Task processing method and device based on data sharing, equipment and medium

Abstract

This invention relates to the field of intelligent decision-making technology, and discloses a task processing method, apparatus, device, and medium based on data sharing. The method includes: determining the process node of the target task in the task flow data based on initial input data; calculating the task value data of the target task based on initial task value data and task operation indicators; determining whether the task value data is less than the task termination threshold; if less, terminating the target task according to the process processing instruction and generating the task termination result; if greater than or equal to the threshold, analyzing the node data status of the process node in the data sharing center and determining the node data scheduling path of the process node based on the node data status; and driving the intelligent agent corresponding to the process node to process the target task based on the node data scheduling path to obtain the task processing result. This invention can be applied to business system platforms such as fintech and healthcare, and can improve the efficiency of multi-agent collaborative task processing.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making technology, and in particular to a task processing method, apparatus, device, and medium based on data sharing. Background Technology

[0002] With the rapid development of artificial intelligence technology, multi-agent collaborative systems have been widely used in complex business scenarios. Most existing agent collaboration mechanisms adopt a rigid chain execution mode, where agents collaborate by directly transmitting raw data or processing results. This lack of a unified task scheduling and knowledge sharing mechanism leads to redundant calculations, resource waste, lack of dynamic evaluation of task value, and lack of efficient intermediate result sharing and reuse mechanisms when the system is processing complex tasks.

[0003] In the fintech field, multi-agent collaborative systems can be applied to the entire credit approval process, where multiple units, such as customer information collection agents, risk assessment agents, and compliance review agents, work together to complete the approval task.

[0004] In the field of healthcare, multi-agent collaborative systems can serve the multi-dimensional monitoring and treatment assistance of hospitalized patients, and consist of a treatment assistance system composed of vital sign monitoring agents, medical record analysis agents, and medication recommendation agents.

[0005] Therefore, existing multi-agent collaborative systems have significant shortcomings in resource scheduling, task evaluation, and knowledge sharing, making it difficult to meet the demand for efficient, low-cost, and intelligent collaboration in complex business scenarios, ultimately resulting in low efficiency in task processing by multi-agent collaborative systems. Summary of the Invention

[0006] This invention provides a task processing method, apparatus, device, and medium based on data sharing to solve the technical problems of task processing based on data sharing.

[0007] Firstly, a task processing method based on data sharing is provided, including: Obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data; Obtain the task operation indicators of the process node, and calculate the task value data of the target task based on the initial task value data and the task operation indicators; When the task value data is less than a preset task termination threshold, the target task is terminated according to a preset process processing instruction, and a task termination result for the target task is generated according to the process processing instruction. When the task value data is greater than or equal to the task termination threshold, the node data status of the process node is analyzed in the pre-acquired data sharing center, and the node data scheduling path of the process node is determined based on the node data status. Based on the node data scheduling path, the preset intelligent agent corresponding to the process node is driven to process the target task and obtain the task processing result.

[0008] Secondly, a task processing device based on data sharing is provided, comprising: The process node determination module is used to obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data. The task value data calculation module is used to obtain the task operation indicators of the process node and calculate the task value data of the target task based on the initial task value data and the task operation indicators. The task termination result generation module is used to terminate the target task according to a preset process processing instruction when the task value data is less than a preset task termination threshold, and to generate the task termination result of the target task according to the process processing instruction. The node data status analysis module is used to analyze the node data status of the process node in the pre-acquired data sharing center when the task value data is greater than or equal to the task termination threshold, and to determine the node data scheduling path of the process node based on the node data status. The task result processing module is used to drive the preset intelligent agent corresponding to the process node to process the target task based on the node data scheduling path, and obtain the task processing result.

[0009] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described task processing method based on data sharing.

[0010] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described task processing method based on data sharing.

[0011] In the aforementioned solution implemented by the data-sharing-based task processing method, device, equipment, and medium, the client can utilize a dynamic task value assessment mechanism to calculate the remaining expected revenue in real time during task execution. When the task value falls below a preset threshold, intelligent termination is automatically triggered, avoiding the allocation of computing resources to low-value or high-risk task segments, thereby significantly reducing ineffective resource consumption. Simultaneously, the knowledge-sharing intermediary layer analyzes the data status of process nodes, intelligently determining whether to reuse intermediate results from the cache or call the agent to recalculate, avoiding repetitive LLM (Large Language Model) calls and token waste in multi-agent collaboration. Based on the optimized data scheduling path, the agent executes tasks, ensuring processing quality while achieving refined scheduling and global savings of computing resources, ultimately reducing operating costs and improving task processing efficiency. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of an application environment for a task processing method based on data sharing according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a task processing method based on data sharing in one embodiment of the present invention; Figure 3 yes Figure 2 A flowchart illustrating a specific implementation method of step S1; Figure 4 yes Figure 2 A flowchart illustrating a specific implementation of step S4; Figure 5 This is a schematic diagram of a task processing device based on data sharing in one embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 7 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] The data-sharing-based task processing method provided in this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can utilize a dynamic task value assessment mechanism to calculate the remaining expected revenue in real time during task execution. When the task value falls below a preset threshold, intelligent termination is automatically triggered, avoiding the allocation of computing resources to low-value or high-risk task segments, thus significantly reducing ineffective resource consumption. Simultaneously, a knowledge-sharing intermediary layer analyzes the data status of process nodes, intelligently determining whether to reuse intermediate results from the cache or call the agent to recalculate, avoiding repetitive LLM calls and token waste in multi-agent collaboration. Based on optimized data scheduling paths, the agent executes tasks, ensuring processing quality while achieving refined scheduling and global resource conservation, ultimately reducing operating costs and improving task processing efficiency. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The following detailed description of specific embodiments further illustrates this invention.

[0016] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a data-sharing-based task processing method provided in an embodiment of the present invention includes the following steps: S1. Obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data.

[0017] In this embodiment of the invention, the target task refers to an independent business request (such as a contract review request) initiated by a user or the system, which requires collaboration between one or more intelligent agents. The initial input data is the raw data information carried when the target task is initiated, such as form data submitted by the user, uploaded documents, business context information, etc. The initial task value data is the estimated commercial value or priority weight of the target task at the time of initiation, used to guide subsequent resource allocation strategies.

[0018] In detail, during the target task initiation phase, the API gateway can receive raw requests from the user interface or external systems, extract structured and unstructured information from the raw requests to form the initial input data, and then perform static calculations on the initial value of the target task based on the task parameters in the initial input data. For example, in a credit approval business request, the initial input data can be used to extract "loan amount applied for" as 1 million yuan and "expected annualized profit rate" as 5%, and combined with the historical user rating weights in the historical database (e.g., 1.2), the initial task value data = application amount × profit rate × customer rating weight (i.e., 1 million × 5% × 1.2) can be statically calculated to obtain the initial task value data (VInitial) of the target task, which serves as the benchmark for subsequent task value evaluation.

[0019] In the embodiments of the present invention, see Figure 3 As shown, determining the flow node of the target task in the preset task flow data based on the initial input data includes: S31. Extract the task type and task parameters from the initial input data; S32. Match the task type with the task flow type in the preset task flow data to determine the task flow type of the target task; S33. Obtain the sequence of process nodes corresponding to the task process type, and map the task parameters to the process nodes corresponding to the sequence of process nodes.

[0020] In detail, the task type identifier and corresponding fields for task parameters can be extracted from the initial input data using predefined JSON path expressions or regular expressions. The task type refers to an enumeration value used to distinguish different business scenarios, such as "credit approval," "contract review," or "customer complaint handling." The task parameters are the specific business data required to execute the target task. For example, in the "credit approval" business scenario, the task parameters include the loan amount, customer ID, and the file path of the asset report.

[0021] Specifically, the extracted task type (such as "credit approval") is used as a keyword to search for matching task flow types in the preset task flow data. Since the task type is a label for the business scenario, and the task flow type is the specific implementation path of that scenario, the two are associated through a pre-configured mapping relationship (such as a task type named "credit approval" being mapped to the task flow type "standard credit approval process" by default), thereby determining the sequence of process nodes to be executed for the target task.

[0022] In this embodiment of the invention, the preset task flow data defines processing templates for various business scenarios. The task flow type is an entry in the task flow data. The task flow type represents a sequence of process nodes consisting of multiple ordered process nodes. For example, the "standard credit approval process" includes four process nodes: [Customer Information Collection] → [Risk Assessment] → [Compliance Review] → [Loan Disbursement].

[0023] Furthermore, after determining the sequence of process nodes for the target task, based on the predefined input parameter requirements (i.e., node configuration information) for each process node in the sequence, corresponding data items are extracted from the task parameters and assembled into an input data package for the process node. For example, for the "risk assessment node," "customer ID" and "asset reporting path" are extracted from the task parameters, packaged, and used as the input for that node; while the input required for the "compliance review node" may also include "loan amount." The input data package is temporarily used as the data to be processed by downstream nodes, while the current process node, i.e., the initial node (such as the "customer information collection node"), directly obtains the complete task parameters as input. A process node refers to an executable unit within a task process type. Each executable unit corresponds to a specific intelligent agent and its processing function; for example, the "risk assessment node" corresponds to a risk control agent.

[0024] For example, in the intelligent credit approval scenario in the fintech field, when a customer submits an application for a business loan of 1 million yuan, the initial task value data is calculated to be 60,000 yuan based on the application amount, customer level and expected profit rate, and the "standard credit approval process" is automatically matched, mapping task parameters such as customer ID card, bank statements, credit report and so on to the "customer information collection node" as initial input.

[0025] For example, in the field of healthcare, when a hospital submits a medical record of a suspected rare disease, the initial task value data is calculated based on the complexity of the medical record, the weight of its scientific research value, and the urgency of the diagnosis. This data is then matched with a "multidisciplinary consultation process," and the patient's gene sequencing data, imaging data, and medical history records are mapped to a "gene analysis node" as initial input to initiate the intelligent agent collaborative judgment process.

[0026] In this embodiment of the invention, static calculation of initial task value data provides a benchmark reference for subsequent dynamic evaluation, enabling the establishment of a cost-benefit analysis framework at the task initiation stage and avoiding excessive computational resources invested in low-value tasks.

[0027] S2. Obtain the task operation indicators of the process node, and calculate the task value data of the target task based on the initial task value data and the task operation indicators.

[0028] In this embodiment of the invention, the task execution index is real-time data generated during the execution of process nodes, used to quantify the execution status and results of process nodes.

[0029] In detail, once a process node (such as the "risk assessment node" corresponding to the risk control agent) completes the processing of the input data, a post-processor can intercept the agent's response and extract predefined key fields. For example, in credit approval, after the risk control agent completes its analysis, in addition to outputting a detailed report, it will also return a standardized data packet (containing two task execution metrics: one is the task failure probability, the failure probability predicted by the risk control agent; the other is the process node execution cost, recording the number of tokens or API fees consumed by the risk control agent in this call). These execution metrics are bound to the node identifier of the current process node and stored in the knowledge-sharing intermediary layer.

[0030] Specifically, after each process node is completed, the value assessment model reads the task performance metrics generated by the target task at that process node from the knowledge-sharing intermediary layer, including the predicted failure probability and execution cost of the current process node, and calculates the cumulative cost. Then, combined with the initial task value data, it is substituted into the dynamic task value assessment formula:

[0031] in, For the first Task value data for each process node. This is the initial task value data. For the first The predicted failure probability of each process node. For the first The execution cost of each process node. and All are indexes.

[0032] The final calculated result is the task value data of the target task at the current process node. It is a dynamically updated quantitative indicator that comprehensively reflects the remaining expected benefits of the task and is used to guide the system to continue to invest resources, divert paths, or terminate the task immediately.

[0033] For example, in the intelligent recommendation task of wealth management products in the field of fintech, when calculating the execution cost of the recommendation strategy analysis node, not only are the token consumption and API fees of LLM calls counted, but also the execution time of strategy analysis and the computing power resources occupied are included in the implicit costs. The value of the task is evaluated after the weighted calculation of the cumulative cost, so as to avoid the continued execution of recommendation tasks with high time consumption and low returns due to ignoring implicit costs.

[0034] For example, in the intelligent medical image judgment task in the healthcare field, when calculating the execution cost of the image analysis node, the value of the task can be accurately calculated by combining the token consumption of image recognition, algorithm call fees, and implicit costs such as GPU computing power occupation time and data transmission time.

[0035] In this embodiment of the invention, by introducing a dynamic task value assessment formula, the real-time tracking of the economic value of tasks is realized, enabling the system to make decisions based on quantitative indicators rather than fixed rules, achieving full-dimensional quantification of the execution cost of process nodes, improving the accuracy of task value data calculation, and avoiding resource allocation decision deviations caused by a single cost dimension.

[0036] S3. Determine whether the task value data is less than a preset task termination threshold.

[0037] In this embodiment of the invention, the preset task termination threshold is a task value boundary dynamically configured according to the task type, used to determine whether continuing to execute subsequent process nodes is economically reasonable.

[0038] S4. If the task value data is less than the preset task termination threshold, the target task is terminated according to the preset process processing instruction, and the task termination result of the target task is generated according to the process processing instruction.

[0039] In this embodiment of the invention, the preset process processing instructions are a set of control instructions, including a termination type code, a resource recycling identifier, and an external output template. The termination type code is an enumerated value preset according to the specific reason for triggering the termination. For example, when the termination is caused by a high probability of task failure, the code corresponds to "high-risk termination"; when the termination is caused by the accumulated resource consumption cost exceeding the expected benefit, the code corresponds to "cost over-limit termination". The task termination result is the final output including termination reason tracing information, details of consumed resources, and alternative processing suggestions.

[0040] In detail, the system identifies trigger events where the task value data falls below a preset task termination threshold, invokes a process processing instruction matching the termination reason, immediately interrupts the subsequent execution link of the preset intelligent agent corresponding to the current process node, and blocks the flow of the target task to the next process node.

[0041] In the embodiments of the present invention, see Figure 4 As shown, generating the task termination result of the target task according to the process processing instructions includes: S41. Extract the task execution indicators of the process node corresponding to the process processing instruction. S42. Obtain the termination type code in the process processing instruction, and generate a task termination description based on the termination type code and the task operation indicators; S43. The task termination description is concatenated with the task operation indicators to obtain the task termination result of the target task.

[0042] In detail, task operation metrics are extracted from the task execution logs. Simultaneously, the list of node storage data generated by the target task and its corresponding caching costs (e.g., "caching space occupied 2MB, caching cost 0.01 yuan") are queried from the knowledge-sharing intermediary layer. All of the above data is then aggregated to form a complete set of task operation metrics. These task operation metrics refer to the real-time data set generated by each process node during the execution of the target task, used to quantify the execution status and results. This includes node identifiers (e.g., "risk assessment node_N2"), preset intelligent agent identifiers (e.g., "risk control Agent_V3"), task failure probability (e.g., "0.85"), current node resource consumption cost (e.g., "this call consumed 5000 Tokens, equivalent to a cost of 0.5 yuan"), cumulative resource consumption cost (e.g., "cumulative consumption of 12000 Tokens, equivalent to a cost of 1.2 yuan"), and a list of node storage data (e.g., "generated customer profile data, credit summary data").

[0043] Specifically, the termination type code in the parsing process instructions is used. When the termination type code is "high-risk termination", a task termination description is generated by combining the task operation indicators. The task termination description includes the specific numerical value of the task failure probability (e.g., "default probability 85%"), the analysis of the main risk dimensions (e.g., "credit risk score 45 points, below the passing score of 60 points"), and the termination suggestion (e.g., "suggest transferring to manual review or requiring supplementary collateral"). When the termination reason identifier is "cost over-limit termination", the task termination description includes the specific numerical value of the accumulated execution cost, the remaining proportion of expected revenue (e.g., "costs consumed account for 82% of expected revenue"), and the termination suggestion (e.g., "suggest adopting a simplified process or retrying with a reduced limit").

[0044] Furthermore, the task termination description and task operation metrics are assembled according to a preset structured template to generate a standardized task termination result. The task termination result includes the following fields: termination type code, termination node position, number of tokens consumed, number of API calls, number of intermediate result entries generated, and suggested low-cost alternative processing solutions.

[0045] For example, in the fintech field, a lower preset task termination threshold is set for high-yield corporate financing approval tasks, while a higher termination threshold is set for low-yield personal microloan approval tasks. At the same time, the termination thresholds of different types of credit tasks are iteratively optimized based on monthly historical approval task data to reduce the false termination of effective corporate financing tasks.

[0046] For example, in the healthcare field, a lower termination threshold is set for remote consultation tasks for complex and difficult diseases, while a higher termination threshold is set for online consultation tasks for common and frequently occurring diseases, and the thresholds are optimized based on historical treatment task data each quarter.

[0047] In this embodiment of the invention, the linkage between dynamic task value assessment and termination mechanism enables early identification and rapid loss prevention of invalid tasks, making the task termination threshold match the task type with different business attributes, and improving the rationality of intelligent pruning.

[0048] S5. If the task value data is greater than or equal to the task termination threshold, the node data status of the process node is analyzed in the pre-acquired data sharing center, and the node data scheduling path of the process node is determined based on the node data status.

[0049] In this embodiment of the invention, the data sharing center refers to the knowledge sharing intermediary layer deployed in the multi-agent collaborative system as the sole channel for communication and collaboration among all agents, responsible for centrally managing the node storage data generated by each agent when processing tasks in the collaborative process; the node data status is a data availability identifier, including a first state and a second state, wherein the first state indicates that the node storage data is complete, valid and conflict-free and can be directly scheduled and used, and the second state indicates that the node storage data is missing, expired or has logical conflicts and needs to be regenerated by the agent.

[0050] In this embodiment of the invention, analyzing the node data status of the process node within the pre-acquired data sharing center includes: The node configuration information of the process node is parsed, and a query instruction for the process node is generated based on the node configuration information; Use the query command to query the node data type of the process node within the pre-acquired data sharing center; Extract the node storage data corresponding to the node data type from the data sharing center; The data status of the nodes is obtained by performing data status analysis on the stored data of the nodes.

[0051] In detail, the node configuration information corresponding to the process node is parsed to obtain the input data specifications required by the process node. The input data specifications include data type identifier, data format requirements, and data source agent identifier. A query instruction is generated based on the data type identifier. By calling the standard query interface provided by the data sharing center, the index database of the data sharing center is searched to see if there is an index tag that matches the data type identifier and is associated with the data source agent identifier. If a matching index tag is found, the node data type is marked as having candidate node storage data. If no matching item is found, the node storage data corresponding to the node data type is marked as missing, and the node data status is marked as the second state.

[0052] Specifically, when candidate node storage data exists for a node data type, the physical storage address in the storage layer of the data sharing center is located based on the retrieved index tag. The data sharing center adopts a key-value pair storage structure, where the key is the hash value of the index tag and the value is the serialized content of the node storage data. The serialized node storage data is read through the physical storage address.

[0053] Furthermore, a triple check is performed on the extracted node storage data to determine the node data status. First, an integrity check is performed to check whether the data structure of the node storage data is complete and whether the required fields exist and are not empty. If the integrity check fails, it is directly determined to be in the second state. Second, a timeliness check is performed to compare the current system time with the generation timestamp of the node storage data. If the time difference exceeds the preset validity period threshold, the data is determined to be expired and marked as the second state. Finally, a logical consistency check is performed. The system calls the conflict verification module to detect logical conflicts between the key conclusions in the node storage data and the current task context. If a contradiction with the conclusions of the preceding node or a violation of business rules is detected, it is marked as the second state. Only when the integrity check, the timeliness check, and the logical consistency check are all passed, is the node data status determined to be in the first state.

[0054] In this embodiment of the invention, the node data scheduling path refers to the data acquisition and processing channel determined based on the determination result of the node data status.

[0055] In this embodiment of the invention, determining the node data scheduling path of the process node based on the node data status includes: When the node data status is a first preset status, a data transmission channel is established between the data sharing center and the process node; The data transmission channel is determined as the reuse path of the process node, and the reuse path is determined as the node data scheduling path of the process node; When the node data status is the second preset status, the computing channel of the agent corresponding to the process node is determined as the execution path of the process node; The execution path is determined as the node data scheduling path of the process node.

[0056] In detail, when the node data status is in the first preset state, a direct data transmission channel is established from the data sharing center to the current process node. The extracted and verified node storage data is encapsulated into a standard data packet format and transmitted to the input buffer of the process node through the internal message bus. After receiving the data, the process node parses and loads it into its working memory, and directly executes the business logic processing of this node based on the process node storage data without triggering LLM calls or complex calculations. This reuse operation is recorded in the scheduling log, and the reference count and last access time of the process node storage data are updated.

[0057] Specifically, the corresponding agent is determined based on the node configuration information of the process node. A processing request is initiated to the agent through the agent call interface. The agent performs LLM analysis or algorithm operation based on the original input data to generate new node storage data. The newly generated node storage data is sent to the data sharing center and stored in the corresponding process node index to complete the cache update. The complete channel that triggers the agent operation and updates the cache is the node data scheduling path. The original input data includes the initial input data of the target task and the node storage data generated by the preceding process node.

[0058] For example, in the fintech field, when performing a logical consistency check on the intermediate risk assessment results generated by the risk control intelligent agent, the business rules engine verifies whether the assessment results comply with the industry rules for insurance underwriting. At the same time, a similarity algorithm is used to compare the similarity between the results and the customer data conclusions from the customer information collection node, so as to avoid logical conflicts such as customer information showing no past medical history but risk assessment results marking a history of serious illness.

[0059] For example, in the healthcare field, when checking the intermediate results of prescription rationality analysis generated by a pharmaceutical intelligence agent, the intermediate results are verified according to business rules to see if they comply with drug incompatibility rules. At the same time, the similarity between the results and the disease conclusions of the doctor's judgment nodes is compared to avoid logical conflicts in the results.

[0060] In this embodiment of the invention, by introducing a knowledge-sharing intermediary layer as a unified data exchange hub, the data silos between intelligent agents are broken down, and the structured caching and intelligent reuse of intermediate results are realized. At the same time, a triple verification mechanism ensures the accuracy and consistency of the reused data.

[0061] S6. Based on the node data scheduling path, drive the preset intelligent agent corresponding to the process node to process the target task and obtain the task processing result.

[0062] In this embodiment of the invention, the preset intelligent agent refers to a pre-configured autonomous computing unit with specific domain task processing capabilities. Each intelligent agent is given a clear responsibility boundary and functional positioning, and is used to perform specific analysis, decision-making or data processing operations in the target task process. The preset intelligent agents include, but are not limited to, customer information collection intelligent agents, risk assessment intelligent agents, compliance verification intelligent agents, etc. Each intelligent agent communicates collaboratively through a knowledge sharing intermediary layer to avoid data redundancy and coupling problems caused by direct interaction.

[0063] In this embodiment of the invention, the step of driving the preset intelligent agent corresponding to the process node to process the target task based on the node data scheduling path and obtaining the task processing result includes: If the node data scheduling path is a reused path, then the node storage data corresponding to the process node is extracted from the data sharing center, and the node storage data is used as the node task result of the process node. If the node data scheduling path is an execution path, then the preset intelligent agent corresponding to the process node is driven to perform task analysis on the target task and generate the node task result of the process node. The node task results are associated with the node identifiers corresponding to the process nodes and stored in the data sharing center; Determine whether there are any unexecuted process nodes in the process node sequence corresponding to the task process type for the target task; If it exists, return to the step of obtaining the task execution metrics of the process node; If it does not exist, then extract the node task results of all process nodes from the data sharing center, and summarize the node task results to obtain the task processing result of the target task.

[0064] In detail, when the data scheduling path of a process node is a reuse path, a data query key value is generated based on the data type and content characteristics required by the process node. Then, a node storage data query request is sent to the knowledge sharing intermediary layer. The cache database is retrieved based on the query key value. If it exists, the generation timestamp, source node identifier, and logical consistency tag of the node storage data are further verified. After confirming the validity and availability, the node storage data is extracted from the cache and returned to the current process node, thereby obtaining the node task result of the current process node, realizing the saving of computing resources and the improvement of processing efficiency.

[0065] In this embodiment of the invention, the node task result is the specific output of the intelligent agent corresponding to the process node after processing the input data, including but not limited to structured data, text summaries, decision conclusions, risk assessment scores, or node analysis reports.

[0066] In this embodiment of the invention, the step of driving the preset intelligent agent corresponding to the process node to perform task analysis on the target task and generate the node task result of the process node includes: Obtain the node input data of the process node and analyze the context information of the process node; The node input data and the context information are encapsulated into an input data packet, and the input data packet is sent to the preset intelligent agent corresponding to the process node; The input data packet is analyzed using the analysis model in the intelligent agent to obtain the node task results of the process node.

[0067] In detail, the raw input data required for the process nodes is extracted from the data storage area of ​​the target task, including the corresponding fields in the initial input data and the node storage data generated by the preceding nodes; at the same time, the task context cache of the target task is read to obtain the task identifier, the position index of the current process node in the task process data, the processing records of executed nodes, and the update values ​​of the task value data.

[0068] In this embodiment of the invention, the extracted node input data and task context information are serialized according to the input interface specification predefined by the agent, assembled into a structured data packet that meets the requirements of the agent's call, and then routed to the agent's message queue or call interface through the internal service bus or API gateway to trigger the agent's execution.

[0069] Specifically, the data packets are deserialized and format parsed to restore the node input data and task context information. Then, the agent calls its built-in analysis model, which includes a large language model, machine learning model, or business rule engine, to perform specific analysis operations on the input data, including text summarization, structured data extraction, risk assessment scoring, or compliance verification. After outputting the original analysis results, the agent's post-processing module standardizes and encapsulates the results, adds confidence scores, processing cost statistics, and failure probability predictions, and generates structured node task results.

[0070] In this embodiment of the invention, the node identifier that has been completed is compared with the process node sequence to identify whether there are any unexecuted nodes after the current node; if there are subsequent nodes in the node sequence, it is determined that there are unexecuted subsequent process nodes, and the identifier and related configuration information of the next node are obtained; if the current node is the last node in the sequence, it is determined that there are no subsequent nodes.

[0071] Furthermore, if there are subsequent process nodes, the processing information generated by the current process node, including the failure probability and cost consumption, as well as the task value data of the current process node, are obtained. The task value data of the target task is recalculated using the dynamic evaluation formula for task value, resulting in an updated task value data value. Then, the updated task value data is compared with the task termination threshold. If it is lower than the threshold, the process is terminated. If it is higher than the threshold, the status of the target task is updated to the pending status of the next process node. The updated task data, context information, and task value data are then passed to the next process node, and the next round of data scheduling and processing is repeated.

[0072] In this embodiment of the invention, if there are no subsequent process nodes, all task processing results generated by the target task at each process node are extracted from the knowledge sharing intermediary layer according to the node execution order. These results include intermediate results obtained from the reuse path and original results generated by the execution path. The processing results of each node are integrated and formatted according to the preset task output template to form a complete final processing result of the target task, such as a complete credit approval report containing customer information, risk assessment conclusions, and compliance verification opinions.

[0073] For example, in the fintech field, when generating the final assessment report for a corporate credit assessment task, the results are arranged according to the execution order of customer information collection, financial data review, and risk level assessment, with the highest weight assigned to the risk level assessment node, and the processing results of different formats such as structured data and text reports are converted into a unified PDF assessment report format.

[0074] For example, in the field of healthcare, when generating the final interpretation report for the intelligent interpretation task of physical examination reports, the results are arranged according to the execution order of basic physical examination data collection, specialized indicator analysis, and disease risk prediction. The specialized indicator analysis node is assigned the highest weight, and numerical data, image analysis results, and textual conclusions are converted into a unified structured physical examination interpretation report format.

[0075] In this embodiment of the invention, by automatically feeding back the task processing results to the knowledge sharing intermediary layer for caching, a continuously accumulating intelligent knowledge base is constructed, enabling the system to achieve cross-task and cross-session knowledge reuse when processing similar tasks, forming a positive cycle of intelligent acceleration effect.

[0076] As can be seen, in the above scheme, intelligent task pruning is achieved by dynamically calculating task value and combining it with threshold judgment, avoiding ineffective resource investment; relying on the data sharing center to analyze data status and plan scheduling paths, intermediate results are efficiently reused, reducing redundant calculations and resource waste; at the same time, intelligent agents are driven to process tasks according to the scheduling path, making multi-agent collaboration more in line with the actual value of the task and the data status, and greatly improving the task processing efficiency of the collaborative system.

[0077] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0078] In one embodiment, a data-sharing-based task processing apparatus is provided, which corresponds one-to-one with the data-sharing-based task processing methods described in the above embodiments. For example... Figure 5 As shown, the data-sharing-based task processing device 100 includes a process node determination module 101, a task value data calculation module 102, a task termination result generation module 103, a node data status analysis module 104, and a task result processing module 105. Detailed descriptions of each functional module are as follows: The process node determination module 101 is used to obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data. The task value data calculation module 102 is used to obtain the task operation indicators of the process node and calculate the task value data of the target task based on the initial task value data and the task operation indicators. The task termination result generation module 103 is used to terminate the target task according to a preset process processing instruction when the task value data is less than a preset task termination threshold, and to generate the task termination result of the target task according to the process processing instruction. The node data status analysis module 104 is used to analyze the node data status of the process node in the pre-acquired data sharing center when the task value data is greater than or equal to the task termination threshold, and determine the node data scheduling path of the process node based on the node data status. The task result processing module 105 is used to drive the preset intelligent agent corresponding to the process node to process the target task based on the node data scheduling path, and obtain the task processing result.

[0079] In one embodiment, the process node determination module 101, when performing the function of determining the process node of the target task in the preset task process data based on the initial input data, is configured to: Extract the task type and task parameters from the initial input data; The task type is matched with the task flow type in the preset task flow data to determine the task flow type of the target task. Obtain the sequence of process nodes corresponding to the task process type, and map the task parameters to the process nodes corresponding to the sequence of process nodes.

[0080] In one embodiment, the task termination result generation module 103, when executing the process to generate the task termination result of the target task according to the process instructions, is used to: Extract the task execution metrics of the process nodes corresponding to the process processing instructions. Obtain the termination type code from the process processing instruction, and generate a task termination description based on the termination type code and the task operation indicators; The task termination description is concatenated with the task operation indicators to obtain the task termination result of the target task.

[0081] In one embodiment, the node data status analysis module 104, when performing the analysis of the node data status of the process node within the pre-acquired data sharing center, is used to: The node configuration information of the process node is parsed, and a query instruction for the process node is generated based on the node configuration information; Use the query command to query the node data type of the process node within the pre-acquired data sharing center; Extract the node storage data corresponding to the node data type from the data sharing center; The data status of the nodes is obtained by performing data status analysis on the stored data of the nodes.

[0082] In one embodiment, the node data status analysis module 104, when performing the node data scheduling path determination of the process node based on the node data status, is used to: When the node data status is a first preset status, a data transmission channel is established between the data sharing center and the process node; The data transmission channel is determined as the reuse path of the process node, and the reuse path is determined as the node data scheduling path of the process node; When the node data status is the second preset status, the computing channel of the agent corresponding to the process node is determined as the execution path of the process node; The execution path is determined as the node data scheduling path of the process node.

[0083] In one embodiment, the task result processing module 105, when executing the preset intelligent agent corresponding to the process node based on the node data scheduling path to process the target task and obtain the task processing result, is used to: If the node data scheduling path is a reused path, then the node storage data corresponding to the process node is extracted from the data sharing center, and the node storage data is used as the node task result of the process node. If the node data scheduling path is an execution path, then the preset intelligent agent corresponding to the process node is driven to perform task analysis on the target task and generate the node task result of the process node. The node task results are associated with the node identifiers corresponding to the process nodes and stored in the data sharing center; Determine whether there are any unexecuted process nodes in the process node sequence corresponding to the task process type for the target task; If it exists, return to the step of obtaining the task execution metrics of the process node; If it does not exist, then extract the node task results of all process nodes from the data sharing center, and summarize the node task results to obtain the task processing result of the target task.

[0084] In one embodiment, the task result processing module 105, when executing the preset intelligent agent corresponding to the process node to perform task analysis on the target task and generate the node task result of the process node, is used to: Obtain the node input data of the process node and analyze the context information of the process node; The node input data and the context information are encapsulated into an input data packet, and the input data packet is sent to the preset intelligent agent corresponding to the process node; The input data packet is analyzed using the analysis model in the intelligent agent to obtain the node task results of the process node.

[0085] This invention provides a task processing device based on data sharing. It achieves intelligent task pruning by dynamically calculating task value and combining it with threshold judgment, avoiding ineffective resource investment. It relies on the data sharing center to analyze data status and plan scheduling paths, enabling efficient reuse of intermediate results and reducing redundant calculations and resource waste. At the same time, it drives intelligent agents to process tasks according to the scheduling path, making multi-agent collaboration more in line with the actual value of the task and the data status, and greatly improving the task processing efficiency of the collaborative system.

[0086] Specific limitations regarding the data-sharing-based task processing device can be found in the limitations of the data-sharing-based task processing method described above, and will not be repeated here. Each module in the aforementioned data-sharing-based task processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0087] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a data-sharing-based task processing method on the server side.

[0088] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 7 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of a data-sharing-based task processing method.

[0089] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data; Obtain the task operation indicators of the process node, and calculate the task value data of the target task based on the initial task value data and the task operation indicators; When the task value data is less than a preset task termination threshold, the target task is terminated according to a preset process processing instruction, and a task termination result for the target task is generated according to the process processing instruction. When the task value data is greater than or equal to the task termination threshold, the node data status of the process node is analyzed in the pre-acquired data sharing center, and the node data scheduling path of the process node is determined based on the node data status. Based on the node data scheduling path, the preset intelligent agent corresponding to the process node is driven to process the target task and obtain the task processing result.

[0090] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data; Obtain the task operation indicators of the process node, and calculate the task value data of the target task based on the initial task value data and the task operation indicators; When the task value data is less than a preset task termination threshold, the target task is terminated according to a preset process processing instruction, and a task termination result for the target task is generated according to the process processing instruction. When the task value data is greater than or equal to the task termination threshold, the node data status of the process node is analyzed in the pre-acquired data sharing center, and the node data scheduling path of the process node is determined based on the node data status. Based on the node data scheduling path, the preset intelligent agent corresponding to the process node is driven to process the target task and obtain the task processing result.

[0091] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0092] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0093] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0094] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0095] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A task processing method based on data sharing, characterized in that, include: Obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data; Obtain the task operation indicators of the process node, and calculate the task value data of the target task based on the initial task value data and the task operation indicators; When the task value data is less than a preset task termination threshold, the target task is terminated according to a preset process processing instruction, and a task termination result for the target task is generated according to the process processing instruction. When the task value data is greater than or equal to the task termination threshold, the node data status of the process node is analyzed in the pre-acquired data sharing center, and the node data scheduling path of the process node is determined based on the node data status. Based on the node data scheduling path, the preset intelligent agent corresponding to the process node is driven to process the target task and obtain the task processing result.

2. The task processing method based on data sharing as described in claim 1, characterized in that, Determining the flow node of the target task in the preset task flow data based on the initial input data includes: Extract the task type and task parameters from the initial input data; The task type is matched with the task flow type in the preset task flow data to determine the task flow type of the target task. Obtain the sequence of process nodes corresponding to the task process type, and map the task parameters to the process nodes corresponding to the sequence of process nodes.

3. The task processing method based on data sharing as described in claim 1, characterized in that, The step of generating the task termination result of the target task according to the process processing instructions includes: Extract the task execution metrics of the process nodes corresponding to the process processing instructions. Obtain the termination type code from the process processing instruction, and generate a task termination description based on the termination type code and the task operation indicators; The task termination description is concatenated with the task operation indicators to obtain the task termination result of the target task.

4. The task processing method based on data sharing as described in claim 1, characterized in that, The step of analyzing the node data status of the process node within the pre-acquired data sharing center includes: The node configuration information of the process node is parsed, and a query instruction for the process node is generated based on the node configuration information; Use the query command to query the node data type of the process node within the pre-acquired data sharing center; Extract the node storage data corresponding to the node data type from the data sharing center; The data status of the nodes is obtained by performing data status analysis on the stored data of the nodes.

5. The task processing method based on data sharing as described in claim 1, characterized in that, Determining the node data scheduling path of the process node based on the node data status includes: When the node data status is a first preset status, a data transmission channel is established between the data sharing center and the process node; The data transmission channel is determined as the reuse path of the process node, and the reuse path is determined as the node data scheduling path of the process node; When the node data status is the second preset status, the computing channel of the agent corresponding to the process node is determined as the execution path of the process node; The execution path is determined as the node data scheduling path of the process node.

6. The task processing method based on data sharing as described in claim 1, characterized in that, The process node is driven by a preset intelligent agent corresponding to the node based on the node data scheduling path to process the target task and obtain the task processing result, including: If the node data scheduling path is a reused path, then the node storage data corresponding to the process node is extracted from the data sharing center, and the node storage data is used as the node task result of the process node. If the node data scheduling path is an execution path, then the preset intelligent agent corresponding to the process node is driven to perform task analysis on the target task and generate the node task result of the process node. The node task results are associated with the node identifiers corresponding to the process nodes and stored in the data sharing center; Determine whether there are any unexecuted process nodes in the process node sequence corresponding to the task process type for the target task; If it exists, return to the step of obtaining the task execution metrics of the process node; If it does not exist, then extract the node task results of all process nodes from the data sharing center, and summarize the node task results to obtain the task processing result of the target task.

7. The task processing method based on data sharing as described in claim 6, characterized in that, The preset intelligent agent corresponding to the process node performs task analysis on the target task and generates the node task result of the process node, including: Obtain the node input data of the process node and analyze the context information of the process node; The node input data and the context information are encapsulated into an input data packet, and the input data packet is sent to the preset intelligent agent corresponding to the process node; The input data packet is analyzed using the analysis model in the intelligent agent to obtain the node task results of the process node.

8. A task processing device based on data sharing, characterized in that, include: The process node determination module is used to obtain the initial input data and initial task value data of the target task, and determine the process node of the target task in the preset task process data based on the initial input data. The task value data calculation module is used to obtain the task operation indicators of the process node and calculate the task value data of the target task based on the initial task value data and the task operation indicators. The task termination result generation module is used to terminate the target task according to a preset process processing instruction when the task value data is less than a preset task termination threshold, and to generate the task termination result of the target task according to the process processing instruction. The node data status analysis module is used to analyze the node data status of the process node in the pre-acquired data sharing center when the task value data is greater than or equal to the task termination threshold, and to determine the node data scheduling path of the process node based on the node data status. The task result processing module is used to drive the preset intelligent agent corresponding to the process node to process the target task based on the node data scheduling path, and obtain the task processing result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the data-sharing-based task processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the data-sharing-based task processing method as described in any one of claims 1 to 7.

Images