A method and apparatus for intelligent service processing based on a context protocol

Abstract

The application provides a kind of intelligent service processing method and device based on context protocol, it is related to artificial intelligence technical field, the intelligent service processing method based on context protocol includes: according to current user demand entity and historical session log entity, corresponding session state package is obtained by preset context perception strategy;According to session state package, corresponding atomic task execution sequence is obtained by task disintegration;According to each atomic task of atomic task execution sequence, corresponding preset external tool is called to generate corresponding structured execution result;According to all the structured execution result and corresponding agent, the agent execution sequence corresponding to the atomic task execution sequence is generated;According to agent execution sequence, corresponding atomic task is executed to obtain the business result corresponding to current user demand entity.The application can improve the accuracy of user service execution.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an intelligent business processing method and apparatus based on a context protocol. Background Technology

[0002] With the rapid development of artificial intelligence technology, large-scale models, with their powerful natural language understanding, logical reasoning, and task processing capabilities, have been widely applied in various intelligent interaction and business processing scenarios, especially in low-code development and multi-agent collaboration. Existing large-scale models can receive various business instructions from users, assisting users in completing related operations, thereby reducing the operational threshold for users and improving business processing efficiency to a certain extent.

[0003] In related technologies, conventional methods typically parse only the user's current input information, which may lead to discrepancies between the parsing results and the user's actual business needs, affecting the accuracy of business execution. Summary of the Invention

[0004] The problem addressed by this invention is how to improve the accuracy of user service execution.

[0005] To address the above problems, the present invention provides an intelligent business processing method and apparatus based on a context protocol.

[0006] In a first aspect, the present invention provides an intelligent service processing method based on a context protocol, comprising: Retrieve the current user request entity and historical session log entity generated by the low-code platform based on user input; Based on the current user demand entity and the historical session log entity, the corresponding session state packet is obtained through a preset context-aware strategy; The corresponding atomic task execution sequence is obtained by decomposing the session state packet. The corresponding structured execution result is generated by calling the corresponding preset external tool for each atomic task in the atomic task execution sequence; Based on all the structured execution results and the corresponding agents, generate the agent execution sequence corresponding to the atomic task execution sequence; The business result corresponding to the current user demand entity is obtained by executing the corresponding atomic task according to the execution sequence of the intelligent agent.

[0007] Optionally, obtaining the corresponding session state packet based on the current user demand entity and the historical session log entity using a preset context-aware strategy includes: The current user demand entity and the historical session log entity are associated and matched to obtain the current complete entity graph; The session state packet is obtained by performing state transitions based on the current complete entity graph.

[0008] Optionally, the step of associating and matching the current user demand entity and the historical session log entity to obtain the current complete entity graph includes: Extract key information based on the current user demand entity; The corresponding supplementary information is obtained by comparing and matching the key information with the historical session log entity. The current complete entity map is generated based on the current user demand entity and the supplementary information.

[0009] Optionally, the step of obtaining the session state packet by performing state transition based on the current complete entity graph includes: Extract key contextual information based on the current complete entity graph; Based on the contextual key information, the corresponding state label is obtained through a preset state transition rule, wherein the state transition rule includes a one-to-one correspondence between the contextual key information and the state label; The session state package is generated based on all the contextual key information and the corresponding state labels.

[0010] Optionally, the step of decomposing the task based on the session state packet to obtain the corresponding atomic task execution sequence includes: Extract atomic tasks based on the session state packet; An atomic task execution sequence is generated based on the dependencies of all atomic tasks, wherein the dependencies include data dependencies and business process dependencies of the atomic tasks.

[0011] Optionally, the step of calling a corresponding preset external tool to generate a corresponding structured execution result based on each atomic task in the atomic task execution sequence includes: Determine the corresponding atomic task type based on the atomic task; The corresponding preset external tool is determined according to the atomic task type and the preset tool matching rule, wherein the tool matching rule includes a one-to-one correspondence between the atomic task type and the preset external tool; The structured execution result corresponding to the atomic task is generated based on the preset external tool.

[0012] Optionally, generating the agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents includes: Based on the structured execution result, match the agent corresponding to the atomic task in the agent role library; The corresponding agent execution sequence is generated based on the atomic task execution sequence and all the corresponding agents.

[0013] Optionally, the low-code platform includes a visual interface, through which the low-code platform receives the user input.

[0014] Optionally, it also includes a database for storing the business results and providing training data for the low-code platform.

[0015] Secondly, the present invention provides an intelligent service processing device based on a context protocol, comprising: The acquisition module is used to acquire the current user request entity and historical session log entity generated by the low-code platform based on user input; The processing module is used to obtain the corresponding session state packet based on the current user demand entity and the historical session log entity through a preset context-aware strategy. The decomposition module is used to decompose tasks based on the session state packet to obtain the corresponding atomic task execution sequence; The calling module is used to call the corresponding preset external tools to generate the corresponding structured execution results according to each atomic task in the atomic task execution sequence; A matching module is used to generate an agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents, wherein the agent is determined by the matching results of the structured execution results and a preset agent role library; The execution module is used to execute the corresponding atomic task according to the execution sequence of the intelligent agent to obtain the business result corresponding to the current user demand entity.

[0016] Thirdly, the present invention provides an electronic device, including a memory and a processor; The memory is used to store computer programs; The processor is configured to implement the context protocol-based intelligent business processing method as described in the first aspect when executing the computer program.

[0017] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent business processing method based on a context protocol as described in the first aspect.

[0018] The beneficial effects of the context-protocol-based intelligent business processing method of the present invention are as follows: By acquiring the current user demand entity and historical session log entity generated by the low-code platform based on user input, it can fully combine historical interaction information with the current intent, avoiding the one-sided understanding caused by relying solely on the current input; then, by integrating the above two types of entities through a preset context-aware strategy to obtain a session state package, it can achieve deep perception and structured expression of the user's true intent, contextual dependencies, and business scenarios, significantly improving the accuracy and completeness of demand understanding; based on the session state package, task decomposition to obtain atomic task execution sequences can break down ambiguous and complex business demands into the smallest execution units with fine granularity, clear logic, and explicit sequence, reducing demand ambiguity and execution deviation from the source. By calling corresponding preset external tools for each atomic task and generating structured execution results, the standardization of task processing and data output can be ensured, avoiding parsing errors and execution chaos caused by unstructured information. Furthermore, by combining all structured execution results with a preset agent role library to generate agent execution sequences, it is possible to accurately match different atomic tasks with agents possessing corresponding processing capabilities, achieving optimal adaptation between tasks and execution agents, and improving the rationality of division of labor and execution reliability. Finally, by executing the corresponding atomic tasks in an orderly manner according to the agent execution sequence, based on multi-agent collaboration, continuous context awareness, and step-by-step controllable tasks, it can fully restore and satisfy the user's true business intent, significantly reducing requirement parsing deviations and significantly improving the accuracy, stability, and consistency of overall business execution. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an intelligent business processing method based on a context protocol according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an intelligent service processing device based on a context protocol according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0021] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0022] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0023] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0024] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0025] In related technologies, conventional methods typically parse only the user's current input information, failing to fully integrate historical interaction records, business scenarios, and contextual information for comprehensive understanding. This can easily lead to incomplete semantic understanding and inaccurate intent recognition, causing the parsing results to deviate from the user's actual business needs. This deviation directly affects the rationality of subsequent task decomposition and execution path planning, resulting in unclear task division, mismatch between execution logic and actual business, and consequently reducing the accuracy and reliability of business process execution. It can even lead to data processing errors, process configuration deviations, and other problems, making it difficult to meet the precise, stable, and coherent interaction and processing requirements in complex business scenarios. It also restricts the application effectiveness of intelligent interaction systems in low-code, data processing, and other fields.

[0026] To address the problems existing in the aforementioned related technologies, embodiments of the present invention provide an intelligent business processing method and apparatus based on a context protocol.

[0027] like Figure 1 As shown in the figure, an intelligent service processing method based on a context protocol provided by an embodiment of the present invention includes: S110, Obtain the current user demand entity and historical session log entity generated by the low-code platform based on user input.

[0028] Specifically, acquiring the current user requirement entity generated by the low-code platform based on user input refers to the platform parsing, extracting, and structuring natural language or structured instructions such as business descriptions, functional requirements, field definitions, and process rules input by the user in real time during the interaction process. This forms a standardized requirement entity that can be recognized by the system and used for subsequent page configuration, interface generation, and logic orchestration. It includes core information such as business object attributes, operation behaviors, constraints, and relationships. At the same time, acquiring the historical session log entity is a complete record and entity encapsulation of the past interaction process between the user and the low-code platform. It covers key log information such as historical question content, system response content, requirement change records, operation sequence, configuration nodes, exception prompts, session identifiers, user identity, and timestamps. The two types of entities respectively carry real-time business requirements and the full interaction trajectory, jointly providing data support and basis for the low-code platform's requirements understanding, intelligent recommendation, process traceability, iterative optimization, and automated generation capabilities.

[0029] The low-code platform performs structured parsing of business requirements in the form of text or voice input by users. By combining named entity recognition, keyword extraction, and business rules to complete the intent details, it generates a current user requirement entity containing key information such as core business objects, time range, operation behavior, and constraints. This entity is also associated with and completed with the existing business entity graph in the system to form a standardized requirement carrier that can support subsequent task decomposition and tool invocation. The historical session log entity is a complete entity encapsulation of all past interactions between the user and the platform, covering core content such as session ID, timestamp, historical question content, previously extracted entity information, previous round of intelligent agent output results, task execution progress, and session state change records.

[0030] S120, based on the current user demand entity and the historical session log entity, obtain the corresponding session state packet through a preset context-aware strategy.

[0031] Specifically, the system first uses the defined current user need entity and historical session log entity as the basic data sources. The current user need entity carries the core intent, target, key parameters, and specific request information of the user's current interaction, while the historical session log entity records and consolidates user questions, system responses, contextual relationships, request evolution trajectories, and key interaction node information from multiple rounds of interaction. Then, a pre-defined context-aware strategy is used, which typically includes core logic such as context semantic association parsing, extraction of key information from historical sessions, matching and verification of current needs with historical context, identification of contextual dependencies, and determination of session continuity. This system deeply integrates, semantically aligns, mines connections, and infers states from structured and unstructured information in two types of entities. It eliminates redundant and invalid information, retains effective contextual features, and clarifies the intrinsic relationship between current needs and historical conversations. Finally, it integrates and forms a standardized conversation state package containing multi-dimensional information such as current conversation intent, contextual dependencies, key features of historical interactions, continuity of user needs, conversation stage states, and key constraints. This provides unified, complete, and reusable contextual state support for subsequent accurate intent recognition, response strategy matching, logical reasoning and decision-making, and personalized interaction execution, ensuring the coherence, consistency, and intelligence of multi-turn conversations.

[0032] S130, the corresponding atomic task execution sequence is obtained by decomposing the task according to the session state packet.

[0033] Specifically, after acquiring and parsing the aforementioned integrated standardized session state package, based on the multi-dimensional structured data contained in the state package, including the current user's core intent, contextual dependencies, historical interaction characteristics, requirement constraints, key parameter information, and session stage status, the complex tasks corresponding to the overall user needs are decomposed layer by layer, logically broken down, and modularized according to preset task decomposition rules, business logic levels, execution dependencies, and priority sorting strategies. The overall composite task is gradually decomposed into multiple atomic tasks with independent execution logic, minimal functional granularity, clear input and output, and no redundant coupling. At the same time, combined with the execution order constraints, prerequisite dependencies, resource call requirements, and exception handling logic contained in the context state, each atomic task is arranged in an orderly manner, with dependency verification and timing planning. Finally, a standardized atomic task execution sequence with sequential execution order, clear dependency relationships, and the ability to directly drive the operation of the underlying execution unit is formed. This provides a clear, orderly, and implementable execution instruction basis for subsequent task scheduling, interface calls, process execution, and result feedback, ensuring that the overall business process can be promoted accurately, efficiently, and coherently according to the context-aware results.

[0034] S140, according to each atomic task in the atomic task execution sequence, call the corresponding preset external tool to generate the corresponding structured execution result.

[0035] Specifically, based on the established and completed temporally ordered and dependency-verified atomic task execution sequence, according to the functional type, execution goal, input parameters, constraints, and preset calling rules of each atomic task in the sequence, pre-configured external tools corresponding to each atomic task are sequentially matched and called. These external tools may include data interfaces, functional components, computing modules, business services, data processing units, or third-party interaction interfaces. During the calling process, parameter passing, permission verification, execution instruction issuance, and execution status synchronization are completed. Through the independent or collaborative operation of each external tool, the corresponding atomic tasks are professionally processed, transforming unstructured, semi-structured, or singular task inputs into structured execution results that are formatted in a standardized manner, have complete fields, clear semantics, and can be directly used for subsequent process parsing and integration. The results include execution status, data content, return parameters, exception identifiers, and associated context information, thereby providing reliable and standardized data support for subsequent result aggregation, logical judgment, response generation, and session state updates, ensuring the accuracy, standardization, and traceability of the overall task execution process.

[0036] S150, Generate an agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents, wherein the agent is determined by the matching result of the structured execution results and the preset agent role library.

[0037] Specifically, after generating the structured execution results corresponding to each atomic task, all structured execution results are uniformly collected, feature-extracted, and semantically integrated. Combining the task type, processing content, data attributes, business scenarios, execution status, and contextual constraints contained in the execution results, a multi-dimensional matching and precise mapping is performed with a pre-built and stored intelligent agent role library. This pre-built intelligent agent role library contains various intelligent agent entities with different functional positioning, business permissions, processing capabilities, interaction logic, and exclusive responsibilities. By comparing task features with intelligent agent adaptation attributes, a high degree of matching with each execution result and corresponding atomic task is determined. The target intelligent agent is defined by clarifying the invocation timing, collaborative relationships, and division of labor boundaries of each intelligent agent. Then, based on the sequential timing, dependencies, and execution logic of the atomic task execution sequence, the matched intelligent agents are arranged and bound in order. Finally, an intelligent agent execution sequence is formed that corresponds one-to-one with the original atomic task execution sequence, has consistent timing, complete dependencies, and has autonomous execution and collaborative processing capabilities. This provides a standardized, schedulable, and scalable execution subject and timing basis for subsequent intelligent agent scheduling, distributed task processing, multi-role collaborative interaction, and overall process closed-loop execution, ensuring that complex tasks are carried out efficiently, orderly, and stably under the collaboration of multiple intelligent agents.

[0038] S160, according to the execution sequence of the intelligent agent, execute the corresponding atomic task to obtain the business result corresponding to the current user demand entity.

[0039] Specifically, following the completed sequence of agent execution with pre-arranged timing, role matching, and dependency associations, each agent in the sequence undertakes and executes its bound atomic tasks in an orderly manner based on its division of labor, execution permissions, and processing logic. During execution, parameter passing, state synchronization, exception handling, intermediate result interaction, and cross-agent collaborative scheduling are completed. The order of atomic tasks, prerequisite dependencies, and business constraints are strictly followed, and data processing, logical operations, interface calls, information integration, and decision-making are carried out sequentially. This fully leverages the unique processing capabilities of each agent in its corresponding task scenario. Through multi-agent collaboration, step-by-step execution, and closed-loop feedback, the phased execution results of each atomic task are aggregated, verified, and merged at each level, ultimately forming a complete, unified business result that conforms to business specifications and user expectations. This business result directly corresponds to and satisfies the core demands, functional goals, and scenario intents expressed by the current user demand entity, providing final and effective data and content support for subsequent result output, content presentation, conversation closure, and business process implementation.

[0040] In this embodiment, by acquiring the current user requirement entity and historical session log entity generated by the low-code platform based on user input, historical interaction information and current intent can be fully combined, avoiding the one-sided understanding caused by relying solely on current input. Then, by integrating these two types of entities through a preset context-aware strategy to obtain a session state package, a deep perception and structured expression of the user's true intent, contextual dependencies, and business scenarios can be achieved, significantly improving the accuracy and completeness of requirement understanding. Based on this session state package, task decomposition yields an atomic task execution sequence, which can break down ambiguous and complex business requirements into the smallest execution units with fine granularity, clear logic, and explicit sequence, reducing requirement ambiguity and execution deviation from the source. By calling each atomic task... By using corresponding preset external tools and generating structured execution results, the standardization of task processing and data output can be ensured, avoiding parsing errors and execution chaos caused by unstructured information. Then, by combining all structured execution results with a preset intelligent agent role library to generate intelligent agent execution sequences, it is possible to accurately match intelligent agents with corresponding processing capabilities for different atomic tasks, achieving optimal adaptation between tasks and execution agents, improving the rationality of division of labor and execution reliability. Finally, the corresponding atomic tasks are executed in an orderly manner according to the intelligent agent execution sequence. Based on multi-agent collaboration, continuous context awareness, and step-by-step controllable tasks, the true business intent of users can be fully restored and satisfied, significantly reducing the deviation of requirement parsing and significantly improving the accuracy, stability and consistency of overall business execution.

[0041] Optionally, obtaining the corresponding session state packet based on the current user demand entity and the historical session log entity using a preset context-aware strategy includes: The current user demand entity and the historical session log entity are associated and matched to obtain the current complete entity graph; The session state packet is obtained by performing state transitions based on the current complete entity graph.

[0042] In this optional embodiment, the current intent, key parameters, business requirements, and constraint information carried by the current user demand entity are globally associated, semantically aligned, and deeply matched with the multi-round interaction content, historical demand characteristics, contextual relationships, previous business states, and historical execution results recorded in the historical session log entity. This constructs a complete current entity graph containing entity relationships, demand chains, contextual dependencies, and business contexts, thereby achieving the organic integration and structured association of current demands and historical session information. Subsequently, based on this complete current entity graph, according to preset state deduction rules, context transition logic, and session lifecycle specifications, state identification and state transition deduction are performed on the evolution trend of user demands, contextual continuity relationships, business stage changes, and key state characteristics. From this, a standardized session state package containing current session intent, contextual state, demand continuity, business stage identifiers, and key constraint information is extracted and integrated, providing a comprehensive, accurate, and context-continuous state foundation for subsequent task decomposition, agent matching, and business execution.

[0043] Optionally, the step of associating and matching the current user demand entity and the historical session log entity to obtain the current complete entity graph includes: Extract key information based on the current user demand entity; The corresponding supplementary information is obtained by comparing and matching the key information with the historical session log entity. The current complete entity map is generated based on the current user demand entity and the supplementary information.

[0044] Specifically, firstly, based on the current user demand entity, key information such as the user's core business intent, key operation instructions, target object, required fields, constraints, and interaction scenario are extracted. For example, if the user only inputs "generate approval process," key information such as "process generation" and "approval type" can be extracted. Subsequently, the extracted key information is compared field by field with the past multi-round interaction content, historical demand statements, determined parameters, context constraints, and incomplete business nodes recorded in the historical session log entity. This involves semantic association and precise matching to trace back and match supplementary information related to the current key information but not explicitly stated at present. For example, information such as the approval entity, process nodes, data sources, and triggering conditions already identified in historical sessions can be used as supplementary information for the current request. Finally, the current user request entity's current demands, basic parameters, and supplementary information obtained through matching are integrated, associated, and structured to clarify the business relationships, dependencies, and contextual links between various pieces of information. This generates a complete entity graph that covers the current request and historical context, with complete information and clear relationships. This provides a comprehensive and unambiguous entity relationship foundation for subsequent state transitions, session state package construction, and accurate understanding of the user's true business intent, effectively avoiding problems such as incomplete requirements and misunderstandings caused by relying solely on current input.

[0045] For example, taking a user's request to "analyze the billing data of product A for the past 3 months and perform year-on-year analysis" as an example, we first extract key information from the current user request entity generated by this request: the product object is product A, the time range is the past 3 months, the operation intent is billing data statistics + year-on-year analysis, and the analysis dimension is billing data. Then, we compare and match the extracted key information with the historical session log entity to retrieve supplementary information about product A from the historical session log, including that product A is a video ringback tone product of a certain platform, the corresponding database table name is tb_video_ring_bill, the fields associated with the billing data include billing amount, billing quantity, and user ID, and that the user had previously requested that the billing amount be aggregated and statistically analyzed by month in a historical session. Year-on-year analysis requires benchmarking against data from the same period of the previous year. Additionally, outlier filtering of the product's billing data must utilize the IQR (Information Quality Reduction) rule – a historical configuration. Finally, the core content of the current user demand entity is integrated with the matched product attributes, database relationships, historical analysis rules, data processing requirements, and other supplementary information to generate a complete entity graph. This graph includes product A's full business attributes, associated database and table fields, precise time range definition, specific rules for analysis operations, and established data processing standards. This graph clearly outlines all entity relationships and business rules related to product A's billing data analysis, providing complete and accurate entity information support for subsequent task decomposition and tool invocation.

[0046] In this optional embodiment, by extracting key information from the current user demand entity, combining it with supplementary information from historical session log entity matching, and generating a complete entity graph, the core user demands can be accurately broken down. Information accumulated from historical interactions is used to comprehensively complete the current demand, making entity information more complete and business relationships clearer. Simultaneously, it achieves a deep association between the current demand and historical sessions, effectively continuing the cross-session business context, avoiding redundant business information definitions, reducing repetitive user input, and improving the efficiency and accuracy of demand parsing. Furthermore, the generated complete entity graph provides standardized and comprehensive entity information support for subsequent tasks such as intelligent agent task decomposition, tool invocation, and multi-agent collaboration, ensuring that subsequent task execution has a clear basis and guaranteeing the continuity and accuracy of business processing flows.

[0047] Optionally, the step of obtaining the session state packet by performing state transition based on the current complete entity graph includes: Extract key contextual information based on the current complete entity graph; Based on the contextual key information, the corresponding state label is obtained through a preset state transition rule, wherein the state transition rule includes a one-to-one correspondence between the contextual key information and the state label; The session state package is generated based on all the contextual key information and the corresponding state labels.

[0048] In this optional embodiment, contextual key information is comprehensively extracted from the current complete entity graph, covering dimensions such as core business objects, precise time range, specific operational intent, task execution history, entity relationships, and business rule constraints. This information includes not only the core requirements of the user's current needs but also supplementary content such as product attributes, data association configurations, and past operation rules accumulated from historical sessions, achieving a structured decomposition and extraction of the user's needs across all dimensions of context. The extracted contextual key information is then precisely matched with preset state transition rules. These state transition rules are constructed based on business scenario analysis and predetermine a unique one-to-one correspondence between contextual key information and state tags. Based on core elements such as the current intent type, session history, task execution progress, and requirement association attributes contained in the key information, the rules match the current session with the corresponding specific contextual key information. The status tags cover all scenario status types, including pending task decomposition, pending task re-decomposition, pending result secondary parsing, task continuation (specified stage), pending process reconfiguration + task re-decomposition, and session idle, ensuring that the status tags are highly consistent with actual business needs. Finally, all extracted key context information and the corresponding matched status tags are systematically integrated and encapsulated, while incorporating session identifiers, entity graph association information, intent sequences, etc., to generate the session status package containing complete context information and a clear processing state direction. This session status package fully carries all the core information and processing instructions of the current session and can be directly used as standardized input data for the subsequent automatic task decomposition of the intelligent agent. It provides a clear and comprehensive execution basis for the atomic decomposition of complex requirements and the precise invocation of tools, ensuring the continuity, accuracy, and efficiency of the intelligent agent's subsequent business processing flow.

[0049] Optionally, the step of decomposing the task based on the session state packet to obtain the corresponding atomic task execution sequence includes: Extract atomic tasks based on the session state packet; An atomic task execution sequence is generated based on the dependencies of all atomic tasks, wherein the dependencies include data dependencies and business process dependencies of the atomic tasks.

[0050] In this optional embodiment, the complex business needs of users are precisely atomically decomposed from the session state packet, which carries complete context information, processing state pointers, and intent sequences, in conjunction with a task type library and business rules. Atomic tasks with a single execution goal and capable of independent completion are extracted, covering all types of tasks such as data querying, data processing, model invocation, result integration, and format conversion. Each atomic task clearly defines its execution goal, input parameters, output requirements, and the type of tool to be invoked, ensuring clear boundaries and feasible execution. Furthermore, all extracted atomic tasks are analyzed from multiple dimensions to identify and clarify data dependencies and business process dependencies between them. Data dependencies are manifested in the requirement that the execution of a subsequent atomic task depends on a preceding atomic task. The output data serves as the input, and business process dependencies are reflected in the fact that, according to established business logic and execution specifications, some atomic tasks need to be executed sequentially after other tasks are completed. Based on these two types of dependencies, a task dependency topology graph is constructed. A topology sorting algorithm is used to eliminate task execution conflicts and clarify the order of task execution. Finally, an atomic task execution sequence that follows business logic and conforms to data flow patterns is generated. This sequence arranges all atomic tasks in an orderly manner and marks the dependency relationships, execution trigger conditions, and connection methods between tasks. This ensures both the continuity and accuracy of data flow and compliance with established business process specifications. It provides a clear and standardized basis for subsequent dynamic tool calls and multi-agent collaborative execution, ensuring that the decomposition and execution of complex business requirements proceed in an orderly and efficient manner throughout the entire process.

[0051] Optionally, the step of calling a corresponding preset external tool to generate a corresponding structured execution result based on each atomic task in the atomic task execution sequence includes: Determine the corresponding atomic task type based on the atomic task; The corresponding preset external tool is determined according to the atomic task type and the preset tool matching rule, wherein the tool matching rule includes a one-to-one correspondence between the atomic task type and the preset external tool; The structured execution result corresponding to the atomic task is generated based on the preset external tool.

[0052] In this optional embodiment, the decomposed atomic tasks are first classified and categorized. Based on the corresponding task type library, and considering the execution goal, operation object, and business action of the atomic task, a corresponding atomic task type is determined for each atomic task. This covers all standardized task types, including data query, data processing, model invocation, result integration, and format conversion. Each type clearly defines its core operation scope and execution requirements, ensuring the accuracy and consistency of atomic task type determination. Next, the determined atomic task types are precisely matched with preset tool matching rules. These rules are built based on business scenarios and tool capabilities, pre-setting a one-to-one correspondence between atomic task types and preset external tools. Based on the type attributes of the atomic task, the rules match the most suitable preset external tool from the tool resource library. For example, data query atomic tasks are matched with database query tools, data processing atomic tasks with data cleaning tools, and model invocation atomic tasks with data cleaning tools. The task matching AI model inference tool incorporates database adaptation rules to match atomic tasks from different data sources with corresponding dedicated tool interfaces, ensuring a high degree of compatibility between the tool and the atomic task. Finally, based on the matched preset external tools, the input parameters of the atomic task are converted into execution instructions recognizable by the tool, driving the preset external tools to complete specific operations according to the execution requirements of the atomic task. Simultaneously, the raw results returned by the tool execution undergo structured parsing, format conversion, and data validation, transforming the non-standardized raw output into structured data conforming to preset specifications. This ultimately generates a structured execution result for each atomic task, containing core content such as task execution status, standardized output data, and result descriptions, and maintaining consistency with the expected output format in the atomic task execution plan. This provides standardized, directly callable input data for subsequent multi-agent collaborative processing, ensuring the coherence of the overall task execution process and the accuracy of data flow.

[0053] Optionally, generating the agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents includes: Based on the structured execution result, match the agent corresponding to the atomic task in the agent role library; The corresponding agent execution sequence is generated based on the atomic task execution sequence and all the corresponding agents.

[0054] In this optional embodiment, firstly, based on the structured execution results output after each atomic task is executed, and combined with the core information contained in the results, such as task execution type, data processing requirements, output data format, and subsequent task execution requirements, a precise agent matching operation is performed in a preset agent role library. This agent role library predefines various agent roles, such as data processing agent, model training agent, result analysis agent, and data query agent. Each role is associated with clear core capabilities, processable task types, compatible data formats, and operational boundaries. During the matching process, a role matching algorithm comprehensively compares the attribute characteristics of the structured execution results with the capability attributes of each agent role, matching a unique and suitable agent for each atomic task to ensure a high degree of compatibility between the agent's capabilities and the execution requirements of the atomic task. Then, combined with the generated atomic task execution sequence, which clarifies the sequential execution order of each atomic task based on data dependencies and business process dependencies, all matched agents are deeply integrated with the atomic task execution sequence. Following the execution sequence logic of the atomic tasks, each atomic task is then... The corresponding intelligent agents are arranged in an orderly manner, and the connection relationships, data flow paths and task execution trigger conditions between the agents are sorted out. The execution results of the preceding agents are used as the input data of the following agents. The message passing rules and cooperation methods between agents are clarified, and finally, an intelligent agent execution sequence corresponding one-to-one with the atomic task execution sequence is generated. This sequence not only follows the business execution logic of the atomic task, but also clarifies the execution order, cooperation relationship and data interaction specifications of each agent. It provides clear and standardized execution guidance for subsequent multi-agent collaboration to complete the overall business requirements, and ensures the continuity of task execution, the accuracy of data flow and the efficiency of overall business processing during multi-agent collaboration.

[0055] Optionally, the low-code platform includes a visual interface, through which the low-code platform receives the user input.

[0056] In this optional embodiment, the low-code platform features a lightweight and easy-to-use visual interface. This interface is designed with a graphical, drag-and-drop interactive approach, eliminating the complex coding required in traditional development and catering to the operating habits of non-technical personnel. The low-code platform uses this visual interface to comprehensively receive user input. Users can submit business requirements through various methods, including text input, voice-to-text input, component drag-and-drop configuration, parameter form filling, and requirement template selection. It supports direct input of business analysis and feature development requirements in natural language, as well as fine-grained operation commands such as configuring domestic database connections, setting data processing rules, optimizing AI model parameters, and orchestrating task execution processes. Simultaneously, the visual interface performs real-time parsing, format validation, and visualization of user input, allowing users to intuitively view the input requirements, configuration parameters, and execution requirements. Users can also instantly modify, supplement, and confirm the input within the interface, ensuring that the user's input is accurately and completely captured and received by the low-code platform. This provides accurate and standardized raw input data for subsequent requirement parsing, entity generation, and intelligent agent task execution.

[0057] Optionally, the low-code platform further includes a database for storing the business results and providing training data for the low-code platform.

[0058] Specifically, the low-code platform also includes a domestically developed database as the core data storage and training data supply carrier. This database not only undertakes the core responsibility of storing various business results, but also provides multi-type, high-quality training data support for the entire AI model training process of the low-code platform. In terms of business result storage, the database can structure and standardize the storage of various business analysis reports, data statistics, model inference conclusions, and other business results generated by the low-code platform after intelligent agent collaborative processing. Furthermore, relying on its distributed architecture support, multi-model storage capabilities, and data lifecycle management mechanism, it can flexibly adopt relational, non-relational, or time-series storage methods according to the type, importance, and access frequency of the business results. It also divides the data into hot, warm, and cold data for hierarchical management, combined with scheduled full backups + The incremental log backup and recovery mechanism ensures the security, integrity, and traceability of business results storage, while also enabling efficient data access and long-term retention. Regarding training data supply, this database serves as the core data source for AI model training on low-code platforms. Through real-time data synchronization technology in a domestically developed database adaptation layer, it supports both full and incremental synchronization. It provides the platform with all the business data required for initial model training and also synchronizes the latest data changes generated in real-time from the database to the low-code platform. The synchronization latency is capped at 1 second, ensuring the freshness of the training data. Furthermore, the database fully supports databases such as DM, OceanBase, and GaussDB. It supports mainstream database types and provides multi-dimensional training data, including relational, non-relational, and time-series data, to meet the diverse data needs of different pre-trained models and training modes within the low-code platform. Furthermore, the low-code platform can use a visual data access and preprocessing module to perform intelligent data mapping, visual data cleaning, automated feature engineering, and data sampling on the raw data extracted from the database. This transforms the raw database data into standardized data that conforms to AI model training standards, providing accurate, adaptable, and high-quality training data support for various training operations such as model fine-tuning, full training, and incremental training on the low-code platform. This achieves an integrated service for data storage and model training data supply.

[0059] like Figure 2 As shown, an embodiment of the present invention provides an intelligent service processing device 200 based on a context protocol, comprising: Module 210 is used to obtain the current user demand entity and historical session log entity generated by the low-code platform based on user input; The processing module 220 is used to obtain the corresponding session state packet based on the current user demand entity and the historical session log entity through a preset context-aware strategy. The disassembly module 230 is used to disassemble tasks based on the session state packet to obtain the corresponding atomic task execution sequence; The module 240 is used to call the corresponding preset external tool to generate the corresponding structured execution result according to each atomic task in the atomic task execution sequence; The matching module 250 is used to generate an agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents, wherein the agent is determined by the matching result of the structured execution results and the preset agent role library; The execution module 260 is used to execute the corresponding atomic task according to the intelligent agent execution sequence to obtain the business result corresponding to the current user demand entity.

[0060] The context protocol-based intelligent service processing device of this embodiment is used to implement the context protocol-based intelligent service processing method as described above. Its advantages over the prior art are the same as the advantages of the context protocol-based intelligent service processing method over the prior art, and will not be repeated here.

[0061] like Figure 3 As shown in the figure, an electronic device 300 provided in this embodiment of the invention includes a memory 310 and a processor 320; the memory 310 is used to store a computer program; the processor 320 is used to implement the intelligent business processing method based on the context protocol as described above when the computer program is executed.

[0062] Alternatively, an electronic device 300 includes a memory 310 and a processor 320 coupled to the memory 310; the memory 310 is configured to store a computer program; and the processor 320 is configured to perform the following operations when the computer program is executed: Retrieve the current user request entity and historical session log entity generated by the low-code platform based on user input; Based on the current user demand entity and the historical session log entity, the corresponding session state packet is obtained through a preset context-aware strategy; The corresponding atomic task execution sequence is obtained by decomposing the session state packet. The corresponding structured execution result is generated by calling the corresponding preset external tool for each atomic task in the atomic task execution sequence; Based on all the structured execution results and the corresponding agents, an agent execution sequence corresponding to the atomic task execution sequence is generated, wherein the agent is determined by the matching result of the structured execution results and the preset agent role library; The business result corresponding to the current user demand entity is obtained by executing the corresponding atomic task according to the execution sequence of the intelligent agent.

[0063] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the intelligent business processing method based on the context protocol as described above.

[0064] Alternatively, a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the following operations: Retrieve the current user request entity and historical session log entity generated by the low-code platform based on user input; Based on the current user demand entity and the historical session log entity, the corresponding session state packet is obtained through a preset context-aware strategy; The corresponding atomic task execution sequence is obtained by decomposing the session state packet. The corresponding structured execution result is generated by calling the corresponding preset external tool for each atomic task in the atomic task execution sequence; Based on all the structured execution results and the corresponding agents, an agent execution sequence corresponding to the atomic task execution sequence is generated, wherein the agent is determined by the matching result of the structured execution results and the preset agent role library; The business result corresponding to the current user demand entity is obtained by executing the corresponding atomic task according to the execution sequence of the intelligent agent.

[0065] The present invention will now be described an electronic device 300 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. Electronic device 300 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 300 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0066] Electronic device 300 includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or a computer program loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0067] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. In this application, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention according to actual needs. Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units can be implemented in hardware or as software functional units.

[0068] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for intelligent business processing based on context protocols, characterized in that, include: Retrieve the current user request entity and historical session log entity generated by the low-code platform based on user input; Based on the current user demand entity and the historical session log entity, the corresponding session state packet is obtained through a preset context-aware strategy; The corresponding atomic task execution sequence is obtained by decomposing the session state packet. The corresponding structured execution result is generated by calling the corresponding preset external tool for each atomic task in the atomic task execution sequence; Based on all the structured execution results and the corresponding agents, an agent execution sequence corresponding to the atomic task execution sequence is generated, wherein the agent is determined by the matching result of the structured execution results and the preset agent role library; The business result corresponding to the current user demand entity is obtained by executing the corresponding atomic task according to the intelligent agent's execution sequence. The step of obtaining the corresponding session state packet based on the current user demand entity and the historical session log entity through a preset context-aware strategy includes: The current user demand entity and the historical session log entity are associated and matched to obtain the current complete entity graph; The session state packet is obtained by performing a state transition based on the current complete entity graph. The step of associating and matching the current user demand entity and the historical session log entity to obtain the current complete entity graph includes: Extract key information based on the current user demand entity; The corresponding supplementary information is obtained by comparing and matching the key information with the historical session log entity. Generate the current complete entity map based on the current user demand entity and the supplementary information; The process of obtaining the session state packet based on the current complete entity graph includes: Extract key contextual information based on the current complete entity graph; Based on the contextual key information, the corresponding state label is obtained through a preset state transition rule, wherein the state transition rule includes a one-to-one correspondence between the contextual key information and the state label; The session state package is generated based on all the contextual key information and the corresponding state labels.

2. The intelligent service processing method based on context protocol according to claim 1, characterized in that, The step of decomposing the task based on the session state packet to obtain the corresponding atomic task execution sequence includes: Extract atomic tasks based on the session state packet; An atomic task execution sequence is generated based on the dependencies of all atomic tasks, wherein the dependencies include data dependencies and business process dependencies of the atomic tasks.

3. The intelligent service processing method based on context protocol according to claim 1, characterized in that, The step of calling corresponding preset external tools to generate corresponding structured execution results based on each atomic task in the atomic task execution sequence includes: Determine the corresponding atomic task type based on the atomic task; The corresponding preset external tool is determined according to the atomic task type and the preset tool matching rule, wherein the tool matching rule includes a one-to-one correspondence between the atomic task type and the preset external tool; The structured execution result corresponding to the atomic task is generated based on the preset external tool.

4. The intelligent service processing method based on context protocol according to claim 1, characterized in that, The step of generating the agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents includes: Based on the structured execution result, match the agent corresponding to the atomic task in the agent role library; The corresponding agent execution sequence is generated based on the atomic task execution sequence and all the corresponding agents.

5. The intelligent service processing method based on context protocol according to claim 1, characterized in that, The low-code platform includes a visual interface, through which the low-code platform receives user input.

6. The intelligent service processing method based on context protocol according to claim 5, characterized in that, It also includes a database for storing the business results and providing training data for the low-code platform.

7. A context-based intelligent service processing device, characterized in that, include: The acquisition module is used to acquire the current user request entity and historical session log entity generated by the low-code platform based on user input; The processing module is used to obtain the corresponding session state packet based on the current user demand entity and the historical session log entity through a preset context-aware strategy. The step of obtaining the corresponding session state packet based on the current user demand entity and the historical session log entity using a preset context-aware strategy includes: The current user demand entity and the historical session log entity are associated and matched to obtain the current complete entity graph; The session state packet is obtained by performing a state transition based on the current complete entity graph. The step of associating and matching the current user demand entity and the historical session log entity to obtain the current complete entity graph includes: Extract key information based on the current user demand entity; The corresponding supplementary information is obtained by comparing and matching the key information with the historical session log entity. Generate the current complete entity map based on the current user demand entity and the supplementary information; The process of obtaining the session state packet based on the current complete entity graph includes: Extract key contextual information based on the current complete entity graph; Based on the contextual key information, the corresponding state label is obtained through a preset state transition rule, wherein the state transition rule includes a one-to-one correspondence between the contextual key information and the state label; The session state package is generated based on all the context key information and the corresponding state tags; The decomposition module is used to decompose tasks based on the session state packet to obtain the corresponding atomic task execution sequence; The calling module is used to call the corresponding preset external tools to generate the corresponding structured execution results according to each atomic task in the atomic task execution sequence; A matching module is used to generate an agent execution sequence corresponding to the atomic task execution sequence based on all the structured execution results and the corresponding agents, wherein the agent is determined by the matching results of the structured execution results and a preset agent role library; The execution module is used to execute the corresponding atomic task according to the execution sequence of the intelligent agent to obtain the business result corresponding to the current user demand entity.

Images