Knowledge base based agent execution method, device, medium and program product

Abstract

This disclosure presents a knowledge-based agent execution method, apparatus, medium, and program product. The method includes: constructing a multi-dimensional knowledge base; acquiring user input data and matching the input data with the multi-dimensional knowledge base to generate a logical plan suitable for the agent; wherein the logical plan includes multiple logical steps and an execution tool adapted for each logical step; driving the agent to call the execution tool according to the logical plan, executing the corresponding logical steps in an orderly manner, and acquiring key information during the execution process in real time; updating the multi-dimensional knowledge base based on the key information; and guiding the agent to perform task execution by combining the updated multi-dimensional knowledge base and the new input data when new input data is acquired. This method, by constructing and updating a multi-dimensional knowledge base and generating a reasonable logical plan to drive the agent to execute in an orderly manner, significantly improves the decision-making quality and task execution capability of the agent in complex professional scenarios.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a knowledge base-based intelligent agent execution method, apparatus, medium, and program product. Background Technology

[0002] In the field of artificial intelligence, the development of multi-agent systems has enabled the widespread application of intelligent agents in various scenarios. The planning ability of an agent is crucial for its effective operation in specialized scenarios, directly impacting its ability to accurately execute tasks, rationally utilize tools, and make informed decisions. As application scenarios become increasingly specialized, the requirements for the planning capabilities of intelligent agents are becoming more stringent.

[0003] Current intelligent agent planning data generation mainly relies on conventional algorithms and general-purpose data. However, this approach lacks effective support from domain knowledge, resulting in unprofessional planning decisions that are difficult to adapt to complex professional scenarios. Furthermore, there are inappropriate tool usage practices, failing to accurately select and apply tools based on specific task requirements, which severely impacts the quality of intelligent agent decision-making in specialized scenarios. Summary of the Invention

[0004] In view of this, the present disclosure provides a knowledge base-based intelligent agent execution method, apparatus, medium, and program product, which can generate logical plans by matching input data with a constructed multi-dimensional knowledge base to drive intelligent agent execution and obtain key information to update the knowledge base. This effectively makes up for the shortcomings of current intelligent agent planning, such as lack of domain knowledge support and unreasonable tool calls, and significantly improves the decision-making quality and task execution capability of intelligent agents in complex professional scenarios.

[0005] In a first aspect, embodiments of this disclosure provide a method for executing an intelligent agent in a knowledge base, employing the following technical solution: Build a multi-dimensional knowledge base; The system acquires user input data, matches the input data with the multi-dimensional knowledge base, and generates a logical plan suitable for the intelligent agent; wherein the logical plan includes multiple logical steps and an execution tool adapted for each logical step. Based on the aforementioned logical plan, the driving agent calls the execution tool to execute the corresponding logical steps in an orderly manner, and obtains key information in the execution process in real time; Based on the aforementioned key information, the multi-dimensional knowledge base is updated; When new input data is acquired, the agent is guided to perform tasks by combining the updated multi-dimensional knowledge base with the new input data.

[0006] Optionally, the construction of the multi-dimensional knowledge base includes: Acquire at least one professional knowledge document in the target field; Based on the preset knowledge types, extract multiple targeted knowledge items from the professional knowledge documents of each target domain; Based on the target domain and knowledge type of each piece of targeted knowledge, all targeted knowledge is organized into a knowledge base of the corresponding target domain and corresponding dimension.

[0007] Optionally, the multi-dimensional knowledge base includes at least one set of knowledge bases for a target domain; Each knowledge base collection includes a professional domain planning library, a factual rule library, an experience case library, a tool parameter library, and a risk warning library.

[0008] Optionally, matching the input data with the multi-dimensional knowledge base to generate a logical plan suitable for the agent includes: Analyze the domain to which the input data belongs, and match the domain to which the data belongs with the target domain of each knowledge base set in the multi-dimensional knowledge base to determine the target knowledge base set; Based on the input data, knowledge mining is performed on the professional domain planning library, fact rule library, experience case library, tool parameter library and risk warning library included in the target knowledge base set to generate logical planning suitable for intelligent agents.

[0009] Optionally, based on the input data, knowledge mining is performed on the professional domain planning library, fact rule library, experience case library, tool parameter library, and risk warning library included in the target knowledge base set to generate logical planning suitable for the intelligent agent, including: The input data is matched with all logical planning frameworks in the professional domain planning library to determine the target logical planning framework; Based on the target logic planning framework, candidate steps containing triggering conditions are selected from the fact rule base and a candidate step set is formed. Based on the input data, multiple logical steps are selected from the candidate step set; All logical steps are embedded into the logical planning framework to form a preliminary plan with a logical structure; Based on the aforementioned experience case library, the logical steps, triggering conditions, and logical structure in the preliminary plan are optimized and adjusted to obtain an improved plan. Based on the tool parameter library, determine the execution tool for each logical step in the improvement plan; Based on the aforementioned risk warning database, contingency plans are determined for each logical step in the improvement plan; The execution tool and the emergency plan are added to the improvement plan to generate a logical plan.

[0010] Optionally, the step of driving the intelligent agent to call the execution tool according to the logical plan, executing the corresponding logical steps in an orderly manner, and obtaining key information in the execution process in real time includes: The agent starts by checking the initial logical step in the logical plan and determines whether the execution requirements are met based on the triggering conditions of the initial logical step. When the execution requirements are met, the agent calls the execution tool adapted to the initial logic steps to begin executing the task. When executing each logical step, the agent checks whether the execution process is normal; If the execution process is normal, after the current logical step is completed, the agent determines and executes the next logical step according to the logical structure. If an abnormal situation occurs during the execution process, the agent stops the execution of the current logical step and calls the emergency plan corresponding to the current logical step; When the emergency plan is successfully executed, the intelligent agent returns to the normal execution flow according to the logical plan and continues to execute subsequent logical steps; The intelligent agent records key information during the execution process.

[0011] Optionally, updating the multi-dimensional knowledge base based on the key information includes: Based on the aforementioned key information, logical planning change items are obtained, and compliance checks are performed on the logical planning change items. When the compliance check of the logical planning change item passes, the logical planning change item is updated to the experience case library of the target knowledge base set; Obtain user feedback information, and determine whether the logical plan is of high quality based on the key information and the feedback information; If so, the logical plan is updated to the professional domain planning library of the target knowledge base set; If not, then based on the key information and the feedback information, confirm the knowledge revision content, and update the target knowledge base set based on the knowledge revision content.

[0012] Secondly, this disclosure also provides a knowledge-based intelligent agent execution system, employing the following technical solution: Build modules are used to construct multi-dimensional knowledge bases; A generation module is used to acquire user input data, match the input data with the multi-dimensional knowledge base, and generate a logical plan suitable for the intelligent agent; wherein, the logical plan includes multiple logical steps and an execution tool adapted for each logical step; The execution module is used to drive the intelligent agent to call the execution tool according to the logical plan, execute the corresponding logical steps in an orderly manner, and obtain key information in the execution process in real time. An update module is used to update the multi-dimensional knowledge base based on the key information. The guidance module is used to guide the agent to perform tasks when new input data is obtained, by combining the updated multi-dimensional knowledge base with the new input data.

[0013] Thirdly, this disclosure also provides a computer device, which adopts the following technical solution: The computer device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed, enable the at least one processor to perform any of the knowledge base-based agent execution methods described above.

[0014] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium storing computer instructions for causing a computer to execute any of the knowledge-based agent execution methods described above.

[0015] Fifthly, embodiments of this disclosure also provide a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of any of the methods described above.

[0016] The knowledge-based agent execution method provided in this disclosure utilizes a multi-dimensional knowledge base that encompasses a rich variety of information, providing a comprehensive and extensive knowledge foundation for the agent's decision-making and execution. This enables the agent to perform logical reasoning and task execution based on reliable knowledge. By matching user input data with the knowledge base, the agent can more accurately understand the user's intent. The knowledge in the knowledge base helps the agent parse and analyze the input data, identifying the user's specific needs and problems, and avoiding erroneous execution due to misunderstanding the user's intent. The logical plan generated based on the matching results includes multiple logical steps and corresponding execution tools for each step. This targeted planning allows the agent to execute tasks in an optimized manner, improving the efficiency and effectiveness of task execution. Different combinations of logical steps and execution tools can be flexibly adjusted according to specific tasks, ensuring that the agent completes the task in the best way. Furthermore, the agent can autonomously determine execution steps and tools based on the knowledge base and user input, reducing reliance on human intervention and improving the agent's autonomy and intelligence level. Executing logical steps in an orderly manner according to a logical plan ensures a smooth and efficient task execution process. Each step has a clear objective and execution tools, allowing the agent to complete each step sequentially according to a predetermined order. This avoids chaos and disorder in task execution, improving the success rate. Real-time acquisition of key information during execution allows the agent to understand the progress and status of the task. If problems or anomalies occur during execution, the agent can make timely adjustments and decisions based on this key information to ensure successful task completion. Through the analysis and processing of key information, the agent can continuously optimize the execution process. For example, if the efficiency of a certain execution step is found to be low, the agent can adjust the execution strategy or change the execution tools based on key information, thereby improving the overall task execution effect. As tasks are executed and the environment changes, the knowledge in the multi-dimensional knowledge base may become outdated or inaccurate. Key information during execution contains the agent's experiences and lessons learned in actual tasks. By updating the multi-dimensional knowledge base based on this key information, it can be continuously enriched and improved, ensuring that the knowledge in the multi-dimensional knowledge base is always up-to-date and accurate. This allows the agent to rely on more reliable knowledge for decision-making and action in subsequent tasks. The updating process of the multi-dimensional knowledge base also reflects the agent's learning ability. The agent can learn and accumulate experience from the execution process, transforming it into knowledge stored in the knowledge base, thereby continuously improving its knowledge reserves and enhancing its problem-solving capabilities.

[0017] The above description is merely an overview of the technical solution disclosed herein. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the knowledge base-based agent execution method provided in this embodiment of the disclosure; Figure 2 A flowchart illustrating the multi-dimensional knowledge base construction method provided in this embodiment of the disclosure; Figure 3 A flowchart illustrating the logic planning generation method provided in this embodiment of the disclosure; Figure 4 A schematic flowchart illustrating the knowledge mining method provided in this embodiment of the disclosure; Figure 5 A flowchart illustrating the agent-driven method provided in this embodiment of the disclosure; Figure 6 A flowchart illustrating the multi-dimensional knowledge base update method provided in this embodiment of the disclosure; Figure 7 A principle block diagram of a knowledge-based intelligent agent execution system provided in the embodiments of this disclosure; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. Detailed Implementation

[0020] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0021] It should be understood that the following specific examples illustrate the implementation of this disclosure, and those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0022] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0023] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0024] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0025] Reference Figure 1 This disclosure provides a knowledge-based agent execution method, including the following steps: S1: Construct a multi-dimensional knowledge base; S2: Obtain user input data, match the input data with a multi-dimensional knowledge base, and generate a logical plan suitable for the intelligent agent; S3: Based on the logical plan, drive the agent to call the execution tool, execute the corresponding logical steps in an orderly manner, and obtain key information in the execution process in real time; S4: Update the multi-dimensional knowledge base based on key information; S5: When new input data is obtained, the agent is guided to perform tasks by combining the updated multi-dimensional knowledge base with the new input data.

[0026] Logical planning includes multiple logical steps and execution tools adapted for each logical step.

[0027] The knowledge-based agent execution method disclosed herein utilizes a multi-dimensional knowledge base encompassing rich and diverse information. This provides a comprehensive and extensive knowledge foundation for the agent's decision-making and execution, enabling the agent to perform logical reasoning and task execution based on reliable knowledge. By matching user input data with the knowledge base, the agent can more accurately understand the user's intent. The knowledge in the knowledge base helps the agent parse and analyze the input data, identifying the user's specific needs and problems, and avoiding erroneous execution due to misunderstanding of user intent. The logical plan generated based on the matching results includes multiple logical steps and corresponding execution tools for each step. This targeted planning allows the agent to execute tasks in an optimized manner, improving task execution efficiency and effectiveness. Different combinations of logical steps and execution tools can be flexibly adjusted according to specific tasks, ensuring the agent completes tasks in the best possible way. Furthermore, the agent can autonomously determine execution steps and tools based on the knowledge base and user input, reducing reliance on human intervention and improving the agent's autonomy and intelligence level.

[0028] Executing logical steps in an orderly manner according to a logical plan ensures a smooth and efficient task execution process. Each step has a clear objective and execution tools, allowing the agent to complete each step sequentially according to a predetermined order. This avoids chaos and disorder in task execution, improving the success rate. Real-time acquisition of key information during execution allows the agent to understand the progress and status of the task. If problems or anomalies occur during execution, the agent can make timely adjustments and decisions based on this key information to ensure successful task completion. Through the analysis and processing of key information, the agent can continuously optimize the execution process. For example, if the efficiency of a certain execution step is found to be low, the agent can adjust the execution strategy or change the execution tools based on key information, thereby improving the overall task execution effect.

[0029] As tasks are executed and the environment changes, the knowledge in the multi-dimensional knowledge base may become outdated or inaccurate. Key information during execution contains the agent's experiences and lessons learned in actual tasks. By updating the multi-dimensional knowledge base based on this key information, it can be continuously enriched and improved, ensuring that the knowledge in the multi-dimensional knowledge base is always up-to-date and accurate. This allows the agent to rely on more reliable knowledge for decision-making and action in subsequent tasks. The updating process of the multi-dimensional knowledge base also reflects the agent's learning ability. The agent can learn and accumulate experience from the execution process, transforming it into knowledge stored in the knowledge base, thereby continuously improving its knowledge reserves and enhancing its problem-solving capabilities.

[0030] In S1, refer to Figure 2 The flowchart illustrating the multi-dimensional knowledge base construction method shows that "building a multi-dimensional knowledge base" includes the following steps: S11: Obtain at least one professional knowledge document in the target field; S12: Extract multiple pieces of targeted knowledge from the professional knowledge files of each target domain according to the preset knowledge type; S13: Organize all targeted knowledge into a knowledge base of the corresponding target domain and corresponding dimension according to the target domain and knowledge type of each piece of targeted knowledge.

[0031] In S11, at least one target domain is identified based on actual needs. These needs may be to solve specific business problems or meet the application requirements of specific scenarios. For example, the target domain for developing intelligent agents in the medical field might be assisting in disease diagnosis and treatment plan formulation, while in the financial field it might cover investment analysis and risk assessment. Professional knowledge documents for each target domain are collected. These documents include normative, tool, case study, and planning documents. Normative documents refer to authoritative and guiding documents that define industry codes of conduct and operational standards within the target domain, such as the "Guidelines for Internal Medicine" and "Financial Regulatory Rules." Tool documents describe the tools, equipment, software, or technical interfaces used by intelligent agents within the target domain, such as ECG machine parameter manuals and Bloomberg data source API documentation. Case studies document past events, projects, or problem-solving processes within the target domain, including detailed background, methods, and results, such as historical medical records, financial analysis reports, and expert experience documents. Planning documents outline forward-looking and systematic plans, processes, or steps for specific tasks or projects within the target domain, such as "Elderly Chest Pain Diagnosis and Treatment Process Planning" and "Quarterly Profit Analysis Step Planning." Collecting these different types of professional knowledge documents can provide a rich, accurate and comprehensive knowledge base for the logical planning of intelligent agents, making their task execution more intelligent, efficient and accurate.

[0032] In S12, the preset knowledge types include factual rules, tool parameters, experience cases, risk warning information, and planning schemes. A hybrid extraction technique is used to extract multiple types of targeted knowledge from all professional knowledge documents. For factual rule knowledge, a rule engine is used to perform semantic analysis on the professional knowledge documents, accurately identifying mandatory words such as "must" and "should," and extracting statements containing industry standards and necessary steps to form explicit factual rule-oriented knowledge. For example, "patients with acute chest pain should complete an electrocardiogram within 10 minutes" is extracted from a medical document. For tool parameter knowledge, Named Entity Recognition (NER) technology is used to deeply mine the documents, locating and extracting various attribute information of the tools, such as applicable scenarios, priorities, and response times. This information is integrated to form targeted knowledge of tool parameters, such as "ECG machine: applicable to ST segment elevation detection, priority A." For experience case knowledge extraction, event extraction technology is used to structure the text in the documents, identifying expert advice and historical optimization schemes, such as "elderly patients suggest increasing screening for underlying diseases"—experience case-oriented knowledge. For risk warning information, correlation analysis is used to mine the data and text in the documents, identify potential risk points and corresponding solutions, and form risk warning-oriented knowledge, such as "delay in financial API opening hours → switching to Reuters interface". For planning scheme knowledge, clustering analysis algorithms are used to cluster past planning schemes in the documents according to different professional fields and problem directions, based on dimensions such as symptoms and task types, to extract general and reusable planning frameworks as planning scheme-oriented knowledge, such as "chest pain patients → basic examination → abnormality handling → detailed diagnosis".

[0033] In S13, the multi-dimensional knowledge base includes a knowledge base set for each target domain. Each knowledge base set includes five sub-bases: a professional domain planning base, a fact rule base, an experience case base, a tool parameter base, and a risk warning base. The dimensions of these sub-bases correspond to previously preset knowledge types. All extracted targeted knowledge is stored in the corresponding knowledge base according to its knowledge type. Knowledge belonging to the fact rule type is stored in the fact rule base for easy subsequent retrieval and adherence. Tool parameter knowledge is entered into the tool parameter base, forming a systematic set of tool features to facilitate understanding the various attributes of the tools. Experience case knowledge is categorized into the experience case base to provide reference and guidance for similar situations. Risk warning knowledge is stored in the risk warning base to quickly obtain response solutions when potential risks arise. Planning scheme knowledge is organized into the planning scheme base, clustered according to different professional domains and problem directions to form a reusable planning framework system, enabling rapid retrieval and application of appropriate planning schemes when facing new tasks. The five sub-bases of the multi-dimensional knowledge base are shown in the following structure table: Structure Table Extracting targeted knowledge according to preset knowledge types allows for focusing on key content in professional knowledge documents, effectively filtering out irrelevant information, and ensuring that the extracted knowledge is targeted and relevant to meet specific needs. Organizing the extracted targeted knowledge into knowledge bases of corresponding dimensions by knowledge type makes knowledge storage more orderly and systematic, facilitating subsequent rapid retrieval, management, and updates. This provides comprehensive, accurate, and structured knowledge support for intelligent agent logical planning, improving the intelligence, efficiency, and accuracy of its task execution.

[0034] In S2, refer to Figure 3 The flowchart illustrating the logical programming generation method demonstrates that "matching input data with a multi-dimensional knowledge base to generate a logical program suitable for the agent" includes the following steps: S21: Analyze the domain of the input data, match the domain with the target domain of each knowledge base set in the multi-dimensional knowledge base, and determine the target knowledge base set; S22: Based on the input data, perform knowledge mining on the professional domain planning library, fact rule library, experience case library, tool parameter library and risk warning library contained in the target knowledge base set to generate logical plans suitable for intelligent agents.

[0035] In the above scheme, the input data is preprocessed to remove noise, invalid characters, etc., and converted into a uniform format for easier subsequent processing. Key features are extracted from the preprocessed input data; for example, keywords and thesaurus can be extracted from text data, and distribution characteristics can be analyzed from numerical data. Then, a domain feature library is established, covering typical features, keywords, and rules of various target domains. The extracted input data features are compared with the domain feature library, and the degree of matching between the input data and the features of each domain is determined by calculating similarity. If the matching degree reaches a certain threshold, the input data is considered to belong to that domain; if it matches multiple domains, it can be sorted according to the matching degree, and the domain with the highest matching degree can be selected as the belonging domain. Manual intervention can be used to assist in the judgment if necessary. Finally, a set of knowledge bases in the same target domain as the input data is selected as the target knowledge base set.

[0036] Reference Figure 4 The flowchart illustrating the knowledge mining method demonstrates that "based on input data, knowledge mining is performed on the target knowledge base set, which includes a professional domain planning library, a fact rule library, an experience case library, a tool parameter library, and a risk warning library, to generate a logical plan suitable for intelligent agents." This includes the following steps: S221: Match the input data with all logical planning frameworks in the professional domain planning library to determine the target logical planning framework; S222: Based on the objective logic programming framework, candidate steps containing triggering conditions are selected from the fact rule base and a candidate step set is formed. S223: Based on the input data, select multiple logical steps from the candidate step set; S224: Embed all logical steps into a logical planning framework to form a preliminary plan with a logical structure; S225: Based on the experience case library, optimize and adjust the logical steps, triggering conditions and logical structure in the preliminary plan to obtain an improved plan; S226: Based on the tool parameter library, determine the execution tool for each logical step in the improvement plan; S227: Based on the risk warning database, determine the contingency plan for each logical step in the improvement plan; S228: Add execution tools and contingency plans to the improvement plan to generate a logical plan.

[0037] In S221, the logical planning frameworks in the domain planning library are clustered according to different domains and problems. Therefore, by analyzing the input data and clarifying its domain and specific problems, a suitable domain planning library has already been matched when determining the target knowledge base set. When matching the input data with all logical planning frameworks in the planning library, it is only necessary to focus on the fit between the problem corresponding to the input data and the problem addressed by each framework. For example, if the input data is about the diagnosis and treatment of chest pain patients, the logical planning framework for the diagnosis and treatment of chest pain patients in the domain planning library is searched and identified as the target logical planning framework. The target logical planning framework includes multiple process stages, such as the "General Process for the Diagnosis and Treatment of Chest Pain Patients," which includes major stages such as triage, preliminary examination, and expert consultation.

[0038] In S222, the fact rule base contains the necessary steps and filtering rules for different problems in each target domain. Based on the process stages of the target logic planning framework, each process stage is analyzed one by one. For each process stage, the corresponding rules are queried from the fact rule base. These rules specify the necessary steps to be executed in the process stage and the triggering conditions for these necessary steps. After extracting these rules, the steps containing the triggering conditions are filtered out; these steps are the candidate steps, and all candidate steps together constitute the candidate step set. For example, in the pre-examination and triage stage of the "General Process for the Diagnosis and Treatment of Chest Pain Patients," the fact rule base contains the rule "If the patient has chest pain accompanied by difficulty breathing," based on which it is determined that an electrocardiogram (ECG) examination is required for the patient, and "perform an ECG examination" is a candidate step.

[0039] In step S223, the input data contains specific scenario information or special requirements. This information is compared with each candidate step in the candidate step set and its triggering conditions. If the input data meets the triggering conditions of a candidate step, that candidate step is selected as a logical step. For example, the input data indicates that the patient has chest pain accompanied by difficulty breathing, and the candidate step set includes the candidate step "If the patient has chest pain accompanied by difficulty breathing, then perform an electrocardiogram (ECG) examination." Since the input data meets its triggering conditions, "perform an ECG examination" becomes a logical step. In this way, multiple logical steps that match the input data are selected from the candidate step set.

[0040] In S224, the logical planning framework has clearly defined process stages and an overall logical sequence. The selected logical steps are embedded into this framework according to its process stages and logical sequence. Logical steps act as nodes, their execution order forms edges, and triggering conditions can be attributes of these edges, forming a preliminary plan with a logical structure. For example, in the "General Process for the Diagnosis and Treatment of Chest Pain Patients," the triage stage has the logical step "perform an electrocardiogram (ECG) examination," and the preliminary examination stage has the logical step "perform a blood test." Following the framework's sequence, the "perform an ECG examination" in the triage stage is executed first, followed by the "perform a blood test" in the preliminary examination stage. The order in which these steps are executed forms an edge, and the triggering condition, "if the patient has chest pain accompanied by difficulty breathing," is an attribute of that edge. In this way, all logical steps are embedded into the logical planning framework, forming a preliminary plan with a network structure.

[0041] In S225, the experience case library can be viewed as a collection of "optimization suggestions," and the preliminary plan is compared and analyzed with cases in the experience case library. For the logical steps in the preliminary plan, refer to the experience case library to see if there are better steps that can be replaced, or if unnecessary steps need to be deleted, or if new steps need to be added. For the triggering conditions of the logical steps, based on the cases in the experience case library, determine whether the thresholds or conditions of the triggering conditions need to be adjusted. For the logical structure, check whether the order between steps is reasonable, and whether the order needs to be adjusted to improve the efficiency and effectiveness of the plan. For example, the experience case library shows that in some cases, "performing a blood test" can be performed before "performing an electrocardiogram," so the logical structure in the preliminary plan can be adjusted. Through these optimization adjustments, an improved plan is obtained.

[0042] In S226, the tool parameter library contains information such as the attributes and priorities of various tools. For each logical step in the improvement plan, a suitable execution tool is selected from the tool parameter library based on its needs and characteristics. Simultaneously, priorities are assigned to the execution tools determined for each logical step. For example, for the logical step of "performing an electrocardiogram (ECG) examination," the tool parameter library may have multiple ECG machines available. Based on factors such as the performance and applicable scenarios of the ECG machines, different ECG machines are assigned different priorities, with higher-priority (e.g., Level A) ECG machines being selected first. If the execution tool of the current priority is unavailable, the next lower priority execution tool can be selected.

[0043] In S227, the risk warning database contains potential risks and corresponding solutions for different problems in each target area during execution steps. For each logical step in the improvement plan, its potential risks are analyzed, and contingency plans corresponding to these risks are retrieved from the risk warning database. For example, for the logical step of "performing an electrocardiogram (ECG) examination," there may be a risk of ECG machine malfunction; the risk warning database may contain a contingency plan such as "if the ECG machine malfunctions, immediately replace it with a backup ECG machine." Another example is the contingency plan of "switching to a backup interface" in case of data interface delay. In this way, appropriate contingency plans are determined for each logical step in the improvement plan.

[0044] In S228, the execution tools and contingency plans determined for each logical step are added to the improvement plan. Specifically, the execution tool and corresponding contingency plan are marked in the attribute column of each logical step in the improvement plan. For example, for the logical step of "perform an electrocardiogram (ECG)", the execution tool is marked as "XX model ECG machine (priority A)" and the contingency plan is "if the ECG machine malfunctions, immediately replace it with a backup ECG machine". By adding the execution tools and contingency plans to the improvement plan, a complete logical plan is finally generated.

[0045] Based on the above scheme, accurate and comprehensive logical plans can be generated for input data, providing precise guidance for the execution of intelligent agents. By matching input data with a professional domain planning library to determine the target framework, the plan is ensured to fit the actual problem; candidate steps are selected from a fact rule library to ensure that planning steps are based on rules; logical steps are selected based on input data to make the plan more targeted; logical steps are embedded into the framework to form a preliminary plan, building a clear logical structure; optimization with an experience case library can improve the efficiency and effectiveness of the plan; determining execution tools and prioritizing ensures the flexibility and reliability of step execution; and contingency plans are developed based on a risk warning library to enhance the plan's ability to cope with risks. The final generated logical plan provides comprehensive, scientific, and reliable guidance for intelligent agents to solve practical problems. Furthermore, the multi-dimensional knowledge base covers the entire planning process. Through targeted extraction and categorized storage, appropriate knowledge can be used in each planning stage to improve decision-making accuracy. At the same time, referencing past solutions can reduce redundant design and improve planning efficiency.

[0046] In S3, refer to Figure 5 The flowchart illustrating the agent-driven method demonstrates that "based on logical planning, the agent is driven to call execution tools, execute corresponding logical steps in an orderly manner, and acquire key information during the execution process in real time," including the following steps: S31: The agent starts by checking the initial logical step in the logical plan and determines whether the execution requirements are met based on the triggering conditions of the initial logical step. S32: When the execution requirements are met, the agent calls the execution tool adapted to the initial logic steps to start executing the task; S33: When executing each logical step, the agent checks whether the execution process is normal; if the execution process is normal, then execute S34; if an abnormal situation occurs during the execution process, then execute S35. S34: After the current logical step is completed, the agent determines and executes the next logical step according to the logical structure; S35: The agent stops the execution of the current logical step and invokes the emergency plan corresponding to the current logical step; S36: When the emergency plan is successfully executed, the intelligent agent returns to the normal execution flow according to the logical plan and continues to execute the subsequent logical steps; S37: The agent records key information during the execution process.

[0047] In the above steps, the agent first locates the initial logical step in the logical plan, reads the preset trigger conditions for that step in detail, and carefully compares the current situation with the trigger conditions to see if they are fully met. For example, if the initial logical step is "to measure body temperature" and the trigger condition is "the patient has fever symptoms," the agent will determine whether the patient has fever symptoms based on the input data or real-time monitoring, thereby determining whether the execution requirements are met. Once it is determined that the execution requirements of the initial logical step are met, the agent selects the execution tool that is compatible with the initial logical step from the previously determined execution tool information, and then sends an execution command to the execution tool to start the task. For example, if the initial logical step is "to measure body temperature" and the compatible execution tool is an "electronic thermometer," the agent will issue a command to start the electronic thermometer; if the execution requirements are not yet met, it will wait, and issue an alarm during the waiting period.

[0048] During the execution of each logical step, the agent continuously monitors the execution status. It collects various data and feedback information in real time based on the expected result and normal execution status of the step. For example, for the "measure body temperature" step, the agent monitors whether the electronic thermometer is displaying data normally and whether the measurement time is within a reasonable range. If there are any discrepancies with expectations, such as abnormal data fluctuations or execution timeouts, the agent determines that the execution process is abnormal.

[0049] When the current logical step is completed and the execution process is normal, the agent determines the next logical step based on the logical structure of the logical plan and checks whether the triggering conditions for the next logical step are met. If they are met, the agent calls the execution tool corresponding to the step and issues an execution command. If they are not met, the agent enters a waiting state, continuously monitoring whether the triggering conditions are met, or processes the situation according to preset rules, such as adjusting certain parameters to ensure that the triggering conditions are met. If an abnormal situation is detected during execution, such as an execution tool error, incomplete data, or network interruption, the agent immediately stops the execution of the current logical step. Subsequently, the agent quickly matches the emergency plan corresponding to the current step from the identified emergency plan information and activates the plan. For example, if the electronic thermometer malfunctions when "performing a temperature measurement," the agent will follow the emergency plan, such as replacing it with a spare thermometer.

[0050] After the emergency plan is executed, the agent checks the execution results. If the emergency plan is successful, that is, the abnormal situation is resolved, the agent returns to the normal execution process according to the logical plan, determines the next logical step to be executed, checks its triggering conditions, and if they are met, calls the corresponding execution tool to continue executing the subsequent steps. If the emergency plan cannot solve the problem, the agent pauses or terminates the task and records the error information in detail, including the specific manifestation of the abnormal situation and the execution process of the emergency plan.

[0051] Throughout the task execution, the agent continuously records key information, including the execution time and results of each logical step, whether any anomalies occurred, a detailed description of the anomalies, and the activation and execution status of contingency plans. This recorded key information can be stored in a specific data storage area for subsequent task summarization, analysis, and optimization, providing a reference for the execution of similar tasks.

[0052] This method enables intelligent agents to achieve more efficient planning through a combination of "templates + regulations + tools + suggestions + contingency plans," achieving a progressive planning approach from "referencing the past" to "precise constraints." By relying on risk warning libraries and experience case libraries, the success rate of handling abnormal situations (such as tool malfunctions) is higher. The professional domain planning library is clustered by problem, allowing similar problems in different domains (such as "analysis of abnormal medical indicators" and "screening of abnormal financial data") to reuse planning logic. The scenario switching time is shortened from "by week" to "by hour," enabling cross-domain knowledge reuse and dynamic updates, improving the adaptability of intelligent agents in multi-professional scenarios, and especially enhancing the planning efficiency of similar problems.

[0053] Furthermore, the intelligent agent includes an executor and a planner. Each time the executor completes a logical step, it feeds back the execution result to the planner. Upon receiving the feedback, the planner calls upon the fact rule base to check whether the current execution process conforms to industry standards. For example, in the logical planning execution of chest pain patient diagnosis and treatment in the medical field, if the executor completes the "perform an electrocardiogram (ECG) examination" step and reports the result, the planner will check whether the examination process is correct based on the industry standards for ECG examinations in the fact rule base. This includes checking whether the electrode placement meets standards, whether the examination time is within a reasonable range, and whether the generated ECG report is complete and accurate. In the industrial production field, if the executor completes the "component assembly" step, the planner will check whether the assembly process meets standards based on the industry standards for component assembly in the fact rule base. This includes checking whether the screw tightening torque is appropriate, whether the assembly sequence of parts is correct, and whether the overall precision after assembly meets requirements. This method ensures the standardization of the execution process.

[0054] In S4, refer to Figure 6 The flowchart illustrating the multi-dimensional knowledge base update method shows that "updating the multi-dimensional knowledge base based on key information" includes the following steps: S41: Based on key information, obtain logical planning change items and perform compliance checks on the logical planning change items; if the compliance check on the logical planning change items passes, proceed to S42; if the compliance check on the logical planning change items fails, proceed to S43. S42: Update the logical planning change items to the experience case library of the target knowledge base set; S43: There is no need to update the logical planning change items to the experience case library of the target knowledge base set; S44: Obtain user feedback information, and based on key information and feedback information, determine whether the logical plan is of good quality; if yes, proceed to S45; if no, proceed to S46. S45: Update the logical plan to the professional domain plan library of the target knowledge base set; S46: Based on key information and feedback information, confirm the knowledge revision content, and update the target knowledge base set based on the knowledge revision content.

[0055] In the above steps, a comprehensive analysis is conducted on the key information of the previously recorded execution process to identify the parts that lead to changes in the logical planning. These include adjustments to the order of execution steps, additions or deletions of steps, and modifications to trigger conditions. These parts are identified as logical planning change items. Based on the fact rule base, industry standards, and relevant laws and regulations, these change items undergo compliance checks to verify whether they violate established rules and requirements. For example, in the logical planning of a medical process, the check examines whether newly added diagnostic steps comply with medical industry treatment guidelines. If a logical planning change item passes the compliance check, it indicates that the change item is compliant with the rules and has certain value. At this point, these change items are added to the experience case library of the target knowledge base set. For example, if it is found that the execution time of a certain step is reduced by 30% by adopting a new method, this change item will be automatically added to the experience case library after compliance checks, providing a reference for subsequent similar plans. If a change item fails the compliance check, it indicates that it may contain unreasonable or non-compliant aspects and cannot be considered a high-quality logical step; therefore, it will not be updated to the experience case library.

[0056] Feedback on the execution results of the logic plan is collected through various channels, including user evaluations of execution efficiency, effectiveness, and ease of use. This feedback is then combined with previously recorded key information for comprehensive analysis. From an efficiency perspective, it's assessed whether the expected time target was achieved; from an effectiveness perspective, whether the practical problem was solved; and from a user experience perspective, whether the operation is convenient. Based on this comprehensive evaluation, the quality of the logic plan is determined. If the comprehensive evaluation determines the logic plan to be high-quality, it indicates good performance in actual execution, has high reference value and promotional significance. In this case, the logic plan is updated to the professional domain planning library of the target knowledge base set, so that it can be directly called upon when encountering similar problems in the future, improving the efficiency and quality of problem-solving. If the logic plan is determined to be low-quality, it indicates that there are areas for improvement. Combining key information and user feedback, the problems in the logic plan are analyzed to determine the knowledge content that needs revision, such as adjusting the execution method of a certain step or improving certain rules. Based on the determined knowledge revision content, the relevant content in the target knowledge base set is updated, including the tool parameter library, fact rule library, and risk warning library, so that the knowledge base can be continuously optimized to provide more accurate and effective support for subsequent logical planning generation.

[0057] In S5, from an accuracy perspective, the updated multi-dimensional knowledge base includes optimized experience cases and professional domain planning. Leveraging this knowledge, the agent can more accurately understand new input data. For example, in a medical diagnosis scenario, new patient symptom data, combined with the rich case studies and treatment guidelines in the updated knowledge base, allows the agent to accurately assess the condition and reduce the probability of misdiagnosis. From an efficiency standpoint, the experience cases in the knowledge base provide efficient solution references for the agent. If new input data is similar to a certain situation in the experience case library, the agent can quickly invoke the corresponding solution steps, avoiding derivation from scratch and saving task execution time. In terms of adaptability, new input data may bring new problems or situations. The updated knowledge base allows the agent to adapt better, flexibly adjusting its execution strategy based on the rules and cases in the knowledge base to cope with different scenarios. For example, in industrial production, faced with new production requirements data, the agent adjusts the production process based on the updated knowledge base to ensure smooth production. It is evident that guiding an agent to perform tasks by combining an updated multi-dimensional knowledge base with new input data can improve accuracy, efficiency, and adaptability, enabling the agent to complete tasks better.

[0058] In summary, this method provides a complete solution for efficient planning by intelligent agents in professional scenarios such as healthcare and finance through the deep integration of systematic knowledge base construction and the entire planning process. It effectively addresses the problems of insufficient knowledge support and poor adaptability in existing technologies. Specifically, by constructing an automated and structured professional knowledge base, fragmented storage and manual reliance are avoided. A timed triggering mechanism collects new guidelines, tool update documents, and new planning schemes from various fields weekly, automatically extracting knowledge and updating the multi-dimensional knowledge base to ensure its timeliness. A dynamic dependency relationship is established between the knowledge base and the entire planning process, enabling precise knowledge support for task decomposition, tool selection, and risk prediction. The newly added professional domain planning library allows for the efficient reuse of past solutions. Its dynamic update mechanism (including the inclusion of user-approved excellent solutions) allows the knowledge base to automatically absorb new content, which is 60% faster than static storage and manual updates. Furthermore, the professional domain planning library continuously accumulates high-quality solutions; after long-term use, the coverage of usable solutions can increase by 50%, achieving "self-evolution" of planning capabilities. Through a closed-loop mechanism of "knowledge base construction - planning execution - knowledge iteration", the knowledge base module is responsible for collecting various types of data, classifying and storing knowledge (including professional domain planning libraries) and updating them dynamically; the planner module realizes knowledge-driven decision-making by querying various sub-libraries (including professional domain planning libraries), combining different types of knowledge, and generating logical plans with alternative solutions; the executor module executes the plan and provides feedback on the results and user evaluations to help the knowledge base continuously optimize. The modules communicate with each other through standardized interfaces to ensure real-time linkage between the knowledge base and the planning process.

[0059] Reference Figure 7 This disclosure provides a knowledge-based intelligent agent execution system, including: Module 101 is used to build a multi-dimensional knowledge base; The generation module 102 is used to acquire user input data, match the input data with a multi-dimensional knowledge base, and generate a logical plan suitable for the intelligent agent; wherein, the logical plan includes multiple logical steps and an execution tool adapted for each logical step; The execution module 103 is used to drive the intelligent agent to call the execution tool according to the logical plan, execute the corresponding logical steps in an orderly manner, and obtain key information in the execution process in real time. Update module 104 is used to update the multi-dimensional knowledge base based on key information; The guidance module 105 is used to guide the agent to perform tasks when new input data is obtained, by combining the updated multi-dimensional knowledge base with the new input data.

[0060] The various variations and specific examples of the knowledge-based agent execution method provided above are also applicable to the knowledge-based agent execution system provided in this disclosure. Through the foregoing detailed description of the knowledge-based agent execution method, those skilled in the art can clearly understand the implementation method of the knowledge-based agent execution system. For the sake of brevity, they will not be described in detail here.

[0061] A computer device according to embodiments of the present disclosure includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0062] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory, causing the computer device to perform all or part of the steps of the knowledge-based agent execution method of the foregoing embodiments of this disclosure.

[0063] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.

[0064] like Figure 8 This is a schematic diagram of a computer device provided for an embodiment of the present disclosure. It illustrates a structural schematic diagram suitable for implementing the computer device in the embodiments of the present disclosure. Figure 8 The computer device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0065] like Figure 8 As shown, a computer device may include a processor (such as a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) or programs loaded from storage devices into random access memory (RAM). The RAM also stores various programs and data required for the operation of the computer device. The processor, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0066] Typically, the following devices can be connected to the I / O interface: input devices, such as sensors or visual information acquisition devices; output devices, such as displays; storage devices, such as magnetic tapes or hard drives; and communication devices. Communication devices allow the computer device to communicate wirelessly or wiredly with other devices (such as edge computing devices) to exchange data. Although Figure 8 A computer apparatus with various devices is shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or included alternatively.

[0067] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the knowledge-based intelligent agent execution method of embodiments of this disclosure are performed.

[0068] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0069] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the knowledge-based agent execution methods described in the foregoing embodiments of the present disclosure are performed.

[0070] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).

[0071] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0072] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0073] In this disclosure, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The block diagrams of devices, apparatuses, devices, and systems involved in this disclosure are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as "comprising," "including," "having," etc., are open-ended terms meaning "including but not limited to," and are used interchangeably with them. The terms "or" and "and" as used herein refer to the terms "and / or," and are used interchangeably with them unless the context clearly indicates otherwise. The term "such as" as used herein refers to the phrase "such as but not limited to," and is used interchangeably with it.

[0074] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0075] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0076] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0077] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0078] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A knowledge-based agent execution method, characterized in that, include: Construct a multi-dimensional knowledge base, wherein the multi-dimensional knowledge base includes at least one set of knowledge bases for a target domain, and each set of knowledge bases includes a professional domain planning base, a fact rule base, an experience case base, and a risk warning base; The construction of the multi-dimensional knowledge base includes: Acquire at least one professional knowledge document in the target field; Based on the preset knowledge types, extract multiple targeted knowledge items from the professional knowledge documents of each target domain; The preset knowledge types include factual rules, tool parameters, experience cases, risk warning information, and planning schemes; For factual rule knowledge, a rule engine is used to perform semantic analysis on professional knowledge documents, extracting statements that contain industry norms and necessary steps to form clear factual rule-oriented knowledge; For tool parameter knowledge, named entity recognition technology is used to deeply mine the files, locate and extract various attribute information of the tools, and integrate this information to form targeted knowledge of tool parameters; In terms of knowledge extraction from experience cases, event extraction technology is used to perform structured processing on the text in the file and identify the expert advice and historical optimization solutions contained therein. For risk warning information, correlation analysis is used to mine the data and text in the file, screen out potential risk points and corresponding countermeasures, and form targeted risk warning knowledge. For planning scheme knowledge, clustering analysis algorithms are used to cluster past planning schemes in the documents according to different professional fields and problem directions, based on symptoms and task types, to extract a general and reusable planning framework as planning scheme orientation knowledge; Based on the target domain and knowledge type of each piece of targeted knowledge, all targeted knowledge is organized into a knowledge base of the corresponding target domain and corresponding dimension. The system acquires user input data, matches the input data with the multi-dimensional knowledge base, and generates a logical plan suitable for the intelligent agent; wherein the logical plan includes multiple logical steps and an execution tool adapted for each logical step; the input data is text data. The step of matching the input data with the multi-dimensional knowledge base to generate a logical plan suitable for the intelligent agent includes: Analyze the domain to which the input data belongs, and match the domain to which the data belongs with the target domain of each knowledge base set in the multi-dimensional knowledge base to determine the target knowledge base set; Based on the input data, knowledge mining is performed on the professional domain planning library, fact rule library, experience case library, tool parameter library and risk warning library included in the target knowledge base set to generate logical planning suitable for intelligent agents; The step of performing knowledge mining on the target knowledge base set, including its professional domain planning library, fact rule library, experience case library, tool parameter library, and risk warning library, based on the input data, to generate logical planning suitable for the intelligent agent, includes: The input data is matched with all logical planning frameworks in the professional domain planning library to determine the target logical planning framework; Based on the target logic planning framework, candidate steps containing triggering conditions are selected from the fact rule base and a candidate step set is formed. Based on the input data, multiple logical steps are selected from the candidate step set; All logical steps are embedded into the logical planning framework to form a preliminary plan with a logical structure; Based on the aforementioned experience case library, the logical steps, triggering conditions, and logical structure in the preliminary plan are optimized and adjusted to obtain an improved plan. Based on the tool parameter library, determine the execution tool for each logical step in the improvement plan; Based on the aforementioned risk warning database, contingency plans are determined for each logical step in the improvement plan; Add the execution tool and the emergency plan to the improvement plan to generate a logical plan; Based on the aforementioned logical plan, the driving agent calls the execution tool to execute the corresponding logical steps in an orderly manner, and obtains key information in the execution process in real time; Based on the aforementioned key information, the multi-dimensional knowledge base is updated; wherein, the key information includes the execution time of each logical step, the execution result, whether any abnormal situations occurred, a detailed description of the abnormal situations, and the activation and execution status of the emergency plan; When new input data is acquired, the agent is guided to perform tasks by combining the updated multi-dimensional knowledge base with the new input data.

2. The knowledge-based agent execution method according to claim 1, characterized in that, Based on the aforementioned logical plan, the intelligent agent is driven to invoke the execution tool, execute the corresponding logical steps in an orderly manner, and acquire key information during the execution process in real time, including: The agent starts by checking the initial logical step in the logical plan and determines whether the execution requirements are met based on the triggering conditions of the initial logical step. When the execution requirements are met, the agent calls the execution tool adapted to the initial logic steps to begin executing the task. When executing each logical step, the agent checks whether the execution process is normal; If the execution process is normal, after the current logical step is completed, the agent determines and executes the next logical step according to the logical structure. If an abnormal situation occurs during the execution process, the agent stops the execution of the current logical step and calls the emergency plan corresponding to the current logical step; When the emergency plan is successfully executed, the intelligent agent returns to the normal execution flow according to the logical plan and continues to execute subsequent logical steps; The intelligent agent records key information during the execution process.

3. The knowledge-based agent execution method according to claim 1, characterized in that, The updating of the multi-dimensional knowledge base based on the key information includes: Based on the aforementioned key information, logical planning change items are obtained, and compliance checks are performed on the logical planning change items. When the compliance check of the logical planning change item passes, the logical planning change item is updated to the experience case library of the target knowledge base set; Obtain user feedback information, and determine whether the logical plan is of high quality based on the key information and the feedback information; If so, the logical plan is updated to the professional domain planning library of the target knowledge base set; If not, then based on the key information and the feedback information, confirm the knowledge revision content, and update the target knowledge base set based on the knowledge revision content.

4. A computer device, characterized in that, The computer device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the knowledge-based agent execution method according to any one of claims 1-3.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the knowledge-based agent execution method according to any one of claims 1-3.

6. A computer program product comprising computer instructions, characterized in that, When executed by a processor, the computer instructions implement the steps of the knowledge-based agent execution method according to any one of claims 1-3.

Images