval-augmented Generation (RAG), Everything of Thoughts (XoT), Iterative Optimization, Context Engineering, or Tool Invocation, to meet customized needs and improve response acceptance rates.

[0005] However, when faced with diverse user needs and complex business environments, the above solutions often exhibit insufficient flexibility and poor adaptability. Summary of the Invention

[0006] This application provides an application management method that dynamically selects and adjusts policy configurations such as prompts, prompt templates, truncation strategies, language models, knowledge bases, retrieval strategies, recall strategies, tools, risk control strategies, and role types through a pluggable configuration mechanism. This avoids the lack of flexibility caused by fixed configurations and effectively manages and optimizes a large number of policy combinations generated by different business needs, programming languages, IDE types, and organizations / departments. This application also provides an application management platform, computing device cluster, computer-readable storage medium, and computer program products corresponding to the above method.

[0007] Firstly, this application provides an application management method. This method is executed by an application management platform. The application management platform can be a software system, which can be a standalone software system or a software system integrated with other software. The software system can be provided to customers as a software package for self-deployment, or it can be provided to users as a cloud service, with users subscribing to the cloud service to use the corresponding capabilities or functions. The aforementioned software system can be deployed in a computing device cluster, where the computing device cluster executes the program code of the software system to perform the application management method of this application. In some examples, the application management platform can also be a hardware system, such as a computing device cluster with application management capabilities. When the computing device cluster is running, it executes the application management method of this application.

[0008] The application management platform includes an interactive device, an intelligent agent extension device (also known as a strategy center), and at least one business intelligent agent for a specific business. Specifically, the interactive device acquires the user's input query, as well as the intelligent agent identifier and conditions. The conditions include at least one of the following: task type, developer affiliation information, programming language, programming framework, or interactive device type. Developer affiliation information can be the organization or department to which the developer belongs. The developer can be a developer using the application management platform for application development. The interactive device can also collect the query context. Then, the intelligent agent extension device determines the strategy configuration information of the target intelligent agent based on the intelligent agent identifier and conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in at least one business intelligent agent for a specific business. The strategy configuration information includes at least one of the following: auxiliary prompts, truncation strategies, model identifiers for generated models, knowledge base identifiers, retrieval strategies, recall strategies, tool identifiers, risk control strategies, or role types. The generated model can be a unimodal language model or a multimodal model. The input to the multimodal model can include various types, such as any one or any combination of text, images, audio, or video. Similarly, the output of the multimodal model can also include various types, such as any one or any combination of text, images, audio, or video. Hints and auxiliary information are used to assist in generating the above-mentioned generative model. The target agent expands the query context according to the policy configuration information. The expanded context includes at least one of the knowledge base retrieval results or tool call results. The target agent can assemble the context and expanded context according to the policy configuration information to obtain hints. The target agent uses the hints as input to the generative model for reasoning to obtain an answer.

[0009] When faced with diverse and customized needs, this method utilizes a pluggable configuration mechanism to dynamically select and adjust prompts, prompt templates, truncation strategies, generative models (such as language models or multimodal models), knowledge bases, retrieval strategies, recall strategies, tools, risk control strategies, role types, etc., avoiding the lack of flexibility caused by fixed configurations. It effectively manages and optimizes a large number of strategy combinations generated by different business requirements, programming languages, types of interactive devices, and organizations / departments. This method allows for rapid adjustment of routing strategies based on customized needs, effectively addressing the combinatorial explosion problem. Based on the above solution, it is possible to achieve business customization requirements at runtime by only modifying the strategy configuration information (extended strategies) without modifying the business code.

[0010] In some possible implementations, the agent extension device can determine the policy configuration information of the target agent based on the agent identifier and conditions through a routing strategy. This method, relying on the routing mechanism of the agent extension device, enables flexible configuration of policies such as prompts, knowledge, tools, and models for the agent.

[0011] In some possible implementations, the application management platform can configure metadata for at least one business intelligence agent. The metadata includes the routing strategy of the business intelligence agent. The routing strategy includes a strategy identifier, routing input information, and routing output information. The routing input information is generated based on at least one of the following: application identifier, intelligence agent identifier, task type, developer affiliation information, programming language, programming framework, or type of interaction device. The routing output information includes at least one of the following: model identifier, prompt identifier, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier.

[0012] In this way, when developing intelligent agents or applications based on intelligent agents, it is possible to customize the configuration of prompts, knowledge, tools, models, etc., so that when the intelligent agent or application based on intelligent agents runs, it can give full play to the optimization capabilities of customized knowledge base and customized tool calls.

[0013] In some possible implementations, the application management platform can receive at least one of the following from the user's configuration interface for at least one business intelligence agent: agent identifier, task type, developer affiliation information, programming language, programming framework, or interaction device type, as well as at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier. Then, the application management platform can generate a routing strategy based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework, or interaction device type, as well as at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier, and instance identifier.

[0014] This method supports generating routing policies based on user-configured metadata, enabling flexible policy configuration through routing policies during the runtime of subsequent intelligent agents or agent-based applications to meet diverse user needs or complex business environments.

[0015] In some possible implementations, the application management platform may also receive at least one of the following user-configured business thesaurus, retrieval strategy, truncation strategy, recall strategy, or risk control strategy. Accordingly, the metadata also includes at least one of the following business thesaurus, retrieval strategy, truncation strategy, recall strategy, or risk control strategy.

[0016] This method configures rich metadata during the agent development or agent-based application development phase, thereby enabling agents or agent-based applications to run efficiently and flexibly configure policies through routing strategies when facing diverse user needs and complex business scenarios.

[0017] In some possible implementations, the auxiliary information for the prompt includes one or more of the following: a prompt identifier, a prompt template, or parameter names within the prompt. In this method, the application management platform provides multiple ways to generate prompts, ensuring high availability.

[0018] In some possible implementations, the prompting information includes the parameter name in the prompt. The application management platform also includes a front-end extension device, and the policy configuration information includes at least one of the extension identifier or instance identifier of the front-end extension device. Accordingly, the interaction device receives at least one of the extension identifier or instance identifier and the parameter name sent by the agent extension device, and the interaction device sends at least one of the extension identifier or instance identifier and the parameter name to the front-end extension device. Then, the interaction device receives the parameter value corresponding to the parameter name returned by the front-end extension device. The query context includes the parameter value. In this way, the context can be collected by the front-end extension device to assemble the prompt, and the acceptance rate of the answer generated by the generative model can be improved by combining the prompt assembled from the context collected by the front-end extension device.

[0019] In some possible implementations, the front-end extension device processes the answer to obtain a processed response. This processing could involve post-processing such as fixing coding style issues, compilation problems, runtime issues, and formatting errors, before the interactive device displays the processed answer to the user. This can further improve the acceptance rate of the answer.

[0020] In some possible implementations, the prompting information includes prompt identifiers. The target agent can obtain a prompt template from the prompt word center based on the prompt identifiers in the policy configuration information. Then, the target agent can fill in the prompt template based on the context and extended context to obtain the prompt. This method supports obtaining prompt templates that match the query from the prompt word center and assembling prompts based on these templates, which can improve the quality of prompts and thus increase the acceptance rate of the answer.

[0021] Secondly, this application provides an application management platform. The application management platform includes an interactive device, an intelligent agent extension device, and at least one business intelligent agent for a given service.

[0022] The interactive device is used to obtain a query input by the user, and to obtain an agent identifier and conditions, wherein the conditions include at least one of the following: task type, developer affiliation information, programming language, programming framework, or type of the interactive device.

[0023] The interactive device is also used to collect the context of the query;

[0024] The intelligent agent extension device is used to determine the strategy configuration information of the target intelligent agent according to the intelligent agent identifier and the conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in the business intelligent agents of the at least one business. The strategy configuration information includes at least one of the following: prompting assistance information, truncation strategy, model identifier of the generated model, knowledge base identifier, retrieval strategy, recall strategy, tool identifier, risk control strategy or role type. The prompting assistance information is used to assist in the generation of the generated model.

[0025] The target agent is configured to expand the context of the query according to the policy configuration information, wherein the expanded context includes at least one of knowledge base retrieval results or tool call results; assemble the context and the expanded context according to the policy configuration information to obtain the prompt; and input the prompt into the generative model for reasoning to obtain an answer.

[0026] In some possible implementations, the agent extension device is specifically used for:

[0027] Based on the agent identifier and the conditions, the policy configuration information of the target agent is determined through a routing strategy.

[0028] In some possible implementations, the platform also includes:

[0029] An intelligent agent development device is used to configure the metadata of the business intelligent agent for the at least one business. The metadata includes the routing strategy of the business intelligent agent. The routing strategy includes a strategy identifier, routing input information, and routing output information. The routing input information is generated based on at least one of the following: application identifier, intelligent agent identifier, task type, developer affiliation information, programming language, programming framework, or type of interaction device. The routing output information includes at least one of the following: model identifier, prompt identifier, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier.

[0030] In some possible implementations, the intelligent agent development apparatus is specifically used for:

[0031] The system receives at least one of the following: the agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier, configured by the user through the agent configuration interface for the at least one service.

[0032] The routing strategy is generated based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier, and instance identifier.

[0033] In some possible implementations, the agent development apparatus is also used for:

[0034] The system receives at least one of the user-configured business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy, and the metadata further includes at least one of the business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy.

[0035] In some possible implementations, the prompting assistance information includes any one or more of the following: prompt identifier, prompt template, or parameter name in the prompt.

[0036] In some possible implementations, the prompt assistance information includes the parameter name in the prompt, the application management platform further includes a front-end extension device, and the policy configuration information further includes at least one of the extension identifier or instance identifier of the front-end extension device;

[0037] The interaction device is further configured to receive at least one of the extension identifier or the instance identifier and the parameter name sent by the intelligent agent extension device; send at least one of the extension identifier or the instance identifier and the parameter name to the front-end extension device; and receive the parameter value corresponding to the parameter name returned by the front-end extension device.

[0038] The context of the query includes the parameter values.

[0039] In some possible implementations, the front-end extension device is also used for:

[0040] The answer is processed to obtain the processed answer;

[0041] The interactive device is also used for:

[0042] The processed answer is displayed to the user.

[0043] In some possible implementations, the prompting assistance information includes a prompt identifier;

[0044] The target intelligent agent is specifically used for:

[0045] Based on the prompt identifier in the strategy configuration information, obtain the prompt template from the prompt word center;

[0046] The prompt template is populated based on the context and the extended context to obtain the prompt.

[0047] Thirdly, this application provides a computing device cluster. The computing device cluster includes at least one computing device, and the at least one computing device includes at least one processor and at least one memory. The at least one processor and the at least one memory communicate with each other. The at least one processor is used to execute instructions stored in the at least one memory to cause the computing device or the computing device cluster to perform the application management method as described in the first aspect or any implementation thereof.

[0048] Fourthly, this application provides a computer-readable storage medium storing instructions that instruct a computing device or a cluster of computing devices to execute the application management method described in the first aspect or any implementation thereof.

[0049] Fifthly, this application provides a computer program product containing instructions that, when run on a computing device or a cluster of computing devices, causes the computing device or cluster of computing devices to execute the application management method described in the first aspect or any implementation thereof.

[0050] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods. Attached Figure Description

[0051] To more clearly illustrate the technical methods of this application, the accompanying drawings used will be briefly described below.

[0052] Figure 1 A schematic diagram illustrating a combination of customized requirements provided in this application;

[0053] Figure 2 A schematic diagram of the architecture of an application management platform provided in this application;

[0054] Figure 3 A flowchart of an application management method provided in this application;

[0055] Figure 4 A flowchart illustrating a metadata configuration method for a business intelligence agent provided in this application;

[0056] Figure 5 A schematic diagram of an intelligent agent configuration interface provided in this application;

[0057] Figure 6 A schematic diagram of the structure of a computing device provided in this application;

[0058] Figure 7This application provides a schematic diagram of the structure of a computing device cluster;

[0059] Figure 8 This application provides a schematic diagram of another computing device cluster structure.

[0060] Figure 9 This is a schematic diagram of another computing device cluster provided in this application. Detailed Implementation

[0061] The terms "first" and "second" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include one or more of that feature.

[0062] First, some technical terms involved in the embodiments of this application will be introduced.

[0063] Artificial intelligence (AI) is the ability to correctly interpret external data, learn knowledge from that data, and use that knowledge to achieve specific goals and tasks. AI research has evolved from focusing on "reasoning" to focusing on "knowledge," and then to focusing on "learning." Currently, an important branch of AI is machine learning (ML). Machine learning is a class of algorithms that automatically analyze data to obtain patterns and use those patterns to predict unknown data.

[0064] Natural Language Processing (NLP) is a branch of artificial intelligence and linguistics, primarily encompassing natural language cognition, understanding, and generation. NLP can be implemented using language models (LMs). Language models can be further categorized by parameter size into small language models and large language models (LLMs). For ease of description, this application uses an LLM as an example. An LLM is a language model constructed from deep neural networks containing hundreds of billions of weights, which can be trained using self-supervised learning methods on a large amount of unlabeled text.

[0065] In recent years, LLM has made significant progress with the emergence of pre-trained Transformer architectures. As the parameter scale of LLMs has expanded to tens of billions and begun to exhibit early signs of artificial general intelligence, LLM applications have rapidly expanded beyond text processing to areas such as software development and testing, image generation, and video generation. Taking software development and testing as an example, the software engineering community applies large-scale code generation models to various downstream tasks to assist in software development and testing. Some software tools also integrate multiple downstream tasks, thus covering application scenarios such as clone detection, defect detection, code understanding, code repair, code translation, program synthesis, and code summarization. These advancements not only demonstrate the powerful capabilities of LLM in code processing but also reveal its enormous potential in promoting software engineering automation and improving development efficiency.

[0066] Despite the enormous potential of large language models in code generation, their adoption rate in real-world projects still faces numerous challenges. Many LLM-based intelligent programming assistants have a code adoption rate (or code acceptance rate) of less than 50%, with some even reaching around 30%. In other words, approximately 70% of their code generation suggestions are not adopted.

[0067] The inventors discovered through research that LLM-based intelligent programming assistants are mainly applicable to specific scenarios, and their adaptability and flexibility need to be improved for complex projects and changing requirements. To enable LLM-based applications such as intelligent programming assistants to meet customized needs, the industry has proposed solutions such as Supervised Fine-Tuning (SFT), Retrieval-augmented generation (RAG), Everything of Thoughts (XoT), Iteration optimization, context engineering, and tool invocation.

[0068] SFT improves model performance in specific domains by using supervised fine-tuning of pre-trained models with vertical domain data. RAG enhances generation by customizing a knowledge base and incorporating reference code during the generation process. XoT breaks down complex problems into easily solvable sub-problems using Chain of Thought (CoT), Tree of Thought (ToT), or Graph of Thought (GoT), improving the accuracy and interpretability of generated code; it also provides a structured thought process, contributing to the logical consistency and coherence of code generation. Iterative optimization techniques improve code quality through multi-turn conversational generation and continuous improvement of the generated results. Context engineering enhances the relevance and accuracy of generated code by strengthening the contextual information required during the generation process. Tool invocation technology enhances the ability to handle complex tasks by connecting to real-time business systems.

[0069] To address complex projects and evolving user needs, related technologies have introduced intelligent agents. An intelligent agent is a general-purpose problem solver. From a software engineering perspective, an intelligent agent can be a computer program based on a large language model, possessing planning and thinking abilities (also known as planning), memory capabilities, and the ability to use tool functions (or tool use), enabling it to autonomously complete a given task. Planning ability can involve breaking down large tasks into subtasks, such as using XoT to decompose large tasks into subtasks and planning the execution flow. Thinking ability can involve reflecting on and analyzing the task execution process to decide whether to continue or terminate the task upon completion. Memory capabilities can include short-term memory and long-term memory. Short-term memory refers to the context generated and temporarily stored during task execution, which is cleared after task completion. Long-term memory refers to information retained for a long period, such as external knowledge bases, typically stored and retrieved using vector databases. The ability to use tools refers to the intelligent agent being configured with application programming interfaces (APIs), such as calculators, search tools, code executors, and database query tools. Typically, the intelligent agent can act as the backend of a business system, collaborating with frontend interaction devices to implement the system's functionality.

[0070] For different business functions, it is usually necessary to develop corresponding business agents. For example, throughout the entire application development lifecycle, business agents can include code agents (CodeAgent) and test agents (TestAgent). CodeAgent can be used for code continuation / incomplete code handling, unit test (UT) generation, code inspection / fixing, code explanation / commenting, and code submission. TestAgent is used to assist in test design, test case inspection, test script generation, test execution, and evaluation.

[0071] When applying large language models to business agents, configuring strategies to meet customized needs is a key technical challenge. Related technologies often exhibit insufficient flexibility and poor adaptability when faced with diverse user requirements and complex business environments. Specifically, the diversity of different business agents, programming languages, integrated development environments (IDEs), and organizational / departmental structures leads to a combinatorial explosion of customized requirements.

[0072] See Figure 1The diagram illustrates a combination of customized requirements. Business agents can include ReqAgent, DesignAgent, CodeAgent, ChatAgent, CheckAgent, Merge Request Agent (MRAgent), TestAgent, BuildAgent, PipelineAgent, CloudAgent, or other agents (denoted as XxxAgent). Programming languages ​​can include Java, JavaScript / TypeScript (JS / TS), Python, C / C++, Go, and Structured Query Language (SQL). IDE types can include IDEs for different programming languages ​​or different types of IDEs developed by different vendors. Organizations / departments can include multiple first-level or second-level departments, etc. This results in a large number of combinations of customized requirements. Developers need to configure prompts, truncation strategies, large models, knowledge bases, retrieval / recall strategies, toolsets, and thesauruses (such as risk control thesaurus, persona thesaurus, and business thesaurus) separately for a large number of customized requirements, making it difficult to achieve efficient and flexible strategy configuration and management. Specifically, the risk control thesaurus includes preset words for risk control, the persona thesaurus includes preset words for the roles of the language model (e.g., in application development, the language model can represent product manager, architect, developer, tester, operations personnel, or Java development expert, cloud service operations expert, etc.), and the persona thesaurus can include these role names. The business thesaurus includes preset words for business operations. Related technologies struggle to efficiently and flexibly configure and manage strategies for these combinations of customized requirements.

[0073] In view of this, this application provides an application management method. This method can be implemented using an application management platform. The application management platform can be a software system, which can be a standalone software system or a software system integrated with other software. The software system can be provided to customers as a software package for self-deployment, or it can be provided to users as a cloud service, with users subscribing to the cloud service to use the corresponding capabilities or functions. The aforementioned software system can be deployed in a computing device cluster, where the computing device cluster executes the program code of the software system to perform the application management method of this application.

[0074] The application management platform may include an interactive device, an intelligent agent extension device, and at least one business intelligent agent for a specific business. The interactive device can be a device that enables human-computer interaction. For example, the interactive device can be an IDE or a browser, where the browser can have built-in interactive modules related to the aforementioned business, or the browser can implement human-computer interaction for the aforementioned business through the installation of plugins. The intelligent agent extension device is used to determine the extension strategy of the business intelligent agent to meet diverse customization needs. The intelligent agent extension device can be a standalone device or an extension or plugin for the business intelligent agent. The business intelligent agent may include business intelligent agents for at least one business involved in the software development lifecycle. For example, the business intelligent agent may include any one or any combination of ReqAgent, DesignAgent, CodeAgent, ChatAgent, CheckAgent, MRAgent, TestAgent, BuildAgent, PipelineAgent, and CloudAgent.

[0075] The interactive device can acquire user-input queries, as well as agent identifiers (IDs) and conditions (or feature profiles). Conditions may include at least one of the following: task type, developer affiliation information, programming language, programming framework, or interactive device type. The interactive device collects the query context. This context can be within a project, including intra-file context or cross-file context within a project. The agent extension device determines the strategy configuration information of the target agent based on the agent identifier and conditions. The target agent is the agent corresponding to the agent identifier in at least one business agent. The strategy configuration information includes at least one of the following: prompt assistance information, truncation strategy, model identifier of the generated model, knowledge base identifier, retrieval strategy, recall strategy, tool identifier, risk control strategy, or role type. The prompt assistance information assists in generating the aforementioned generated model; for example, it may include any one or more of the following: prompt identifier, prompt template, or parameter names in the prompt. The role type indicates the persona or role of the person creating the language model, such as product manager, architect, etc. The target agent can expand the query context based on policy configuration information. This expanded context includes at least one of the following: knowledge base retrieval results (or simply retrieval results) or tool call results. The target agent assembles the context and expanded context based on the policy configuration information to obtain a prompt. The target agent then uses the prompt as input to generate a model for reasoning to obtain an answer.

[0076] When faced with diverse and customized needs, this method utilizes a pluggable configuration mechanism to dynamically select and adjust prompts, prompt templates, truncation strategies, generation models, knowledge bases, retrieval strategies, recall strategies, tools, risk control strategies, role types, etc., avoiding the lack of flexibility caused by fixed configurations. It effectively manages and optimizes a large number of strategy combinations generated by different business requirements, programming languages, IDE types, and organizations / departments. This method allows for rapid adjustment of routing strategies based on customized needs, effectively addressing the combinatorial explosion problem. The application management method provided in this application saves time and costs for teams collaborating on massive customized needs, improving overall development efficiency and deployment schedules.

[0077] To make the technical solution of this application clearer and easier to understand, the system architecture of the application management platform provided in this application is described below with reference to the accompanying drawings.

[0078] See Figure 2 The diagram shows an architecture of an application management platform 20. The application management platform 20 includes an interaction device 202, an intelligent agent extension device 204, and a business intelligent agent 206 for at least one business function. Further, the application management platform 20 may also include a front-end extension device 208. The components of the application management platform 20 are described in detail below.

[0079] The interaction device 202 is used to obtain the user's input query (denoted as query), as well as the agent identifier (agent_id) and conditions. The conditions can also be called feature profiles. Conditions may include at least one of the following: task type, developer affiliation information, programming language, programming framework, or the type of the interaction device. The task type can be used to distinguish different tasks; for example, task types may include code completion and code testing. Developer affiliation information may include the organization / department to which the developer belongs. The programming language can be the language currently used in development, such as C, Java, or JavaScript. A programming framework is an abstract toolset that provides general functionality, designed to help developers build and maintain applications more easily. A framework typically includes a set of predefined classes, modules, and functions that developers can extend and customize. The goal of a framework is to provide a structured development approach, reduce repetitive work, and accelerate the application development process. The interaction device 202 is also used to collect the context of the query. For example, the interactive device 202 can capture the current position of the cursor and obtain the code snippets before and after the current position. Furthermore, the interactive device 202 can also obtain the cross-file context based on the import of the code file where the current position is located. The cross-file context includes the context in other files within the project.

[0080] The intelligent agent extension device 204 is used to determine the policy configuration information of the target intelligent agent based on the intelligent agent identifier and conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in at least one business intelligent agent. The policy configuration information includes at least one of the following: prompt identifier, truncation policy, model identifier, knowledge base identifier, retrieval policy, recall policy, tool identifier, instance identifier of the front-end extension device, risk control policy, or role type. When the application management platform includes the front-end extension device 208, the policy configuration information also includes the extension identifier and instance identifier of the front-end extension device 208. It should be noted that when the front-end extension device 208 starts an instance, the policy configuration information may not include the aforementioned instance identifier.

[0081] The business intelligence agent 206 is used to expand the query context according to the strategy configuration information. The expanded context includes at least one of the knowledge base retrieval results or tool call results. Further, the strategy configuration information obtained by the business intelligence agent 206 from the intelligence agent extension device 204 may also include a processing strategy, which may include the pre-processing strategy and / or post-processing strategy of the front-end interaction device 202. When the application management platform 20 includes a front-end extension device 208, the processing strategy may also include the pre-processing strategy and / or post-processing strategy of the front-end extension device 208. The pre-processing strategy of the front-end extension device 208 can be used to pre-process the context, and the post-processing strategy can be used to post-process the answer generated by the generative model. Pre-processing may include, but is not limited to, data cleaning, and post-processing may include, but is not limited to, fixing coding standard issues, fixing compilation issues, fixing runtime issues, and fixing formatting. The business intelligence agent 206 is also used to assemble the context and expanded context according to the strategy configuration information to obtain prompts, input the prompts into the generative model for reasoning, and obtain an answer. The generative model may be a language model, which is typically a single-model model. For example, a language model can be a large language model (LLM), including but not limited to bidirectional encoder representations from transformers (BERT) and generative pre-trained transformers (GPT). Figure 2 The example of a generative model as a unimodal language model illustrates that, in practical applications, generative models can also be multimodal models. The input to a multimodal model can include various types, such as any one or any combination of text, images, audio, or video. Similarly, the output of a multimodal model can also include various types, such as any one or any combination of text, images, audio, or video. It should be noted that, with the evolution of technology, multimodal models can support even more types or more modalities of input and output; this application does not limit this.

[0082] like Figure 2 As shown in step ①, the interactive device 202 can receive user-input queries and obtain the agent identifier (agent_id) and conditions. Conditions may include at least one of the following: task type of the downstream task, organization / department / employee number, programming language, programming framework, or IDE type. The interactive device 202 can present a question-and-answer interface to the user and receive the user query input through the interface. The agent_id and conditions can be user input or actively obtained by the interactive device 202; for example, the interactive device 202 can actively identify the agent_id and conditions including programming language and programming framework. Furthermore, the interactive device 202 can also collect the context of the user query through a context collection module. The context collected by the interactive device 202 can be a common context. The common context can be a context applicable to different scenarios.

[0083] Furthermore, the application management platform 20 also includes a front-end extension device 208. The front-end extension device 208 is used to extend the front-end, including processing the context or post-processing the response. The front-end extension device 208 can be in the form of an extension or a plugin. The front-end extension device 208 can be identified by an extension identifier (extension_id) or a plugin identifier (plugin_id). For ease of description, the following example uses extension_id. The front-end extension device 208 can include one or more instances, where an instance can be a code segment implementing a class or method of the front-end extension device 208. Figure 2 As shown in steps ② and ③, the interaction device 202 can send a condition to the intelligent agent extension device 204. The intelligent agent extension device 204 can then return to the interaction device 202 the extension identifier (extension_id) of the front-end extension device 208, the instance identifier (instance_id) of the front-end extension device 208, or the parameter name from the prompt, based on the condition. The parameter name in the prompt is also called the key name in the prompt, denoted as prompt_key. It should be noted that... Figure 2Using `prompt_key` as just one example of prompt assistance information, in practical applications, the agent extension device 204 can also return other prompt assistance information, such as a prompt identifier (`prompt_id`) or a prompt template (`prompt template`). That is, the prompt assistance information can be any one or more of the parameter names, prompt identifiers, or prompt templates in the prompt. The prompt template returned by the agent extension device 204 can be a pre-defined prompt template of the agent extension device 204. A prompt template is a predefined template used to generate prompts. A prompt template is essentially a parameterizable model input. The prompt identifier returned by the agent extension device 204 can be an identifier of a pre-defined prompt of the agent extension device 204, through which the pre-defined prompt can be used directly.

[0084] Specifically, the agent extension device 204 can obtain the aforementioned extension_id through extended routing according to a pre-configured routing policy. Similarly, the agent extension device 204 can obtain the aforementioned instance_id through instance routing according to a pre-configured routing policy. The agent extension device 204 can obtain the aforementioned prompt_key through prompt routing according to a pre-configured routing policy.

[0085] The interactive device 202 can request the front-end extension device 208 to return the parameter value corresponding to the parameter name, such as prompt_value. Figure 2 As shown in steps ④ and ⑤, the interactive device 202 can send an extension identifier, an instance identifier, or a parameter name from a prompt to the front-end extension device 208. The front-end extension device 208 retrieves the corresponding parameter value based on the parameter name. This parameter value can also serve as the query context. The front-end extension device 208 can preprocess the context according to a preprocessing strategy to obtain the processed context. The processed context includes the processed parameter values.

[0086] like Figure 2As shown in steps ⑥, ⑦, and ⑧, the interaction device 202 can send the user query, agent_id, condition, and context to the business intelligence agent (such as business intelligence agent 206) corresponding to the intelligence agent identifier. This context can be the context collected by the interaction device 202; in some examples, the context can also include the context collected by the front-end extension device 208. Further, the context collected by the interaction device 202 or the front-end extension device 208 can undergo preprocessing to obtain a processed context. Then, the business intelligence agent can send the agent_id and condition to the intelligence agent extension device 204, which determines the policy configuration information of the business intelligence agent 206 based on the agent_id and condition. In this example, the policy configuration information includes at least one of the following: prompt_id, model_id, tool_id, risk control policy, and role type.

[0087] See also Figure 2 Middle steps 9, 10 The business agent 206 can send a prompt_id to the prompt word center, which can then return the corresponding prompt template to the target agent. The business agent 206 also sends search terms, a knowledge base identifier (knowledgebase_id), and a search strategy to the knowledge base. The search terms can be obtained from the user's input query (e.g., user query). Specifically, the business agent 206 can expand or abbreviate the user query and then perform keyword extraction to obtain the search terms. The knowledge base identifier identifies the knowledge base to be searched; this knowledge base is typically highly relevant to the user query. The knowledge base identifier can be the knowledge base name, or simply the database name. Furthermore, the business agent 206 can also send a knowledge base table identifier (knowledgeasset_id) to narrow the search scope. The knowledge base table identifier identifies the tables within the knowledge base; for example, it can be the table name, or simply the table name. The knowledge base can then search for the search terms according to the search strategy and return the search results. Similarly, business intelligence agent 206 can send a `tool_id` to the toolset to invoke the corresponding tool and obtain the tool invocation result. The toolset can include local tools, OpenAPI, or code interpreters. It should be noted that the above are just some examples of tools in a toolset; in actual applications, the toolset can also include other tools, including but not limited to search engines, SQL queries, and web crawlers. Figure 2 intermediate steps As shown, the business intelligence agent 206 can also assemble the context, search results, tool call results, and prompt template into a complete prompt, send the prompt and model identifier `model_id` to the inference service, and the inference service can call the corresponding language model to perform inference based on `model_id`, and then return the answer. Figure 2 Using a language model as an example of LLM, when performing inference using LLM, the prompts can first be encoded or vectorized to obtain a prompt vector. The inference service can first return a response to the target agent, and the target agent can then return a response to the interaction device 202.

[0088] It should be noted that, as Figure 2 intermediate steps As shown, when the application management platform 20 includes a front-end extension device 208, the interaction device 202 can also send a response to the front-end extension device 208. The front-end extension device 208 processes the response according to a post-processing strategy, obtains the processed response, and returns the processed response to the interaction device 202. The interaction device 202 can then present the response to the user. It should be noted that when the interaction device sends a response to the front-end extension device 208, it also sends the extension identifier (extension_id) and the instance identifier (instance_id) of the front-end extension device 208, so that the instance corresponding to the instance identifier can perform post-processing on the response.

[0089] exist Figure 2 In the example, if the interactive device 202 obtains the prompt or prompt template through the front-end extension device 208, the step of obtaining the prompt template from the prompt word center can be omitted.

[0090] based on Figure 2 The application management platform 20 shown in this application provides an application management method. The application management method of this application will be described below with reference to the accompanying drawings.

[0091] See Figure 3 The flowchart shown illustrates an application management method applied to an application management platform 20. The application management platform 20 includes an interaction device 202, an intelligent agent extension device 204, and a business intelligent agent 206 for at least one business. The method includes:

[0092] S302, the interactive device 202 obtains the query input by the user.

[0093] Specifically, the interactive device 202 can present a question-and-answer interface to the user, receiving queries (or questions) entered by the user through the interface. These queries can be based on natural language or programming languages. The interface can include a question-and-answer area or an editing area. Users can enter queries based on natural language in the question-and-answer area, or queries based on SQL, eXtensible Markup Language (XML), YAML, JavaScript Object Notation (JSON), MD, JAVA, or Python in the editing area. Queries can be applied throughout the entire application development lifecycle. For example, during the requirements gathering phase, queries can be used for requirements management, defect management, and project management; during the design phase, queries can be used for view design, design for X (DFX), API design, and data table design; and during the code development phase, queries can be used for code completion / writing, unit testing (UT) generation, code inspection / fixing, code explanation / commenting, and code submission. Furthermore, during development, queries can also be used to access research and development knowledge. During the code review phase, queries can also be used for code inspection. Similarly, during the testing phase, queries are used to assist in test design, test case review, test script generation, or test execution and evaluation. For API design, development, and operations, queries are also used for managing cloud service resources, application cloud deployment design, code / script development, and cloud service operation / maintenance.

[0094] It should be noted that this application can also be applied to businesses other than application development (such as DevOps), and this application does not limit it in this regard. For example, this application can also be used for automated office business, where the intelligent agent extension device can extend the strategies of the business intelligent agents in automated office business to meet diverse user needs or complex business scenarios.

[0095] S304, Interactive device 202 acquires agent identifier and conditions.

[0096] The interactive device 202 can receive agent IDs and conditions input by the user through a question-and-answer interface, or automatically identify agent IDs and conditions. Conditions may include at least one of the following: task type, developer affiliation information, programming language, programming framework, or the type of the interactive device. Developer affiliation information may be the developer's organization, department, or employee ID. The programming language refers to the language used in project development, such as C, Java, or Python. A programming framework is a collection of reusable software components used in the programming process. Programming frameworks may include AI frameworks or development frameworks, where AI frameworks may include, but are not limited to, various open-source deep learning (DL) frameworks, and development frameworks may include development frameworks for various programming languages. The type of interactive device may be an IDE or a browser plugin.

[0097] The above-mentioned S302 and S304 can be executed in parallel or sequentially according to a set order; this application does not impose any restrictions on this. By providing the above detailed input, customized configuration can be achieved according to specific needs.

[0098] S306, Interactive device 202 collects the context of the query.

[0099] For queries, the interactive device 202 can collect the cursor position when the query is input and collect the query context based on the cursor position. Specifically, during the requirements gathering and design phases, the interactive device 202 can collect the cursor position through a dialog window and collect the query context within that window. In this case, the query context can be the content of the first m rounds of dialogue, including the queries and responses from the first m rounds. During the code development and testing phases, the interactive device 202 can collect the cursor position within the code file and collect the query context within that file. In this case, the query context can include the preceding and following k lines of code in the code file. Here, m and k can be positive integers.

[0100] In some possible implementations, the query context can also include business-related context. For example, if the query is entered in a business Q&A area, the interactive device 202 can also obtain business-related context from the business window. Business-related context can include files opened in the business window and content loaded in the business window. For example, in chart-based applications, business-related context can include open data tables and graphs loaded in the data tables.

[0101] S308, the agent extension device 204 determines the strategy configuration information of the target agent based on the agent identifier and conditions.

[0102] The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in at least one business intelligent agent. In the code development and testing domain, the business intelligent agent may include any one or any combination of ReqAgent, DesignAgent, CodeAgent, ChatAgent, CheckAgent, MRAgent, TestAgent, BuildAgent, PipelineAgent, or CloudAgent. The intelligent agent extension device 204 can match the target intelligent agent from the business intelligent agents of at least one business based on the intelligent agent identifier, and determine the policy configuration information of the target intelligent agent according to conditions. The policy configuration information includes at least one of the following: prompt assistance information, truncation policy, model identifier, knowledge base identifier, retrieval policy, recall policy, tool identifier, risk control policy, or role type.

[0103] Hints and cues are used to assist in generating hints for the language model. In some examples, hints and cues may include hint identifiers, hint templates, or parameter names within the hints. Truncation strategies refer to strategies that truncate hints to meet length requirements. Model identifiers are used to uniquely identify the generated model and are typically a universally unique identifier (UUID). In some examples, the model identifier can also be the model name, and further, it may include the model version number. Knowledge base identifiers are used to uniquely identify the knowledge base. Retrieval strategies refer to strategies for retrieving information from the knowledge base, including but not limited to vector search and keyword search. Recall strategies refer to algorithmic strategies for generating recommendation candidate sets through big data computation or deep training. Tool identifiers are used to uniquely identify tools, which can be services or APIs, etc. Risk control strategies refer to strategies that trigger and execute risk control. Role types indicate the roles or personas of the language model; role types can be determined from a persona lexicon.

[0104] In its specific implementation, the agent extension device 204 supports configuring routing policies. Routing policies can be in the form of rules, also known as routing rules, which can define different routing inputs and corresponding routing outputs. Based on this, the agent extension device 204 can determine the policy configuration information of the target agent through the routing policy, according to the agent identifier and conditions. Specifically, the agent extension device 204 can first determine the target agent based on the agent identifier, and then determine the policy configuration information of the target agent based on the conditions and the routing policy. For example, the agent extension device 204 can match the conditions as routing inputs with the routing policy to determine the policy configuration information.

[0105] Taking prompting routing and model routing as examples, the application management platform 20 can provide different prompts and language models for different services. The intelligent agent extension device 204 can determine the prompt identifier and the model identifier of the language model corresponding to the condition based on the task type or intelligent agent identifier in the condition through the routing strategy.

[0106] The strategy configuration information may further include pre-processing strategies for handling the context and post-processing strategies for handling the response statements. Specifically, the application management platform 20 may also include a front-end extension device 208. For example, the agent extension device 204 may determine the parameter name in the prompt corresponding to the condition, such as prompt_key, based on the task type, programming language, and programming framework in the condition, and return prompt_key to the interaction device 202. When the agent extension device 204 returns prompt_key, it may also return the extension identifier and instance identifier of the front-end extension device 208, so that the interaction device 202 can request the parameter value corresponding to prompt_key from the corresponding front-end extension device 208 or an instance of the front-end extension device 208. Accordingly, the front-end extension device 208 may obtain the corresponding parameter value, such as prompt_value, based on the parameter name. This parameter value can serve as the context. Furthermore, the front-end extension device 208 may also process the context to obtain the processed context. This method returns strategy configuration information, including pre-processing strategies and post-processing strategies, through an advanced strategy extension mechanism. By configuring this strategy, business intelligence agents can execute tasks more accurately and adapt to different application scenarios.

[0107] S309, The target agent receives the strategy configuration information returned by the agent extension device 204.

[0108] Specifically, the target agent can invoke the agent extension device 204 to obtain its return value. The return value of the agent extension device 204 may include the aforementioned policy configuration information.

[0109] S310, The target agent expands the context of the query based on the policy configuration information.

[0110] The extended context includes at least one of the knowledge base retrieval results or tool call results. The process of extending the query context is explained in detail below.

[0111] In some possible implementations, the target agent can perform searches in the corresponding knowledge base based on the retrieval strategy and knowledge base identifier (e.g., knowledgebase_id) in the strategy configuration information to obtain search results. Specifically, the target agent can extract search terms from the query (e.g., the query statement) and, according to the retrieval strategy, retrieve these search terms from the knowledge base corresponding to the knowledge base identifier to obtain search results. In this way, a customized knowledge selection strategy can be used to ensure that the retrieved knowledge is highly relevant and contributes to the generation of high-quality content.

[0112] In other possible implementations, the target agent can send a tool invocation request to the tool corresponding to the tool identifier in the toolset based on the tool identifier in the policy configuration information, and then obtain the tool invocation result returned by the tool. The tool invocation request can include a query and a tool identifier. In this method, the target agent invokes relevant tools in the toolset according to the policy configuration information to enhance task processing capabilities. The toolset returns the tool invocation result, providing the agent with necessary functional support. This method provides personalized tool selection capabilities, improving processing efficiency and accuracy in complex tasks.

[0113] It should be noted that when the strategy configuration information includes a prompt identifier, the target agent can also request a prompt template matching that prompt identifier from the prompt word center. The prompt word center provides guidance for the generation process based on the returned prompt template. This method, by using predefined prompt templates, ensures that the generated content is consistent with expectations in both format and semantics.

[0114] In some possible implementations, when searching the knowledge base, the target agent can first determine the similarity between the user's query and historical queries. The target agent can incorporate the query context when determining similarity to improve the accuracy of the similarity calculation. When the similarity between the user's query and historical queries reaches a threshold, the target agent can also return the answers to the historical queries without needing to execute subsequent steps S312 and S314.

[0115] In some other possible implementations, when the user configures the agent's strategy to retrieve only or retrieve only without generating, for example, when the user enters a query, the strategy can be configured to retrieve only without generating, then the target agent can return the retrieval results without executing subsequent steps S312 and S314.

[0116] S312, The target agent assembles the context and extended context based on the policy configuration information and receives a prompt.

[0117] Specifically, the target agent can fill the prompt template with the context and extended context respectively to obtain the prompt. When the policy configuration information includes a truncation policy, the target agent can also truncate the context and extended context according to the truncation policy to compress the prompt length and reduce the amount of inference computation. The truncation policy can include sequential truncation or weight-based truncation.

[0118] For example, when the truncation strategy is sequential truncation, the target agent can sequentially fill in the parameter values ​​corresponding to the parameter names in the prompt template. These parameter values ​​can be parameter values ​​from the context and extended context. When the number of tokens in the prompt after filling in the parameter values ​​reaches a threshold, the filling of the remaining parameter values ​​corresponding to the parameter names can be stopped.

[0119] For example, when the truncation strategy is weight-based, the target agent can sequentially fill in the parameter values ​​corresponding to the parameter names in the prompt template according to their weights. When the number of prompt tokens after filling in the parameter values ​​reaches a threshold, the filling of the remaining parameter values ​​can be stopped.

[0120] S314. The target agent will prompt the input language model to perform reasoning and obtain an answer.

[0121] The language model can be a generative model used to generate answers to queries, such as GPT or BERT, or a small-scale language model. The target agent can send prompts and model identifiers to the inference service. The inference service, based on the model identifiers, calls the corresponding language model and performs inference based on the input prompts to generate an answer. The target agent can also return the answer as a response. For example, the target agent can return the answer to itself, or the target agent can return the answer to the interaction device 202. This method, through the inference service and a dynamic context weight adjustment mechanism, can generate high-quality answers that meet the user's specific needs.

[0122] Furthermore, when the application management platform 20 includes a front-end extension device 208, the front-end extension device 208 can also send the answer to itself. The front-end extension device 208 can process the answer to obtain a processed answer. The interaction device 202 can also present the processed answer to the user. In this method, the front-end extension device 208 further optimizes and adjusts the answer to ensure that the returned answer meets the user's needs in terms of format, content, and applicability.

[0123] It should be noted that when this application is used in other scenarios, the generative model can also be an image generation model or a multimodal model such as an audio generation model or a video generation model.

[0124] Based on the above description, this application provides an application management method. This method relies on an agent extension device 204 to configure the strategies of business agents, such as prompt templates, knowledge bases, tools, and language models. During the operation of the business agent, based on the user-input query and the obtained agent identifier and conditions, the method determines the corresponding strategy configuration information of the target agent. This allows the target agent to perform knowledge base retrieval and tool invocation based on the strategy configuration information to extend the context. Prompts are then assembled based on the extended context, and the prompt input is used for reasoning in the corresponding language model. This allows for the fulfillment of complex and ever-changing customized needs without modifying the business code, only by modifying the strategy configuration information (extended strategy).

[0125] The above provides a detailed introduction to the application management methods during the runtime of business intelligence agents. The following section describes the strategy configuration during business intelligence agent development. The application management platform 20 supports strategy configuration for business intelligence agents through metadata. Specifically, the application management platform 20 can provide an intelligence agent configuration interface to users (such as administrators or application developers). This interface can be a graphical user interface (GUI) or a command user interface (CUI), and then receive the metadata configured by the user through this interface. The metadata can include the routing strategy of the business intelligence agent, which includes a strategy identifier, routing input information, and routing output information. The routing input information is generated based on at least one of the following: application identifier, intelligence agent identifier, task type, developer affiliation information, programming language, programming framework, or type of interaction device. The routing output information includes at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of the front-end extension device 208, and instance identifier. The prompt assistance information can include parameter names, prompt identifiers, or prompt templates. The following section uses prompt assistance information as an example of a prompt identifier.

[0126] The following section, with reference to the accompanying diagram, describes the method for configuring metadata for business intelligence agents.

[0127] See Figure 4 The flowchart shown illustrates a metadata configuration method for a business intelligence agent, which includes:

[0128] S402, Application Management Platform 20 receives business requests.

[0129] The business requirements section specifies the tasks that the business intelligence agent needs to complete. This information can include the task type, the agent ID (agent_id) of the business intelligence agent on which the task relies, and the application ID of the application built upon the business intelligence agent. The task type can include code continuation, unit testing (UT) generation, code debugging, or test code generation, test case checking, and optimization. The agent ID can include the agent name, such as CodeAgent, DesignAgent, or TestAgent. The application ID can be the application name.

[0130] Furthermore, business requirements may also include information about the developer who triggered the task, including the developer's affiliation, the programming language used, the programming framework, or the type of interactive device. Developer affiliation information includes the organization or department to which the developer belongs, or the developer's employee ID.

[0131] Specifically, the application management platform 20 can present an agent configuration interface to the user and receive business requirements input by the user through the agent configuration interface. The application management platform 20 may include an agent development device for configuring agents in development mode. This agent development device can be an agent development production line, which can present the agent configuration interface to the user and receive business requirements input by the user.

[0132] For ease of understanding, this application also provides an example.

[0133] See Figure 5The diagram illustrates a smart agent configuration interface 500. The smart agent configuration interface 500 includes a business requirement configuration component 502, which comprises a department configuration control 5022, an application configuration control 5024, a business smart agent configuration control 5026, and a task configuration control 5028. The department configuration control 5022 is used to configure developer affiliation information, such as the developer's department. The application configuration control 5024 is used to configure the application name, such as "xxx code development assistant" or "yyy test development assistant." The business smart agent configuration control 5026 is used to configure the smart agent identifier, such as "CodeAgent" or "TestAgent." The task configuration control 5028 is used to configure the task types of downstream tasks supported by the business smart agent. Among them, the downstream task types supported by CodeAgent can include, but are not limited to, code continuation, unit test generation, code debugging, code explanation, and code commenting. The downstream task types supported by TestAgent can include, but are not limited to, test keyword extraction, test code generation, test case inspection and optimization, anomaly detection and root cause analysis (RCA), malicious code detection, API testing, GUI self-testing, test point recommendation, test design to assist production, and requirement content formatting.

[0134] In some possible implementations, the business requirement configuration component 502 may further include a programming language configuration control 5027 and a client configuration control 5029. The programming language configuration control 5027 is used to configure the programming languages ​​supported by developers, such as Java, Python, Go, TS, C, C++, etc. The client configuration control 5029 is used to configure at least one of the following: client type (client_type), client name (client_name), or client version (client_version). Here, "client" refers to the interaction device 202; therefore, the client type can be an IDE or a browser. The client name can be the IDE name or the browser name. The client version can be the IDE version or the browser version. For example, the client version could be 2024.2.3. Figure 5 Taking the client configuration control 5029 as an example of configuring the IDE type, in other possible implementations of this application, the agent configuration interface can also support user configuration of client name and client version.

[0135] S404, Application Management Platform 20 receives the prompt icon configured by the user.

[0136] S406, Application Management Platform 20 receives the knowledge base identifier configured by the user.

[0137] S408, Application Management Platform 20 receives the tool identifier configured by the user.

[0138] S410 and Application Management Platform 20 receive the model identifier configured by the user.

[0139] Similar to the configuration of business requirements, the application management platform 20 can receive prompt identifiers, knowledge base identifiers, tool identifiers, or model identifiers configured by the user through the intelligent agent configuration interface 500. Furthermore, the application management platform 20 can also receive extension identifiers and instance identifiers of the front-end extension device 208 configured by the user through the intelligent agent configuration interface 500.

[0140] like Figure 5 As shown, the agent configuration interface 500 includes a strategy configuration component 504, which includes a template configuration control 5042, a knowledge configuration control 5044, a tool configuration control 5046, and a model configuration control 5048. The template configuration control 5042 is used to configure prompt templates, for example, allowing users to select a template identifier via a dropdown menu. Similarly, the knowledge configuration control 5044 allows users to configure a knowledge base identifier via a dropdown menu, the tool configuration control 5046 allows users to configure a tool identifier via a dropdown menu, and the model configuration control 5048 allows users to configure a model identifier via a dropdown menu.

[0141] In some possible implementations, the application management platform 20 may also receive at least one of the following user-configured business terms, retrieval strategies, truncation strategies, recall strategies, or risk control strategies. For example... Figure 5 As shown, the strategy configuration component 504 may further include at least one of the following: a vocabulary configuration control 5041, a retrieval strategy configuration control 5043, a truncation strategy configuration control 5045, a recall strategy configuration control 5047, and a risk control strategy configuration control 5049. The application management platform 20 may receive the business vocabulary, retrieval strategy, truncation strategy, recall strategy, or risk control strategy configured by the user through the above configuration controls. The retrieval strategy may include vector search or keyword search. The application management platform 20 may also receive the corresponding metadata of the business intelligence agent, which may include at least one of the business vocabulary, retrieval strategy, truncation strategy, recall strategy, or risk control strategy. In some examples, the application management platform 20 may also receive the re-ranking algorithm, pre-processing strategy, and post-processing strategy configured by the user; for example, the re-ranking algorithm may be configured as edit distance or word segmentation edit distance.

[0142] It should be noted that S404 to S410 can be executed in parallel or sequentially, and this application does not impose any restrictions on this.

[0143] S412, The application management platform 20 generates a routing strategy based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of model identifier, prompt identifier, knowledge base identifier, and tool identifier.

[0144] Specifically, the application management platform 20 can use at least one of the following as routing input information: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework, or type of interaction device; and use model identifier, prompt identifier, knowledge base identifier, or tool identifier as routing output information. It establishes a mapping relationship between the routing input and output information to generate a routing strategy. The routing strategy can include the mapping relationship between the routing input and output information. Furthermore, when generating the routing strategy, the application management platform 20 can also generate a strategy identifier, which can be the number or name of the routing strategy. Accordingly, the routing strategy also includes a strategy identifier.

[0145] It should be noted that when the application management platform 20 includes a front-end extension device 208, the application management platform 20 also receives the extension identifier and instance identifier of the front-end extension device 208 configured by the user. Specifically, the application management platform 20 can generate a routing policy based on at least one of the following: application identifier of the business intelligence agent, intelligence agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, knowledge base identifier, tool identifier, extension identifier of the front-end extension device 208, and instance identifier.

[0146] In this application, metadata, including routing strategies, can be stored in a configuration file. The configuration file can be in JavaScript Object Notation (JSON) format, or in XML or YAML format; this application does not impose any restrictions on this. The following example illustrates the metadata configuration file.

[0147] In this example, the metadata configuration file can be:

[0148]

[0149]

[0150] Here, `app_name` is the application name; `app_type` is the application type, including no-code applications, low-code applications (workflow), and full-code applications (agent). `tasks` is an array of downstream tasks. One agent can have multiple downstream tasks. Given an agent and downstream tasks, there can be multiple extension strategies. Each extension strategy typically corresponds to only one set of prompt templates, one or more models (e.g., 1-2 models), multiple knowledge bases, and multiple tools. `task_name` is the task name of the downstream task; `task_description` is the task description of the downstream task. `routing_rules` are routing rules (or routing strategies), specifically generated by configuring extension strategies for models, prompt templates, knowledge, tools, and front-end extensions based on conditions given an Agent and downstream tasks. `routing_rule_id` is the routing rule ID, typically a 32-bit Universally Unique Identifier (UUID).

[0151] `routing_key` represents the routing input information. When multiple routing inputs are included, they can be concatenated. These inputs can be conditions, and multiple conditions can be linked together to form a conditional concatenation string. In one example, application name, agent name, task, organization / department / product, programming language, IDE type, etc., can be concatenated with underscores to form the following conditional concatenation string: `Application Name_Agent Name_Task Name_Company / First-Level Department / Second-Level Department / Third-Level Department_Programming Language_IDE Type`. In some instances, the application management platform 20 may also skip concatenating the routing inputs and directly construct a routing strategy based on the multiple inputs. When routing according to the above strategy, multi-condition retrieval can be used directly.

[0152] routing_agent_model_id is for model routing; routing_agent_prompt_id is for Prompt routing; routing_agent_knowledge_ids is for knowledge routing; routing_tool_ids is for available toolsets; and routing_extension_id is for Extension routing.

[0153] In the example above, app_info can include the following information:

[0154]

[0155] In this context, `app_description` is the application description; `app_provider` is the application provider / department name; `app_creator` is the application creator / developer name; `app_version` is the application version, such as v1.1; `created_time` is the application creation time; `updated_time` is the application update time; `icon` is the application icon; `icon_background` is the application icon background color; `instruction` is the user-defined instruction; `opening_statement` is the opening statement, such as "Hello, I am your assistant"; `suggested_questions` are recommended (high-frequency) questions, such as "What are today's hot topics?"; `moderation_type` is the risk control type, including user-defined keywords, `openai_moderation`, third-party APIs, etc.; `sensitive_words` is a list of sensitive words, such as "violent" and "hijacking". In `input_moderation`, `enabled` indicates whether input risk control is enabled, and `preset_response` is the preset response when risk control is triggered, such as "Your input contains sensitive words, please re-enter". In `output_moderation`, `enabled` indicates whether output risk control is enabled, and `preset_response` is the preset response when risk control is triggered, such as "The model output contains sensitive words, please retry".

[0156] In some possible implementations, the application management platform 20 can also render a debugging interface to display the aforementioned metadata to the user. When a certain configuration of the metadata is not working properly, the user can also debug the metadata configuration through the debugging interface. The application management platform 20 can receive the debugged metadata from the user and perform metadata updates.

[0157] Based on the above description, the metadata configuration method for business intelligent agents in this application supports customized configuration of prompt templates, knowledge, tools, models, vocabularies, etc., during the development of business intelligent agents. This fully leverages the optimization capabilities of customized knowledge bases, customized context engineering, and tool invocation. Furthermore, this method achieves a balance between generality and customization, adhering to the open / closed principle at the architectural level. It is open to extensions but closed to modifications, allowing for independent release and evolution of main version capabilities and extension capabilities, facilitating problem location and resolution, and reducing development and maintenance costs. In addition, this method supports co-creation with customers, rapid exploration and support for diverse scenarios, and the ability to enable or disable extensions as needed for different customization requirements. It even supports integrating common components into the main plugin (such as interactive devices or intelligent agents) as built-in functionality. Moreover, this method supports agile development and deployment to support customer feedback and customized needs.

[0158] Based on the aforementioned application management method, this application also provides an application management platform 20. For example... Figure 2 As shown, the application management platform 20 includes an interaction device 202, an intelligent agent extension device 204, and a business intelligent agent 206 for at least one business.

[0159] Interactive device 202 is used to obtain a query input by the user, and to obtain an agent identifier and conditions, the conditions including at least one of task type, developer affiliation information, programming language, programming framework or type of interactive device;

[0160] The interactive device 202 is also used to collect the context of the query;

[0161] The intelligent agent extension device 204 is used to determine the strategy configuration information of the target intelligent agent based on the intelligent agent identifier and conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in the business intelligent agent 206 of at least one business. The strategy configuration information includes at least one of the following: prompting assistance information, truncation strategy, model identifier of the generated model, knowledge base identifier, retrieval strategy, recall strategy, tool identifier, risk control strategy or role type. The prompting assistance information is used to assist in generating the above-mentioned generated model.

[0162] A target agent is configured to expand the context of the query based on the policy configuration information, wherein the expanded context includes at least one of knowledge base retrieval results or tool call results; assemble the context and the expanded context based on the policy configuration information to obtain the prompt; and input the prompt into a generative model for reasoning to obtain an answer.

[0163] For example, the aforementioned interactive device 202, intelligent agent extension device 204, and business intelligent agent 206 can be implemented by software.

[0164] When implemented through software, the aforementioned interactive device 202, intelligent agent extension device 204, and business intelligent agent 206 can be applications running on computing devices. For example, the business intelligent agent 206 can be a computing engine running on a computing device. These applications can also be virtualized and provided to users as virtualization services. Virtualization services can include virtual machine (VM) services, bare metal server (BMS) services, or container services. Specifically, a VM service can be a service that uses virtualization technology to create a pool of virtual machine (VM) resources on multiple physical hosts to provide VMs for users to use on demand. A BMS service is a service that uses virtualization technology to create a pool of BMS resources on multiple physical hosts to provide BMS for users to use on demand. A container service is a service that uses virtualization technology to create a pool of container resources on multiple physical hosts to provide containers for users to use on demand. A VM is a simulated virtual computer, that is, a logical computer. A BMS is a scalable, high-performance computing service with computing performance indistinguishable from traditional physical machines and features secure physical isolation. Containers are a kernel virtualization technology that provides lightweight virtualization to isolate user space, processes, and resources. It should be understood that the VM service, BMS service, and container service mentioned above are merely specific examples. In practical applications, virtualization services can also include other lightweight or heavyweight virtualization services, which are not specifically limited here.

[0165] In some examples, the interaction device 202, the agent extension device 204, and the business agent 206 can also be implemented in hardware. For example, they can be hardware devices with corresponding capabilities. When implemented in hardware, the interaction device 202, the agent extension device 204, and the business agent 206 may include at least one computing device, such as a server. Alternatively, the interaction device 202, the agent extension device 204, and the business agent 206 can also be devices implemented using application-specific integrated circuits (ASICs) or programmable logic devices (PLDs). The PLD can be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), generic array logic (GAL), or any combination thereof.

[0166] In some possible implementations, the agent extension device 204 is specifically used for:

[0167] Based on the agent identifier and the conditions, the policy configuration information of the target agent is determined through a routing strategy.

[0168] In some possible implementations, the application management platform 20 also includes:

[0169] Intelligent Agent Development Device ( Figure 2 (Not shown in the image), used to configure the metadata of the business intelligence agent 206 for the at least one service, the metadata including the routing strategy of the business intelligence agent 206, the routing strategy including strategy identifier, routing input information and routing output information, the routing input information being generated based on at least one of application identifier, intelligence agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, the routing output information including at least one of model identifier, prompt identifier, knowledge base identifier, tool identifier, extension identifier of front-end extension device, instance identifier.

[0170] The intelligent agent development device can be implemented through software or hardware. When implemented through software, the intelligent agent development device can be an application running on a computing device. For example, the intelligent agent development device can be an intelligent agent development production line running on a computing device. The application can also be virtualized and provided to users as virtualization services such as VM services, BMS services, or container services. When implemented through hardware, the intelligent agent development device can include at least one computing device, such as a server. Alternatively, the intelligent agent development device can also be a device implemented using ASIC or PLD.

[0171] In some possible implementations, the agent development apparatus is specifically used for:

[0172] The intelligent agent 206 that receives at least one of the following: intelligent agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier, configured by the user through the intelligent agent configuration interface.

[0173] A routing strategy is generated based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier, and instance identifier of the business agent 206 of at least one business.

[0174] In some possible implementations, the agent development apparatus is also used for:

[0175] The system receives at least one of the following user-configured business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy, and the metadata further includes at least one of the following business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy.

[0176] In some possible implementations, the prompting information includes one or more of the following: prompt identifier, prompt template, or parameter name in the prompt.

[0177] In some possible implementations, the prompting assistance information includes the parameter name in the prompt, the application management platform 20 also includes the front-end extension device 208, and the policy configuration information also includes at least one of the extension identifier or instance identifier of the front-end extension device 208;

[0178] The interaction device 202 is also used to receive at least one of the extension identifier or instance identifier and the parameter name sent by the intelligent agent extension device 204; send at least one of the extension identifier or instance identifier and the parameter name to the front-end extension device 208; and receive the parameter value corresponding to the parameter name returned by the front-end extension device 208.

[0179] The context of the query includes the parameter values ​​mentioned above.

[0180] The front-end extension device 208 can be implemented in software or hardware. When implemented in software, the front-end extension device 208 can be an application running on a computing device. For example, the front-end extension device 208 can be an extension or plug-in of the interactive device 202. The application can also be virtualized and provided to users as virtualization services such as VM services, BMS services, or container services. When implemented in hardware, the front-end extension device 208 can include at least one computing device, such as a server. Alternatively, the front-end extension device 208 can also be a device implemented using an ASIC or a PLD.

[0181] In some possible implementations, the front-end extension device 208 is also used for:

[0182] Process the answer to obtain the processed answer;

[0183] The interactive device 202 is also used for:

[0184] Show the user the processed answer.

[0185] In some possible implementations, the prompting information includes prompt labels;

[0186] The target intelligent agent is specifically used for:

[0187] Based on the prompt identifier in the strategy configuration information, obtain the prompt template from the prompt word center;

[0188] The hint template is populated based on the context and extended context to obtain the hint.

[0189] This application also provides a computing device 600. For example... Figure 6 As shown, the computing device 600 includes a bus 602, a processor 604, a memory 606, and a communication interface 608. The processor 604, the memory 606, and the communication interface 608 communicate with each other via the bus 602. The computing device 600 can be a server or a terminal device. It should be understood that this application does not limit the number of processors and memories in the computing device 600.

[0190] Bus 602 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 6 The bus 602 may be represented by a single line, but this does not mean that there is only one bus or one type of bus. The bus 602 may include a path for transmitting information between various components of the computing device 600 (e.g., memory 606, processor 604, communication interface 608).

[0191] Processor 604 may include any one or more processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0192] Memory 606 may include volatile memory, such as random access memory (RAM). Memory 606 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD). Memory 606 stores executable program code, which processor 604 executes to implement the aforementioned application management method. Specifically, memory 606 stores instructions for the application management platform 20 to execute the application management method. In specific implementations, memory 606 may store instructions for implementing the functions of the interactive device 202, the intelligent agent extension device 204, and the business intelligent agent 206. Furthermore, memory 606 may also store instructions for implementing the functions of the front-end extension device 208 and the intelligent agent development device.

[0193] The communication interface 608 uses transceiver modules, such as, but not limited to, network interface cards and transceivers, to enable communication between the computing device 600 and other devices or communication networks.

[0194] This application also provides a computing device cluster. The computing device cluster includes at least one computing device. The computing device can be a server, such as a central server, an edge server, or a local server in a local data center. In some embodiments, the computing device can also be a terminal device such as a desktop computer, a laptop computer, or a smartphone.

[0195] like Figure 7 As shown, the computing device cluster includes at least one computing device 600. The memory 606 of one or more computing devices 600 in the computing device cluster may store instructions from the same application management platform 20 for executing application management methods.

[0196] In some possible implementations, one or more computing devices 600 in the computing device cluster can also be used to execute some of the instructions used by the application management platform 20 to execute application management methods. In other words, a combination of one or more computing devices 600 can jointly execute the instructions used by the application management platform 20 to execute application management methods.

[0197] It should be noted that the memory 606 in different computing devices 600 in the computing device cluster can store different instructions for executing some functions of the application management platform 20.

[0198] Figure 8One possible implementation is shown. For example... Figure 8 As shown, two computing devices 600A and 600B are connected via a communication interface 608. The memory in computing device 600A stores instructions for executing the functions of the interactive device 202. The memory in computing device 600B stores instructions for executing the functions of the intelligent agent extension device 204 and the business intelligent agent 206. In other words, the memory 606 of computing devices 600A and 600B jointly stores the instructions used by the application management platform 20 to execute application management methods.

[0199] Figure 8 The connection method between the computing device clusters shown can be based on the consideration that the application management method provided in this application requires a large amount of computing resources to extend the policies of the business intelligence agent 206 and to respond to user queries through the business intelligence agent 206. Therefore, it is considered that the functions implemented by the interaction device 202 are executed by the computing device 600A, and the functions implemented by the intelligence agent extension device 204 and the business intelligence agent 206 are executed by the computing device 600B. Furthermore, the application management platform 20 also includes a front-end extension device 208. Considering that the front-end extension device 208 usually extends the functions of the front-end interaction device 202, in order to reduce communication overhead, the functions implemented by the front-end extension device 208 can be executed by the computing device 600A.

[0200] It should be understood that Figure 8 The functions of computing device 600A shown can also be performed by multiple computing devices 600. Similarly, the functions of computing device 600B can also be performed by multiple computing devices 600.

[0201] In some possible implementations, one or more computing devices in a computing device cluster can be connected via a network. This network can be a wide area network (WAN) or a local area network (LAN), etc. Figure 9 One possible implementation is shown. For example... Figure 9 As shown, the two computing devices 600C and 600D are connected via a network. Specifically, they are connected to the network through the communication interfaces in each computing device. In this possible implementation, the memory 606 in computing device 600C stores instructions for executing the functions of the interaction device 202. Simultaneously, the memory 606 in computing device 600D stores instructions for executing the functions of the intelligent agent extension device 204 and the business intelligent agent 206.

[0202] Figure 9The connection method between the computing device clusters shown can be based on the fact that the application management method provided in this application requires a large amount of computing resources to expand the strategy configuration of the business intelligence agent 206, and to respond to user queries through the business intelligence agent 206. Therefore, it is considered that the functions implemented by the interaction device 202 are executed by an independent computing device, such as computing device 600C, and the functions implemented by the remaining intelligence agent expansion device 204 and business intelligence agent 206 are executed by computing device 600D.

[0203] It should be understood that Figure 9 The functions of the computing device 600C shown can also be performed by multiple computing devices 600. Similarly, the functions of the computing device 600D can also be performed by multiple computing devices 600.

[0204] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a computing device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the application management method described above, applied to the application management platform 20.

[0205] This application also provides a computer program product containing instructions. The computer program product may be a software or program product containing instructions, capable of running on a computing device or stored on any usable medium. When the computer program product is run on at least one computing device, it causes the at least one computing device to execute the application management method described above.

[0206] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. An application management method, characterized in that, The method is executed by an application management platform, which includes an interactive device, an intelligent agent extension device, and a business intelligent agent for at least one business. The method includes: The interactive device acquires the query input by the user, as well as the agent identifier and conditions, wherein the conditions include at least one of the following: task type, developer affiliation information, programming language, programming framework, or type of the interactive device. The interactive device collects the context of the query; The intelligent agent extension device determines the strategy configuration information of the target intelligent agent based on the intelligent agent identifier and the conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in the business intelligent agents of the at least one business. The strategy configuration information includes at least one of the following: prompting assistance information, truncation strategy, model identifier of the generated model, knowledge base identifier, retrieval strategy, recall strategy, tool identifier, risk control strategy, or role type. The prompting assistance information is used to assist in generating the generated model. The target agent expands the context of the query according to the strategy configuration information, and the expanded context includes at least one of the knowledge base retrieval results or tool call results; The target agent assembles the context and the extended context according to the policy configuration information to obtain the prompt; The target agent inputs the prompt into the generative model to perform reasoning and obtain an answer.

2. The method according to claim 1, characterized in that, The agent extension device determines the policy configuration information of the target agent based on the agent identifier and the conditions, including: The agent extension device determines the policy configuration information of the target agent through a routing strategy based on the agent identifier and the conditions.

3. The method according to claim 1 or 2, characterized in that, The method further includes: Configure the metadata of the business intelligence agent for the at least one business, the metadata including the routing strategy of the business intelligence agent, the routing strategy including strategy identifier, routing input information and routing output information, the routing input information being generated based on at least one of application identifier, intelligence agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and the routing output information including at least one of model identifier, prompt identifier, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier.

4. The method according to claim 3, characterized in that, The metadata of the business intelligence agent configuring the at least one service includes: The system receives at least one of the following: the agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier, configured by the user through the agent configuration interface for the at least one service. The routing strategy is generated based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier, and instance identifier.

5. The method according to claim 3 or 4, characterized in that, The method further includes: The system receives at least one of the user-configured business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy, and the metadata further includes at least one of the business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy.

6. The method according to any one of claims 1 to 5, characterized in that, The prompting assistance information includes any one or more of the following: prompt identifier, prompt template, or parameter name in the prompt.

7. The method according to any one of claims 1 to 6, characterized in that, The prompt assistance information includes the parameter name in the prompt, the application management platform further includes a front-end extension device, the policy configuration information further includes at least one of the extension identifier or instance identifier of the front-end extension device, and the method further includes: The interactive device receives at least one of the extended identifier or the instance identifier and the parameter name sent by the intelligent agent extension device; The interactive device sends at least one of the extension identifier or the instance identifier and the parameter name to the front-end extension device; the interactive device receives the parameter value corresponding to the parameter name returned by the front-end extension device. The context of the query includes the parameter values.

8. The method according to claim 7, characterized in that, The method further includes: The front-end extension device processes the answer to obtain the processed answer; The interactive device displays the processed answer to the user.

9. The method according to any one of claims 1 to 7, characterized in that, The auxiliary information includes a prompt icon; The target agent assembles the context and the extended context according to the policy configuration information to obtain the prompt, including: The target agent obtains a prompt template from the prompt word center based on the prompt identifier in the policy configuration information; The target agent fills in the prompt template based on the context and the extended context to obtain the prompt.

10. An application management platform, characterized in that, The application management platform includes an interactive device, an intelligent agent extension device, and a business intelligent agent for at least one business. The interactive device is used to obtain a query input by the user, and to obtain an agent identifier and conditions, wherein the conditions include at least one of the following: task type, developer affiliation information, programming language, programming framework, or type of the interactive device. The interactive device is also used to collect the context of the query; The intelligent agent extension device is used to determine the strategy configuration information of the target intelligent agent according to the intelligent agent identifier and the conditions. The target intelligent agent is the intelligent agent corresponding to the intelligent agent identifier in the business intelligent agents of the at least one business. The strategy configuration information includes at least one of the following: prompting assistance information, truncation strategy, model identifier of the generated model, knowledge base identifier, retrieval strategy, recall strategy, tool identifier, risk control strategy or role type. The prompting assistance information is used to assist in the generation of the generated model. The target agent is configured to expand the context of the query according to the policy configuration information, wherein the expanded context includes at least one of knowledge base retrieval results or tool call results; and to assemble the context and the expanded context according to the policy configuration information to obtain the prompt. The prompt is input into the generative model for reasoning to obtain an answer.

11. The platform according to claim 10, characterized in that, The intelligent agent extension device is specifically used for: Based on the agent identifier and the conditions, the policy configuration information of the target agent is determined through a routing strategy.

12. The platform according to claim 10 or 11, characterized in that, The platform also includes: An intelligent agent development device is used to configure the metadata of the business intelligent agent for the at least one business. The metadata includes the routing strategy of the business intelligent agent. The routing strategy includes a strategy identifier, routing input information, and routing output information. The routing input information is generated based on at least one of the following: application identifier, intelligent agent identifier, task type, developer affiliation information, programming language, programming framework, or type of interaction device. The routing output information includes at least one of the following: model identifier, prompt identifier, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier.

13. The platform according to claim 12, characterized in that, The intelligent agent development device is specifically used for: The system receives at least one of the following: the agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier of front-end extension device, and instance identifier, configured by the user through the agent configuration interface for the at least one service. The routing strategy is generated based on at least one of the following: application identifier, agent identifier, task type, developer affiliation information, programming language, programming framework or interaction device type, and at least one of the following: model identifier, prompt assistance information, knowledge base identifier, tool identifier, extension identifier, and instance identifier.

14. The platform according to claim 12 or 13, characterized in that, The intelligent agent development device is also used for: The system receives at least one of the user-configured business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy, and the metadata further includes at least one of the business terminology, retrieval strategy, truncation strategy, recall strategy, or risk control strategy.

15. The platform according to any one of claims 10 to 14, characterized in that, The prompting assistance information includes any one or more of the following: prompt identifier, prompt template, or parameter name in the prompt.

16. The platform according to any one of claims 10 to 15, characterized in that, The prompt assistance information includes the parameter name in the prompt, the application management platform also includes a front-end extension device, and the policy configuration information also includes at least one of the extension identifier or instance identifier of the front-end extension device; The interaction device is further configured to receive at least one of the extension identifier or the instance identifier and the parameter name sent by the intelligent agent extension device; send at least one of the extension identifier or the instance identifier and the parameter name to the front-end extension device; and receive the parameter value corresponding to the parameter name returned by the front-end extension device. The context of the query includes the parameter values.

17. The platform according to claim 16, characterized in that, The front-end extension device is also used for: The answer is processed to obtain the processed answer; The interactive device is also used for: The processed answer is displayed to the user.

18. The platform according to any one of claims 10 to 16, characterized in that, The auxiliary information includes a prompt icon; The target intelligent agent is specifically used for: Based on the prompt identifier in the strategy configuration information, obtain the prompt template from the prompt word center; The prompt template is populated based on the context and the extended context to obtain the prompt.

19. A computing device cluster, characterized in that, The computing device cluster includes at least one computing device, the at least one computing device including at least one processor and at least one memory, the at least one memory storing computer-readable instructions; the at least one processor executes the computer-readable instructions to cause the computing device cluster to perform the application management method as described in any one of claims 1 to 9.

20. A computer-readable storage medium, characterized in that, Includes computer-readable instructions; the computer-readable instructions are used to implement the application management method according to any one of claims 1 to 9.

21. A computer program product, characterized in that, Includes computer-readable instructions; the computer-readable instructions are used to implement the application management method according to any one of claims 1 to 9.

Images