Task assistance method and apparatus, electronic device, and storage medium

Abstract

The application provides a task assistance method and device, electronic equipment and storage medium, belonging to the technical field of artificial intelligence, comprising: determining a task implementation process required for completing an omics task through a coarse-grained planning group; sequentially traversing each process step in the task implementation process, and performing the following operations: determining specific implementation steps required for completing each process step through a fine-grained planning group; executing the specific implementation steps through an action execution group to obtain a target omics result after completing the specific implementation steps. The task assistance method and device, electronic equipment and storage medium provided by the application provide a method for combining a multi-agent system based on a large language model and bioinformatics research for omics research, and innovatively design an agent group grouping management mode, which can reduce information redundancy and meaningless conversations between agents, ensure efficient and accurate information flow, and effectively improve the efficiency and accuracy of omics research between the two.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a task assistance method, apparatus, electronic device, and storage medium. Background Technology

[0002] In recent years, driven by the exponential growth of complex and multimodal biological data and the rapid development of artificial intelligence (AI) technology, the field of bioinformatics research is undergoing profound changes. Despite significant progress in combining bioinformatics and AI, current life science research still faces many challenges. In particular, processing large-scale, high-dimensional, and multimodal biological data requires advanced AI technologies to support scalable, adaptive, and interpretable data processing. However, existing single-agent systems have limitations in handling these complex tasks.

[0003] While single-agent systems can support task planning and tool invocation to complete user-specified tasks, their capabilities are limited by the performance of large language models when handling complex omics tasks. For example, when dealing with extremely long texts, multi-turn dialogues, or complex contexts, single agents may misunderstand user intent and provide incompletely relevant responses due to limited memory retention and understanding of contextual relevance. Furthermore, single agents typically decompose tasks into linear workflows, lacking multi-perspective and macro-level problem analysis capabilities, and exhibiting low execution efficiency, making it difficult to respond promptly to dynamic environmental changes.

[0004] To overcome the limitations of single-agent systems, multi-agent systems based on large language models have emerged. Multi-agent systems consist of multiple agents with different capabilities and roles working collaboratively. Through interactions between agents, they simulate complex real-world environments, forming collective intelligence to solve more dynamic and complex tasks. However, existing multi-agent systems also face challenges, such as high difficulty in system design and implementation, complex interaction processes leading to information redundancy, high computational resource consumption, and high customization costs. Summary of the Invention

[0005] This invention provides a task assistance method, device, electronic device, and storage medium to address the shortcomings of existing technologies, such as the lack of multi-perspective and macro-level problem analysis capabilities and low execution efficiency of single-agent systems, as well as the high design and implementation difficulty, complex interaction process, large consumption of computing resources, and high customization costs of multi-agent systems.

[0006] This invention provides a task assistance method, comprising the following steps:

[0007] Obtain omics tasks assigned by users;

[0008] The task implementation process required to complete the omics task is determined by coarse-grained planning groups;

[0009] Iterate through each step of the task implementation process sequentially and perform the following operations:

[0010] The specific implementation steps required to complete each of the process steps are determined by fine-grained planning groups; the specific implementation steps are executed by action execution groups to obtain the target omics results after the specific implementation steps are completed.

[0011] The coarse-grained planning group, the fine-grained planning group, and the action execution group are groups composed of multiple identical or different intelligent agents.

[0012] According to a task assistance method provided by the present invention, the coarse-grained planning group includes at least a coarse-grained planning agent, a first agent agent, and a coarse-grained auditing agent;

[0013] The process of determining the task implementation required to complete the omics task through coarse-grained planning groups includes:

[0014] The coarse-grained planning agent generates an initial implementation process based on the omics task.

[0015] Obtain the first review result of the coarse-grained review agent on the initial implementation process;

[0016] If it is determined that the first review result includes process improvement suggestions, then the coarse-grained planning agent is controlled to generate a new initial implementation process based on the process improvement suggestions, and the coarse-grained review agent re-reviews the new initial implementation process until the first review result is determined to be passed;

[0017] The initial implementation process when the first review result is passed is sent to the first agent intelligent agent as the task implementation process, so as to initiate the first human-computer interaction through the first agent intelligent agent;

[0018] The first human-computer interaction is used to obtain the user's adjustment suggestions for the task implementation process, and the adjustment suggestions are used by the coarse-grained planning agent to readjust the initial implementation process.

[0019] According to a task-assistance method provided by the present invention, the coarse-grained planning group is temporarily constructed based on the omics task, and different coarse-grained planning groups are constructed for different omics tasks.

[0020] According to a task assistance method provided by the present invention, the fine-grained planning group includes at least a fine-grained planning agent, a second agent agent, and a fine-grained auditing agent;

[0021] The determination of the specific implementation steps required to complete each process step through fine-grained planning groups includes:

[0022] The fine-grained planning agent generates initial implementation steps according to the process steps.

[0023] Obtain the second audit result of the fine-grained audit agent on the initial implementation step;

[0024] If it is determined that the second review result includes step improvement suggestions, then the fine-grained planning agent is controlled to generate new initial implementation steps based on the step improvement suggestions, and the fine-grained review agent is used to re-review the new initial implementation steps until the second review result is determined to be passed;

[0025] The initial implementation steps when the second review result is passed are sent to the second agent intelligent agent as the specific implementation steps, so that the second agent intelligent agent can initiate the second human-computer interaction;

[0026] The second human-computer interaction is used to obtain the user's adjustment suggestions for the specific implementation steps, and the adjustment suggestions are used by the fine-grained planning agent to readjust the initial implementation steps.

[0027] According to a task assistance method provided by the present invention, the fine-grained planning group is temporarily constructed based on the process steps, and different fine-grained planning groups are constructed for different process steps.

[0028] According to a task assistance method provided by the present invention, the action execution group includes at least a coding agent, a third agent agent, and a debugging agent;

[0029] The process of executing the specific implementation steps through an action execution group to obtain the target omics results after completing the specific implementation steps includes:

[0030] The coded agent generates initial omics code according to the specific implementation steps.

[0031] The initial omics code is debugged by the debugging agent, so that the coding agent regenerates new initial omics code based on the debugging results, until it is determined that the debugging result of the debugging agent on the initial omics code is successful;

[0032] The initial omics code obtained when the debugging result is successful is used as the target omics code;

[0033] The target omics code is executed by the third agent to generate the target omics results.

[0034] According to a task-assisting method provided by the present invention, after the coded agent generates initial omics code according to the specific implementation steps, the method further includes:

[0035] A third-party human-computer interaction is initiated through the third agent intelligent agent;

[0036] The third human-computer interaction is used to obtain the user's suggestions for adjusting the initial omics code.

[0037] According to a task-assistance method provided by the present invention, after debugging the initial omics code through the debugging agent, the method further includes:

[0038] If the debugging agent determines that the preset triggering conditions are met, then the fourth human-computer interaction is initiated through the third agent agent;

[0039] The fourth human-computer interaction is used to obtain the user's adjustment suggestions for the debugging results, and the adjustment suggestions are used by the coding agent to readjust the initial omics code.

[0040] According to a task assistance method provided by the present invention, the preset triggering conditions include:

[0041] The debugging agent generates the same code error message when running the initial omics code twice consecutively.

[0042] Alternatively, the number of code error messages generated by the debugging agent when running the initial omics code exceeds a preset threshold;

[0043] Alternatively, the debugging agent may generate a code error message containing preset sensitive words when running the initial omics code;

[0044] Alternatively, the debugging agent may lack necessary files or parameters when running the initial omics code.

[0045] A task assistance method provided by the present invention further includes:

[0046] Receive information call requests from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group;

[0047] In response to the information retrieval request, a connection is established for any intelligent agent to retrieve knowledge from an interdisciplinary knowledge base;

[0048] The interdisciplinary knowledge base is constructed by extracting knowledge from various disciplines from the environment and encoding it in a form readable by intelligent agents.

[0049] A task assistance method provided by the present invention further includes:

[0050] Receive a self-learning information call request from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group;

[0051] In response to the self-learning information call request, historical files related to any of the intelligent agents are sent to any of the intelligent agents so that any of the intelligent agents can reflect and learn based on the historical files;

[0052] The historical files include human-computer interaction information and / or task interaction logs.

[0053] According to a task-assistance method provided by the present invention, after obtaining the target omics results after completing the specific implementation steps, the method further includes:

[0054] The target omics results corresponding to each specific implementation step are stored as objects.

[0055] The present invention also provides a task assistance device, comprising the following components:

[0056] The task receiving unit is used to obtain the omics tasks issued by the user;

[0057] A coarse-grained auxiliary unit is used to determine the task implementation process required to complete the omics task through coarse-grained planning groups;

[0058] The fine-grained auxiliary unit is used to sequentially traverse each process step in the task implementation process and perform the following operations:

[0059] The specific implementation steps required to complete each of the process steps are determined by fine-grained planning groups; the specific implementation steps are executed by action execution groups to obtain the target omics results after the specific implementation steps are completed.

[0060] The coarse-grained planning group, the fine-grained planning group, and the action execution group are groups composed of multiple identical or different intelligent agents.

[0061] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the task assistance method as described above.

[0062] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the task-assisting method as described above.

[0063] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the task assistance method as described above.

[0064] The task assistance method, device, electronic device, and storage medium provided by this invention offer a method for combining a multi-agent system based on a large language model with bioinformatics research for omics research. It also innovatively designs a grouping and management method for intelligent agents, which can reduce information redundancy and meaningless dialogue between agents, ensure efficient and accurate information flow, and effectively improve the efficiency and accuracy of omics research between the two. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0066] Figure 1 This is a flowchart illustrating the task assistance method provided by the present invention.

[0067] Figure 2 This is a schematic diagram of the architecture of the task assistance system provided by the present invention.

[0068] Figure 3 This is a schematic diagram of the workflow of the task assistance system provided by the present invention.

[0069] Figure 4 This is a schematic diagram of the task assistance device provided by the present invention.

[0070] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this 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 this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0072] It should be noted that in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. The terms "upper," "lower," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Unless otherwise expressly specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection; a mechanical connection or an electrical connection; a direct connection or an indirect connection through an intermediate medium; or a connection within two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0073] The terms "first," "second," etc., used in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0074] Currently, combining the advantages of artificial intelligence technology with biological big data to drive bioinformatics research is an ongoing and arduous pursuit. Researchers possess specialized knowledge in a single bioinformatics field but lack interdisciplinary expertise and cross-disciplinary research capabilities. The effective application of artificial intelligence in bioinformatics typically requires a deep understanding of biological knowledge and advanced computing technologies; therefore, strengthening interdisciplinary collaboration is necessary to bridge the interdisciplinary gap and unlock AI-driven bioinformatics research and analysis capabilities.

[0075] Furthermore, the rapid accumulation of high-dimensional and multimodal biological data, characterized by its large scale, complexity, diversity, noise, and context dependence, has spurred the demand for advanced artificial intelligence technologies. Artificial intelligence must not only support the scalable, adaptive, and interpretable processing capabilities of biological big data, but also support the complementary integration with expert knowledge to promote improved intelligence levels.

[0076] With the breakthrough development of deep learning technology, especially the introduction of large language models, intelligent agents based on large language models possess powerful natural language processing capabilities. They can autonomously perceive the external environment, think and make decisions, and invoke tools to execute decisions, making it one of the most promising directions towards general artificial intelligence. Intelligent agents perform tasks using a perception-decision-execution model. After execution, they reflect on and evaluate the results, then learn knowledge and store it in their memory module to optimize future actions and decisions. Intelligent agents exhibit near-human intelligent behavior due to their autonomous perception of the external environment, their ability to think and make decisions, and their learning and adaptation capabilities during interactions with the environment. Human expertise can continuously improve and enhance intelligent agents, while intelligent agents, through the fusion of multidisciplinary knowledge and the mining of biological big data, generate new insights, expanding human understanding of life sciences. Utilizing their complementary advantages, bioinformatics research has gained the ability to traverse unknown knowledge domains, generate testable hypotheses, identify new research directions, and optimize experimental designs to accelerate the pace of discovery, revealing previously unexplored connections and emerging features in complex biological systems.

[0077] Commonly used single-agent frameworks include BabyAGI and Hugging, which support task planning, tool invocation, and the completion of user-specified tasks. For example, the Scientist system can autonomously conduct chemical research, including automating the planning and execution of chemical experiments, as well as predicting and optimizing reaction pathways through models. Researchers completed an instance by communicating with the Scientist system, which is based on a large language model, to perform six key tasks: planning synthetic pathways of known compounds using publicly available data, efficiently retrieving and navigating hardware documentation, executing advanced commands in a cloud-based laboratory, precisely controlling liquid handling equipment, simultaneously using multiple hardware modules to handle complex tasks and integrate multi-source data, and analyzing past experimental data to solve optimization problems.

[0078] For example, the ChemCrow system, based on a large language model, can autonomously plan and implement tasks such as organic compound synthesis, drug discovery, and materials design. By integrating tools such as web search, literature search, and an interactive programming environment, the ChemCrow system enhances its ability to acquire and understand external information, demonstrating the potential to improve scientific task processing capabilities.

[0079] In the field of pathology, PathChat, a multimodal generative AI-assisted system, demonstrates its ability to perform interactive differential diagnosis in cases of unknown primary cancer by combining histological images, clinical data, and subsequent immunohistochemical (IHC) results. It aims to assist pathologists in diagnosing and understanding complex cases, improving the accuracy and efficiency of pathological diagnosis. PathChat is designed to handle open-ended questions and multiple-choice pathology-related queries, showcasing its potential in supporting pathological diagnosis.

[0080] While large language models have driven the development of agent technology, single-agent systems still have some limitations. First, the agent's capabilities are constrained by the performance of the large language model. For example, due to poor training data quality, the agent is prone to biased, erroneous, or outdated action decisions when handling complex tasks. When dealing with extremely long texts, multi-turn dialogues, or complex contexts, the agent may misunderstand the user's intent and provide incompletely relevant responses due to limitations in the memory retention of the large language model and its understanding of cue words and contextual relevance. Second, although a single agent can play multiple roles, it quickly collapses into a specific perspective with the help of cue words or the first few rounds of dialogue, thus lacking multi-perspective and macro-level problem analysis capabilities. Finally, single agents typically decompose tasks into linear workflows, resulting in low execution efficiency and difficulty in responding promptly to dynamic environmental changes.

[0081] Single-agent systems based on large language models focus on internal thought and decision-making processes and interactions with the external environment. Inspired by human collaboration in solving complex tasks, multi-agent systems based on large language models emphasize diverse agent groups, interactions between agents, and collective decision-making processes. Compared to single-agent systems, multi-agent systems involve multiple agents with different capabilities and roles working collaboratively. Through interactions between agents, they simulate complex real-world environments, fully leveraging the strengths of different agents to form collective intelligence, thereby solving more dynamic and complex tasks such as software development, scientific experiments, and embodied intelligence.

[0082] In recent years, several noteworthy multi-agent frameworks have emerged in the open-source community, such as CAMEL and AutoGen. CAMEL uses cue words to define the roles of agents, promoting autonomous collaboration among agents and guiding agent dialogue toward task objectives. AutoGen is a highly customizable multi-agent framework that allows researchers to customize large language models, communication mechanisms, and workflow management for agents to solve complex tasks.

[0083] However, multi-agent systems also face some challenges, mainly in that the interaction process is complex when multi-agent systems face dynamic environments, which can easily lead to information redundancy and meaningless dialogue, affecting task execution efficiency.

[0084] Advances in high-throughput sequencing technology and artificial intelligence have provided unprecedented opportunities for bioinformatics research. However, addressing the challenges posed by the exponential growth of omics data and the rapid development of AI technologies requires intelligent big data analysis capabilities and interdisciplinary knowledge-driven scientific insights.

[0085] Omics is a comprehensive discipline that mainly studies the collection of various components of organisms, including genomes, transcriptomes, proteomes, metabolomes, etc. It aims to conduct systematic and comprehensive analysis and research on the various molecular components of organisms through high-throughput experimental techniques and data analysis methods.

[0086] The following is combined Figures 1-5 This invention describes the task assistance method, apparatus, electronic device, and storage medium provided by the present invention, aiming to introduce an innovative design concept for interdisciplinary collaboration, shared knowledge resources, and continuous human-computer interaction feedback, particularly in providing innovative designs for efficient human-computer interaction, agent group management, interdisciplinary knowledge empowerment, and continuous agent learning. These innovative designs enhance the task assistance method and apparatus provided by the present invention in terms of language intelligence, social intelligence, knowledge empowerment, and learning intelligence, effectively compensating for or at least alleviating some shortcomings of existing multi-agent systems.

[0087] Figure 1 This is a flowchart illustrating the task assistance method provided by the present invention, as shown below. Figure 1 As shown, the executing entity can be a processor, computer, industrial control computer, or other device capable of data processing, storage, computation, and communication with the external environment. For ease of description, it will be referred to as a coordinating processor in subsequent embodiments. The provided task assistance method includes, but is not limited to, the following steps:

[0088] First, obtain the omics tasks assigned by the users.

[0089] Users first input a specific omics task through a user interface (UI) or application programming interface (API) that communicates with the coordinator. For example, a user might want to assist in studying gene variations of a specific genetic disease; the task description input by the user might include the research objective, expected output, and relevant sample information.

[0090] Then, the task implementation process required to complete the omics task is determined by coarse-grained planning groups.

[0091] After receiving the omics task from the user, the coordinator passes it to the coarse-grained planning group. This coarse-grained planning group can be composed of multiple agents with different domain knowledge and planning capabilities. It can perform preliminary decomposition and planning of the omics task from a macro perspective, and obtain the task implementation process required to complete the omics task.

[0092] In a coarse-grained planning group, multiple agents can be selected from a cluster of agents based on the requirements for completing an omics task. Through the collaboration of these agents, the main process steps required to complete the user task can be determined. For example, for gene variation research on genetic diseases, the steps that may be performed include data collection, data preprocessing, gene variation detection, and variation result analysis.

[0093] The coarse-grained planning group feeds back the determined task implementation process to the coordinator, which can also display it synchronously on the user interface.

[0094] As an optional embodiment, after receiving the omics task from the user, the coordinating processor can first invoke a retrieval agent to perform searches in existing knowledge bases and other existing technologies. The retrieval agent is an agent capable of performing a comprehensive search of existing technologies.

[0095] If a search reveals an existing technology (such as an existing knowledge base) that can fully solve the user's omics task, the search results are directly used as the target omics result and fed back to the user through a coordinator, thus ending the task assistance.

[0096] If the retrieval agent fails to find an auxiliary method in the existing technology that can completely solve the omics task input by the user, then the implementation process of the omics task is analyzed by constructing a coarse-grained planning group.

[0097] After obtaining the task implementation flow determined by the coarse-grained planning group, the coordinator calls the constructed fine-grained planning group and action execution group to sequentially traverse each step of the task implementation flow, mainly performing the following operations:

[0098] The specific implementation steps required to complete each of the aforementioned process steps are determined through a fine-grained planning group. Specifically, for each process step in the task implementation process, it is first passed to the fine-grained planning group. The fine-grained planning group consists of multiple intelligent agents with detailed domain knowledge and execution capabilities, capable of further breaking down and refining the process steps from a microscopic perspective.

[0099] Fine-grained planning groups can determine the specific implementation steps required to complete a given process step based on its content and requirements. For example, for a data preprocessing step, the specific implementation steps that may be performed include data cleaning, data formatting, and data standardization.

[0100] The fine-grained planning group returns the specific implementation steps determined for the current process step to the coordinator, which then passes them on to the action execution group. The action execution group consists of multiple agents with actual operational and execution capabilities, capable of performing relevant operations according to the requirements of the specific implementation steps.

[0101] Each agent in the action execution group performs relevant operations and obtains results based on the content and requirements of the specific implementation steps. For example, the data cleaning step may involve operations such as removing invalid data, filling in missing values, and handling outlier data.

[0102] After completing the specific implementation steps related to the current process steps, the action execution group will return the obtained target omics results to the coordinator.

[0103] After determining that the current process step has been completed, the coordinator continues to send the next process step to the fine-grained planning group for step division.

[0104] The above steps are executed iteratively until all process steps in the entire task implementation process are executed, and the target omics results corresponding to each process step are obtained.

[0105] The task assistance method provided by this invention offers a way to combine multi-agent systems based on large language models with bioinformatics research for omics research. It also innovatively designs a grouping and management method for agent groups, which can reduce information redundancy and meaningless dialogue between agents, ensure efficient and accurate information flow, and effectively improve the efficiency and accuracy of omics research between the two.

[0106] As an optional embodiment, after the coordinator obtains the target omics results after completing the specific implementation steps related to each process step, the method further includes:

[0107] The target omics results corresponding to each specific implementation step are stored as objects.

[0108] Taking a specific bioinformatics analysis workflow as an example, the coordinator is configured to manage and coordinate the execution of multiple workflow steps related to omics tasks, which are designed to analyze specific genomic data.

[0109] First, the coordinator triggers and executes each step of the task implementation process sequentially, according to the sequence determined by the coarse-grained planning group. For example, it executes four steps sequentially: quality control, sequence alignment, variant detection, and gene function annotation. Specifically, the quality control step cleans and filters out low-quality DNA sequence data through multiple implementation steps; the sequence alignment step compares the filtered DNA sequence data with a reference genome through multiple implementation steps; the variant detection step identifies variant information in the sequence through multiple implementation steps, such as detecting single nucleotide polymorphisms (SNPs); and the gene function annotation step provides biological interpretations of the detected variants through multiple implementation steps.

[0110] Furthermore, the coordinator will follow the method provided in the above embodiments to sequentially traverse the above four process steps and obtain the target omics results corresponding to each process step.

[0111] Specifically, after each process step is completed, the coordinator retrieves the target omics results output by the action execution group. For example, after the quality control step, it retrieves a cleaned, high-quality DNA sequence dataset; after the sequence alignment step, it retrieves the alignment result file, including the location information of each DNA sequence on the reference genome; after the variant detection step, it retrieves a list containing information on all detected variants; and after the gene function annotation step, it retrieves annotation information such as the gene name and functional impact corresponding to each variant.

[0112] After obtaining the target omics results corresponding to each specific implementation step, the coordinator performs an additional step: object storage of these target omics results. This invention stores the result data as objects (such as files or data blocks) in a dedicated object storage system because object storage systems offer high scalability, data persistence, and fast access, making them ideal for storing and processing large-scale bioinformatics data.

[0113] The coordinator can create a unique object identifier (such as an object key or filename) for the target omics results of each process step by calling the API interface of the object storage system, and upload the data related to the target omics results as the content of the object to the specified bucket (or container). Simultaneously, it can add metadata to these objects, such as description information, creation time, step name, etc., to facilitate subsequent data retrieval and management.

[0114] Once the results of all process steps are securely stored in the object storage system, the coordinator can perform further operations, such as generating a final report, triggering downstream analysis processes, or notifying the user that the results are ready.

[0115] The task assistance method provided by this invention not only effectively manages and coordinates the analysis and execution of process steps in omics tasks, but also ensures that the omics results generated by each process step can be stored and managed securely and efficiently, providing a solid foundation for subsequent data analysis and mining.

[0116] To more clearly illustrate the task assistance method provided by this invention, the following explanation will focus on the system architecture in which the task assistance method operates.

[0117] Figure 2 This is a schematic diagram of the architecture of the task assistance system provided by the present invention, such as... Figure 2 As shown, the task assistance system upon which the task assistance method provided by this invention relies can be mainly considered to consist of three parts: an intelligent agent cluster, a coordination layer, and an environment.

[0118] The intelligent agent cluster comprises multiple intelligent agents, which are the core components of the multi-agent system. In subsequent embodiments of this invention, the intelligent agents employ a large language model as their control core, possess autonomy, and are capable of perceiving information in the environment, understanding complex natural language instructions (such as expert prompts), performing thought and decision-making, and invoking tools to execute actions. Different intelligent agents can assume different roles and responsibilities, such as information gathering, task planning, code generation, and task execution.

[0119] The coordination layer provides a bridge for agents to perceive and influence the environment. The virtual coordination layer processor can be considered the executor of the task assistance method provided by this invention, capable of coordinating the behavior of various agents simultaneously. Building upon existing tasks such as task planning and allocation, resource scheduling, conflict resolution, and collaborative strategies performed by agents, this invention further develops functions in the coordination layer, including agent group management strategies, more efficient human-agent interaction mechanisms, interdisciplinary shared knowledge bases, and continuous learning strategies for agents, effectively enhancing task assistance capabilities.

[0120] The environment refers to the external conditions under which a multi-agent system operates. The state of the environment affects the behavior of the agents, and the actions of the agents also change the state of the environment. Specifically, the environment can include elements of the real world, such as users, instruments, equipment, and tools, as well as elements of the virtual world, such as datasets, documents, knowledge bases, and software.

[0121] The coarse-grained planning group is composed of multiple agents selected from the agent cluster that can collaborate to achieve coarse-grained planning for omics tasks. These agents mainly include, but are not limited to, coarse-grained planning agents, first agent agents, and coarse-grained audit agents.

[0122] The process of determining the task implementation required to complete the omics task through coarse-grained planning groups includes:

[0123] The coarse-grained planning agent generates an initial implementation process based on the omics task.

[0124] Obtain the first review result of the coarse-grained review agent on the initial implementation process;

[0125] If it is determined that the first review result includes process improvement suggestions, then the coarse-grained planning agent is controlled to generate a new initial implementation process based on the process improvement suggestions, and the coarse-grained review agent re-reviews the new initial implementation process until the first review result is determined to be passed;

[0126] The initial implementation process when the first review result is passed is sent to the first agent intelligent agent as the task implementation process, so as to initiate the first human-computer interaction through the first agent intelligent agent;

[0127] The first human-computer interaction is used to obtain the user's adjustment suggestions for the task implementation process, and the adjustment suggestions are used by the coarse-grained planning agent to readjust the initial implementation process.

[0128] Specifically, the coarse-grained planning agent is responsible for generating an initial implementation process (called the initial implementation process) based on the input omics task.

[0129] After the coarse-grained planning agent completes the initial planning, another agent will be responsible for initiating human-computer interaction with the user. The purpose of this interaction is to receive the user's suggestions for adjusting the initial implementation process and to feed these suggestions back to the coarse-grained planning agent for subsequent process optimization.

[0130] The coarse-grained review agent is primarily used to review the initial implementation process generated by the coarse-grained planning agent. It mainly checks the rationality and feasibility of the initial implementation process, as well as whether there are potential errors or areas for improvement. If the review result output by the coarse-grained review agent (referred to as the first review result) contains process improvement suggestions, the coordinator will forward it to the coarse-grained planning agent. The coarse-grained planning agent can then generate a new initial implementation process based on these suggestions, which will be reviewed again by the coarse-grained review agent until the review result is approved.

[0131] Figure 3 This is a schematic diagram of the workflow of the task assistance system provided by the present invention. The following is in conjunction with... Figure 3 As shown, the task assistance method provided by the present invention utilizes coarse-grained planning groups to determine the task implementation process.

[0132] Assuming the task description input by the user mainly includes "perform hierarchical clustering. The total number of cluster levels is 4. For the final human lung cell atlas, it is also necessary to calculate the marker gene for each cluster," then it can be determined that the user's input omics task is to use machine learning algorithms to perform hierarchical clustering on single-cell RNA sequencing data to identify different human lung cell types. The goal is to assist in completing this process and provide the marker gene for each cluster.

[0133] The coarse-grained planning intelligence will perform preliminary analysis of the omics tasks given by the user and generate a preliminary implementation process. This preliminary implementation process describes the main steps required to complete the omics task and the logical relationship between these main steps, which mainly includes data preprocessing, hierarchical clustering, and calculation of marker genes.

[0134] After receiving the initial implementation process, the coarse-grained auditing agent conducts a detailed audit, mainly checking whether each process step is clear and reasonable, and whether there are any potential errors or omissions.

[0135] Suppose that during the review process, the coarse-grained review agent discovers two problems with hierarchical clustering in the initial implementation process: "1. Step combination and decomposition; 2. Method selection". In other words, the coarse-grained review agent puts forward some process improvement suggestions and requires the coarse-grained planning agent to supplement or improve the problems pointed out.

[0136] Based on the process improvement suggestions from the coarse-grained review agent, the coarse-grained planning agent adjusted the initial implementation process, supplemented the specific algorithms in the hierarchical clustering step (such as using the Leiden algorithm for hierarchical clustering) and improved the step combination and decomposition (such as initializing a loop in the third step to execute clustering at each level), and resubmitted it to the coarse-grained review agent for review.

[0137] The review process of the coarse-grained review agent and the improvement process of the coarse-grained planning agent may be iterated multiple times. Each iteration involves the coarse-grained planning agent adjusting the current preliminary implementation process based on the review suggestions, and the coarse-grained review agent re-reviewing the adjusted preliminary implementation process.

[0138] The iteration process terminates when the coarse-grained review agent confirms that the final generated preliminary implementation process is correct and issues a pass review result. At this point, the initial implementation process is determined as the final task implementation process, which includes clearly defined implementation procedures such as data preprocessing, hierarchical clustering, and calculation of marker genes.

[0139] Once the coordinator learns that the task implementation process has been determined, it will call the first agent to receive the task implementation process and prepare to initiate human-computer interaction with the user.

[0140] During the human-computer interaction phase, the first agent AI will display the task implementation process to the user and invite the user to provide suggestions for adjustments. Users can modify or improve certain implementation processes or steps based on their professional knowledge and experience, such as the method for computationally marker genes.

[0141] After receiving user feedback and adjustments, the optimized task implementation process from the coarse-grained planning agent will be submitted again to the coarse-grained review agent for approval. This is to ensure that the final task implementation process meets both the agent's review requirements and the user's needs.

[0142] When the coordinator receives a final approval result from the granularity review agent, the task implementation process is formally determined and can begin execution. During execution, the coordinator will trigger and execute each step in the task implementation process sequentially according to the steps and logical relationships, including data preprocessing, hierarchical clustering, and calculation of marker genes.

[0143] During execution, fine-grained planning and adjustments may be necessary to ensure precise execution of each step. For example, in the hierarchical clustering stage, the clustering algorithm and parameters may be adjusted based on the actual data; in the marker gene calculation stage, appropriate statistical methods and thresholds may be selected to determine the marker genes.

[0144] The task assistance method provided by this invention addresses the complex process of determining the task implementation process related to omics tasks, which involves collaboration, iteration, and human-computer interaction among multiple intelligent agents. This process ensures that the final determined task implementation process not only meets the objective requirements of solving omics tasks but also satisfies the subjective needs of users, thereby improving the efficiency and accuracy of omics research.

[0145] Based on the above embodiments, as an optional embodiment, the coarse-grained planning group is temporarily constructed based on the omics task, and different coarse-grained planning groups are constructed for different omics tasks.

[0146] Specifically, the task-assistance method provided by this invention optimizes the task execution process and improves analysis efficiency and accuracy by constructing coarse-grained planning groups. A coarse-grained planning group is a cluster of multiple identical or different intelligent agents (such as artificial intelligence algorithms, machine learning models, expert systems, etc.) that collaborate to complete complex omics tasks.

[0147] During each execution, the omics task can be decomposed to identify each subtask and its requirements. Based on the requirements of the subtask, the required agent type and number can be determined, and one or more coarse-grained planning groups can be temporarily constructed. Each coarse-grained planning group consists of multiple identical or different agents. These agents are selected and combined according to the requirements of the subtask. The agents in the group cooperate and communicate to complete the omics task together.

[0148] Within the coarse-grained planning groups, specific tasks are assigned based on the expertise and experience of each agent. Each agent performs corresponding operations according to its assigned task, such as data preprocessing, algorithm application, and result analysis. During execution, agents maintain necessary communication and collaboration to ensure task continuity and efficiency. Each coarse-grained planning group then integrates its results to form the final task implementation flow.

[0149] Optionally, after completing the coarse-grained planning work for the omics task, the coarse-grained planning group can be disbanded, that is, the relevant resources of each agent (such as computing resources, storage devices, etc.) are reclaimed to serve other omics tasks.

[0150] For example, when the coordinator receives a coarse-grained planning group with an upcoming omics task involving deep analysis of a batch of genomics data, including gene variant detection, gene expression analysis, and gene function annotation, the coordinator will temporarily construct a coarse-grained planning group. This group will include multiple agents such as variant detection algorithm agents, data analysis agents, and computational resource agents. All these agents will collaborate to complete steps such as data preprocessing, variant detection, expression analysis, and functional annotation, ultimately outputting a feasible task implementation flow.

[0151] As another alternative embodiment, a coarse-grained planning group containing multiple agents can be pre-built and repeatedly invoked each time an omics task is performed.

[0152] The task assistance method provided by this invention can flexibly allocate various intelligent agent resources by constructing coarse-grained planning groups, optimize the task execution process, and improve the efficiency and accuracy of omics data analysis. This method is not only applicable to the fields of genomics and proteomics, but can also be extended to other omics fields and complex task scenarios.

[0153] refer to Figure 3 As shown, in one optional embodiment, the fine-grained planning group includes at least a fine-grained planning agent, a second agent agent, and a fine-grained auditing agent;

[0154] The determination of the specific implementation steps required to complete each process step through fine-grained planning groups includes:

[0155] The fine-grained planning agent generates initial implementation steps according to the process steps.

[0156] Obtain the second audit result of the fine-grained audit agent on the initial implementation step;

[0157] If it is determined that the second review result includes step improvement suggestions, then the fine-grained planning agent is controlled to generate new initial implementation steps based on the step improvement suggestions, and the fine-grained review agent is used to re-review the new initial implementation steps until the second review result is determined to be passed;

[0158] The initial implementation steps when the second review result is passed are sent to the second agent intelligent agent as the specific implementation steps, so that the second agent intelligent agent can initiate the second human-computer interaction;

[0159] The second human-computer interaction is used to obtain the user's adjustment suggestions for the specific implementation steps, and the adjustment suggestions are used by the fine-grained planning agent to readjust the initial implementation steps.

[0160] The following section will elaborate on how, based on the task implementation process already determined using coarse-grained planning groups, fine-grained planning groups are used to further refine the specific implementation steps of each process step in the task implementation process.

[0161] Fine-grained planning groups include at least three types of agents, such as fine-grained planning agents, agent agents (for ease of description, they are referred to as second agent agents), and fine-grained auditing agents.

[0162] Among them, the fine-grained planning agent is mainly responsible for generating initial specific implementation steps (hereinafter referred to as initial implementation steps) based on each process step determined by the coarse-grained planning group. These initial implementation steps need to be detailed enough to directly guide the execution of subsequent tasks.

[0163] The fine-grained auditing agent is primarily used to audit the initial implementation steps generated by the fine-grained planning agent, ensuring that they comply with the relevant requirements, safety standards, and other related specifications of the omics task. If problems or areas for improvement are found, suggestions for improving the steps will be provided.

[0164] The second agent mainly acts as a bridge for interaction with the user, responsible for receiving the specific implementation steps of the fine-grained planning group output, and initiating human-computer interaction with the user by calling the human-computer interaction interface to obtain user feedback and adjustment suggestions.

[0165] The following section will explain in detail how the three types of intelligent agents collaborate to formulate specific implementation steps for each process step.

[0166] Based on the task implementation process determined by the coarse-grained planning group, the fine-grained planning agent generates preliminary detailed implementation steps (called initial implementation steps) according to the detailed description and context information of each process step. These initial implementation steps are required to include information such as specific operation steps, required resources, and expected results.

[0167] The fine-grained review agent receives the initial implementation steps generated by the fine-grained planning agent and conducts a detailed review. The review includes, but is not limited to, the feasibility, security, efficiency, and whether the steps align with the overall task objectives.

[0168] If the fine-grained auditing agent identifies any issues or areas for improvement, it generates suggestions for improvement and returns them to the fine-grained planning agent via the coordinator. These suggestions may include reordering steps, reallocating resources, or adjusting operational details.

[0169] In the above scenario, the fine-grained planning agent will revise the initial implementation steps it generates based on the improvement suggestions provided by the fine-grained review agent, thus generating new initial implementation steps. These new initial implementation steps will then be submitted to the fine-grained review agent for review.

[0170] The process of generating the initial implementation steps and reviewing the initial implementation steps will be repeated until the fine-grained review agent confirms that the fine-grained planning agent has fully met the task requirements in the currently generated initial implementation steps, that is, the second review result is passed.

[0171] Once the coordinator learns that the second review result is passed, it will send the initial implementation steps generated by the fine-grained planning agent as the final specific implementation steps to the second agent agent.

[0172] The coordinator will notify the second agent to initiate human-computer interaction with the user (this is called the second human-computer interaction), show the user the specific implementation steps, and invite the user to make adjustment suggestions.

[0173] At this point, users can fine-tune the implementation steps or offer other suggestions based on their experience and needs. The second agent will then relay the user's suggestions to the fine-grained planning agent via the coordinator. This allows the fine-grained planning agent to further adjust and optimize the specific implementation steps based on the user's feedback, ensuring that the steps meet both the objective requirements of the omics task and the user's actual needs.

[0174] The task assistance method provided by this invention achieves further refinement and optimization of the task implementation process through the collaborative work of various agents in a fine-grained planning group. The fine-grained planning agent is responsible for generating the initial implementation steps, the fine-grained review agent is responsible for reviewing and providing improvement suggestions, and the second agent agent is responsible for interacting with the user, obtaining user feedback and adjustment suggestions. This collaborative working method effectively improves the efficiency and accuracy of omics task execution, and in particular, can meet the actual needs of users.

[0175] In omics research, such as genomics, transcriptomics, or proteomics, data analysis tasks typically involve multiple complex and interdependent implementation steps. Each step may require different expertise, computational resources, and review criteria. To improve the flexibility and efficiency of understanding and decomposing omics tasks, this invention provides a method for dynamically constructing fine-grained planning groups based on process steps.

[0176] First, the entire omics task is broken down into a series of ordered process steps using coarse-grained planning groups. In genomic data analysis, these steps might include data reception and preprocessing, sequence alignment, variant detection, functional annotation, and results reporting. Each step has clearly defined inputs, outputs, and required expertise.

[0177] Then, based on the execution requirements of each process step, the required intelligent agents are dynamically selected to construct a coarse-grained planning group.

[0178] For example, for the data receiving and preprocessing process, a fine-grained planning group can be constructed, comprising a data cleaning agent, a format conversion agent, and a preprocessing review agent. The data cleaning agent is responsible for removing low-quality data, the format conversion agent ensures the data format is compatible with subsequent analysis software, and the preprocessing review agent performs quality checks on the preprocessing results.

[0179] For example, regarding the sequence alignment process, a fine-grained planning group is reconstructed, comprising an alignment algorithm selection agent, an alignment execution agent, and an alignment result review agent. The alignment algorithm selection agent chooses a suitable alignment algorithm based on data characteristics and alignment requirements; the alignment execution agent performs the alignment operation; and the alignment result review agent verifies the accuracy and completeness of the alignment results. Through collaboration and information transmission mechanisms among the groups, continuity in the execution of omics tasks and data consistency are achieved.

[0180] Optionally, during the execution of an omics task, if different fine-grained planning groups are constructed for each process step, all fine-grained planning groups can collaborate and exchange information through a coordinator, including providing a unified communication protocol and data interface.

[0181] The task assistance method provided by this invention achieves flexible and efficient execution of omics data analysis tasks by dynamically constructing fine-grained planning groups based on process steps. Each process step constructs a corresponding fine-grained planning group according to specific needs, ensuring the targeted and professional nature of task execution.

[0182] As an optional embodiment, the action execution group includes at least a coding agent, a third-party agent, and a debugging agent;

[0183] The process of executing the specific implementation steps through an action execution group to obtain the target omics results after completing the specific implementation steps includes:

[0184] The coded agent generates initial omics code according to the specific implementation steps.

[0185] The initial omics code is debugged by the debugging agent, so that the coding agent regenerates new initial omics code based on the debugging results, until it is determined that the debugging result of the debugging agent on the initial omics code is successful;

[0186] The initial omics code obtained when the debugging result is successful is used as the target omics code;

[0187] The target omics code is executed by the third agent to generate the target omics results.

[0188] The task-assistance method provided by this invention offers a way to execute specific implementation steps through action execution groups and obtain target omics results after these steps are completed. The action execution group includes at least three types of agents: encoding agents, proxy agents (referred to as third proxy agents for ease of description), and debugging agents. These agents work collaboratively to execute the specific implementation steps related to each process step and ultimately generate target omics results.

[0189] The following is based on Figure 3 Taking the content shown as an example, this article details how the dynamic execution group performs specific implementation steps to obtain the target omics results after completing the specific implementation steps.

[0190] The coded agent generates initial omics code (referred to as initial omics code) according to the specific implementation steps, specifically as follows:

[0191] The coding agent first imports the scampy library and sets the dataset variable. Then, it uses the command `sc.read_h5ad('integrated_dataset_copy.h5ad')` to read a dataset named `integrated_dataset_copy.h5ad`. If the dataset contains the key 'SCANVI', the coding agent will generate the corresponding code to print the value associated with that key.

[0192] The debugging agent debugs the initial omics code to ensure its correctness and reliability. Specifically, if the detection fails, a debugging result KeyError is generated, indicating "No 'neighbors' in uns". This KeyError indicates that the neighborhood graph has not been computed, which is a prerequisite for the Leiden clustering algorithm. After capturing this error, the debugging agent analyzes it and determines the cause. Then, the debugging result is fed back to the coding agent through the coordinator, allowing the coding agent to regenerate new initial omics code based on the debugging results.

[0193] The process of generating and adjusting the initial omics code may be repeated multiple times until the debugging agent determines that the initial omics code has passed the debugging process, meaning that the code has no errors and can be executed correctly.

[0194] When the debugging agent passes the debugging test for the initial omics code, the initial omics code is identified as the target omics code. This means that the initial omics code has been fully validated and tested and can be safely used to generate target omics results.

[0195] By executing the target omics code, the third-party agent generates the target omics results. These results can include data analysis reports, gene expression maps, protein-protein interaction networks, and more.

[0196] It should be noted that during the generation of the target omics code by the action execution group, human-computer interaction will be dynamically executed. That is, throughout the process, the action execution group can not only communicate and collaborate with the internal intelligent agents, but also interact with external users or other systems to obtain more information and guidance.

[0197] In addition, the task assistance method provided by the present invention performs an orderly analysis of all process steps in the task implementation process through fine-grained planning groups and action execution groups. That is, after completing the analysis of the current process step and obtaining its corresponding target omics results, the analysis of the next process step is repeated until the analysis of all process steps is completed and the target omics results corresponding to each process step are obtained.

[0198] The task assistance method provided by this invention realizes a complete process from specific implementation steps to target omics results through the collaborative work of an encoding agent, a debugging agent, and a third agent, thereby improving the efficiency and accuracy of omics analysis or data processing tasks.

[0199] In data- and intelligence-intensive research paradigms, both human experts and intelligent agents are key participants in scientific research. To support close human-machine collaborative work, the task assistance method provided in this invention designs a user-friendly and efficient human-machine interaction mechanism.

[0200] Compared with the static human-computer interaction mechanism where the opening time of the human-computer interaction interface is fixed, the present invention provides a dynamic human-computer interaction mechanism, that is, when certain triggering conditions are met, the time window for human-computer interaction is automatically opened, providing highly real-time interaction.

[0201] This dynamic human-computer interaction mechanism is widely used in agents responsible for action execution, helping to handle complex and dynamic execution processes. By combining static human-computer interaction in all agent groups with dynamic human-computer interaction in action execution groups, unnecessary human intervention and involvement are minimized while ensuring that the multi-agent system can autonomously execute tasks.

[0202] The following section details how static and dynamic human-computer interaction are implemented within the action execution group. Specifically, the third agent is equipped with a Python kernel and code execution environment, responsible for executing the target omics code and generating the target omics results. Simultaneously, the third agent provides the action execution group with two types of human-computer interaction: static and dynamic.

[0203] As an optional embodiment, the static human-computer interaction performed by the action execution group occurs after the coded agent generates the initial omics code according to the specific implementation steps, and specifically includes:

[0204] A third-party human-computer interaction is initiated through the third agent intelligent agent;

[0205] The third human-computer interaction is used to obtain the user's suggestions for adjusting the initial omics code.

[0206] Specifically, after the initial omics code is generated using the coding agent, a third agent initiates a third-party human-computer interaction, allowing the user to view and evaluate the initial omics code and provide adjustment suggestions. For example, the user can interact with the third agent through a graphical user interface (GUI), a command-line interface (CLI), or other interactive methods. During the interaction, the user can view the content, structure, and execution results of the initial omics code and propose modifications based on their needs or expectations.

[0207] The debugging agent receives adjustment suggestions from users through third-party human-computer interaction and will perform corresponding debugging and modifications on the initial omics code, such as correcting errors in the code, optimizing the code structure, and adjusting parameter settings.

[0208] Of course, the above process may involve multiple iterations of the debugging agent based on user suggestions until an initial omics code that meets user requirements and passes the debugging test is determined.

[0209] When the debugging agent determines that the debugging result of the initial omics code is successful, the initial omics code is identified as the target omics code. Then, the third agent continues to execute the target omics code and generate the target omics results.

[0210] As another optional embodiment, the dynamic human-computer interaction performed by the action execution group may occur after the debugging agent debugs the initial omics code, specifically including:

[0211] If the debugging agent determines that the preset triggering conditions are met, then the fourth human-computer interaction is initiated through the third agent agent;

[0212] The fourth human-computer interaction is used to obtain the user's adjustment suggestions for the debugging results, and the adjustment suggestions are used by the coding agent to readjust the initial omics code.

[0213] Specifically, the debugging agent first receives the initial omics code generated by the encoding agent and debugs it. The debugging process may include checking the code's syntax, logic, execution efficiency, etc., to ensure that the code can run correctly and achieve the expected results.

[0214] During the debugging process, the debugging agent records the debugging results, including key information such as whether the code ran successfully, execution time, and resource consumption. This key information will serve as the basis for subsequent decisions on whether to initiate a fourth human-computer interaction.

[0215] After completing debugging, the debugging agent will determine whether to initiate a fourth human-computer interaction based on preset trigger conditions. If the debugging agent determines that the preset trigger conditions are met, it will proceed to the next step; otherwise, it will continue to execute other tasks or wait for new instructions.

[0216] If preset triggering conditions are met, the debugging agent will initiate a fourth human-computer interaction through a third agent. The purpose is to obtain user suggestions for adjusting the debugging results in order to further optimize the initial omics code. The fourth human-computer interaction can adopt an interaction method similar to the third human-computer interaction, such as GUI, CLI, or voice interaction.

[0217] During the fourth human-computer interaction process, users will analyze and evaluate the debugging results based on their professional knowledge and experience, and will also provide adjustment suggestions for the debugging results. These adjustment suggestions may involve code modification, parameter adjustment, algorithm optimization, etc.

[0218] After receiving adjustment suggestions from the user, the coding agent will modify and optimize the initial omics code accordingly to ensure its correctness, efficiency, and scalability, thereby meeting the user's needs and expectations. The adjusted code will then be verified and tested again by the debugging agent to ensure that its quality and performance meet the preset standards and requirements.

[0219] As an optional embodiment, the above-mentioned preset triggering conditions may include, but are not limited to:

[0220] The debugging agent generates the same code error message when running the initial omics code twice consecutively.

[0221] Alternatively, the number of code error messages generated by the debugging agent when running the initial omics code exceeds a preset threshold;

[0222] Alternatively, the debugging agent may generate a code error message containing preset sensitive words when running the initial omics code;

[0223] Alternatively, the debugging agent may lack necessary files or parameters when running the initial omics code.

[0224] When the debugging agent runs two consecutive initial omics code, if the generated code error information is exactly the same, it may mean that there is a problem in the code that the debugging agent cannot solve. At this time, it is considered that the trigger condition is met, and a fourth human-computer interaction is initiated through the third agent to request the user to provide further adjustment suggestions.

[0225] If the number of code error messages generated by the debugging agent when running the initial omics code exceeds a preset threshold (e.g., more than 5 errors), it indicates that the code may have multiple problems or be unstable. At this time, it can also be considered that the triggering condition is met, and the fourth human-computer interaction process is initiated, requesting the user to make adjustment suggestions for solving this phenomenon based on professional knowledge and experience.

[0226] In some cases, specific error messages may indicate serious code problems or potential security risks. Therefore, if the error message generated by the debugging agent when running the initial omics code contains preset sensitive words (such as the actual gene names contained in the dataset), a fourth human-computer interaction can be immediately triggered so that the user can quickly intervene and resolve the problem (if necessary, requesting the user to manually specify the name in the error message).

[0227] If the debugging agent finds that necessary files or parameters are missing when running the initial omics code, this usually means that the code cannot continue execution or the execution results may be inaccurate. In this case, the debugging agent may also consider that the triggering condition has been met and initiate a fourth human-computer interaction through a third agent, prompting the user to check and provide the missing files or parameters.

[0228] Of course, the triggering conditions need to be specifically set according to the actual situation. For example, triggering conditions may also include: debugging results not meeting preset quality standards, code execution time exceeding a threshold, excessive resource consumption, etc. By setting reasonable preset triggering conditions and combining them with a dynamic human-computer interaction mechanism, efficient debugging and optimization of initial omics code is achieved. This method not only improves the automation of code generation but also provides users with a more convenient and efficient omics analysis experience.

[0229] The task assistance method provided by this invention enhances the cognitive abilities of intelligent agents with different roles by constructing a shared interdisciplinary knowledge base.

[0230] As an optional embodiment, the task assistance method provided by the present invention further includes:

[0231] Receive information call requests from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group;

[0232] In response to the information retrieval request, a connection is established for any intelligent agent to retrieve knowledge from an interdisciplinary knowledge base;

[0233] The interdisciplinary knowledge base is constructed by extracting knowledge from various disciplines from the environment and encoding it in a form readable by intelligent agents.

[0234] During task execution, any agent in the coarse-grained planning group, fine-grained planning group, or action execution group may need to retrieve relevant knowledge or information from an interdisciplinary knowledge base based on the requirements of the current task. At this time, the agent will send an information retrieval request to the system.

[0235] Upon receiving an information retrieval request from any agent, the coordinating processor responds immediately. This response process includes identifying the source of the request, parsing the request content, and determining the specific knowledge or information in the interdisciplinary knowledge base that needs to be retrieved. Then, based on the parsed request content, it establishes a connection for the requesting agent to retrieve knowledge from the interdisciplinary knowledge base. This connection ensures that the agent can efficiently access and obtain the required knowledge or information.

[0236] For example, during coarse-grained planning, the granular planning agent initiates its process by retrieving relevant information from an interdisciplinary knowledge base. The retrieved data is integrated into the operational framework of the granular planning agent as supplementary hints, thereby increasing the agent's insight based on knowledge and enriching the decision-making process with the support of existing knowledge, thus improving its planning quality.

[0237] The interdisciplinary knowledge base is constructed by extracting knowledge from various disciplines in the environment, such as research papers from multiple disciplines, and encoding them in a form readable by intelligent agents. This means that the knowledge or information in the interdisciplinary knowledge base is stored in a format that intelligent agents can understand and apply, thereby supporting the intelligent agent's knowledge retrieval and decision-making during task execution.

[0238] The task assistance method provided by this invention takes into account that bioinformatics involves multiple disciplines. Therefore, it constructs a shared interdisciplinary knowledge base, encodes professional knowledge from different fields into a form that the intelligent agent can understand and stores it. The interdisciplinary knowledge empowerment enables the intelligent agent to extract relevant information from the knowledge base according to task requirements and provide reasonable decision-making plans, which significantly improves the efficiency and decision-making ability of the intelligent agent in the task execution process. At the same time, it promotes the sharing and integration of interdisciplinary knowledge and enhances the flexibility and scalability of the system.

[0239] In order to quickly adapt to new and challenging life science research tasks, the task assistance method provided by this invention designs a continuous learning strategy for intelligent agents, namely reflective learning (also known as self-learning).

[0240] As an optional embodiment, the task assistance method may further include:

[0241] Receive a self-learning information call request from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group;

[0242] In response to the self-learning information call request, historical files related to any of the intelligent agents are sent to any of the intelligent agents so that any of the intelligent agents can reflect and learn based on the historical files;

[0243] The historical files include human-computer interaction information and / or task interaction logs.

[0244] During task execution or planning, any agent in the coarse-grained planning group, fine-grained planning group, or action execution group may need to review and learn from past experiences or historical information based on current task requirements or its own development needs. At this time, the agent will issue a request to retrieve self-learning information.

[0245] Upon receiving a self-learning information request from any agent, the coordinator will respond immediately. This response process includes identifying the source of the request, parsing the request content, and determining the type and content of the historical files to be sent.

[0246] Based on the parsed request content, historical files related to the requesting agent will be sent to that agent. These historical files include, but are not limited to, human-computer interaction information and / or task interaction logs. Human-computer interaction information records the interaction process between the agent and the user, including user instructions, feedback, and the agent's responses; task interaction logs record various state changes, decision-making processes, and output results of the agent during task execution.

[0247] Upon receiving historical files, the agent can reflect and learn based on these files. By reviewing and analyzing past interaction information and task logs, the agent can identify its shortcomings and errors in task execution or planning, and then adjust and optimize its strategies and behaviors.

[0248] For example, a coarse-grained review agent is obtained in the coarse-grained planning group, a fine-grained review agent is obtained in the fine-grained planning group, and a debugging agent is obtained in the action execution group. They are responsible for checking the plans, steps and code generated by the coarse-grained planning agent or the fine-grained planning agent and the coding agent. By using the reflective learning provided by this invention, after multiple rounds of interaction, they can effectively optimize their own capabilities and automatically provide modification suggestions to improve the quality of the answers.

[0249] The task assistance method provided by this invention enables intelligent agents to reflect on and learn from their own behavior, thereby continuously optimizing their decision-making and execution capabilities over time, which helps to improve intensive intelligence.

[0250] Finally, it should be emphasized that the large language models of each agent in the task assistance method provided by this invention can be different types and versions of large language models, such as the LLaMa series, and the agents can be implemented using various frameworks, such as LangChain.

[0251] In addition, the task-aiding method provided by this invention can be applied to the construction of human lung tissue cell atlases, and can also be extended to other bioinformatics analysis tasks, such as multi-species, multi-tissue type, and multi-omics data analysis.

[0252] Figure 4 This is a schematic diagram of the task assistance device provided by the present invention, as shown below. Figure 4 As shown, it mainly includes:

[0253] Task receiving unit 41 is used to obtain omics tasks issued by the user;

[0254] The coarse-grained auxiliary unit 42 is used to determine the task implementation process required to complete the omics task through coarse-grained planning groups;

[0255] Fine-grained auxiliary unit 43 is used to sequentially traverse each process step in the task implementation process and perform the following operations:

[0256] The specific implementation steps required to complete each of the process steps are determined by fine-grained planning groups; the specific implementation steps are executed by action execution groups to obtain the target omics results after the specific implementation steps are completed.

[0257] The coarse-grained planning group, the fine-grained planning group, and the action execution group are groups composed of multiple identical or different intelligent agents.

[0258] It should be noted that the task assistance device provided by the present invention can execute the task assistance method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0259] The task assistance device provided by this invention offers a method for combining a multi-agent system based on a large language model with bioinformatics research for omics research. It also innovatively designs a group management method for agent groups, which can reduce information redundancy and meaningless dialogue between agents, ensure efficient and accurate information flow, and effectively improve the efficiency and accuracy of omics research between the two.

[0260] To effectively demonstrate the feasibility and advancement of the auxiliary method and apparatus provided by this invention, they were evaluated from multiple perspectives, including task planning evaluation, code quality evaluation, and task execution efficiency evaluation. The relevant evaluation methods and corresponding evaluation results are briefly described below.

[0261] To fully evaluate the performance of the task-assistance method and apparatus provided by the present invention, a wide range of evaluation metrics were used. For example, the system constructed using the present invention (hereinafter referred to as the system) was applied to all bioinformatics tasks in the workflow of constructing a large-scale human lung tissue cell atlas, and the performance results of the system were compared with those of the control group and the ablation group.

[0262] The control group consisted of AutoBA and AutoGen, where AutoBA is an influential single AI agent in the field of bioinformatics, and AutoGen is an influential multi-agent system. The ablation group included variants of our system that excluded knowledge empowerment related to interdisciplinary knowledge bases and reflective learning. By comparing the execution results, it was determined that our system achieved state-of-the-art overall performance in performing the aforementioned omics tasks.

[0263] Furthermore, the task planning evaluation employed was used to analyze the system's ability to decompose bioinformatics tasks and formulate task implementation processes. Evaluation indicators included clarity, technical feasibility, structure, comprehensiveness, rationality of method selection, risk management, data management, scientific rigor, resource allocation, and scalability. Clearly, the system achieved state-of-the-art task planning performance across all tasks. Compared to the control groups AutoBA and AutoGen, the task implementation processes generated by this system demonstrated higher executability in bioinformatics research tasks. This further confirms that intensive intelligence, through close collaboration among multiple agents and users, contributes to the design of excellent scientific research plans. Moreover, comparisons with the ablation group also indicate that knowledge empowerment, self-reflective learning, and literature retrieval can improve task planning capabilities.

[0264] The code quality assessment evaluates the system's ability to understand the specific implementation steps required for each described process step. Evaluation metrics include accuracy, efficiency, completeness, readability, logicality, portability, maintainability, usability, reusability, robustness, security, and scalability. Similarly, this system achieved state-of-the-art code quality performance across all omics tasks, demonstrating a more comprehensive understanding of process steps. Comparison with the control group shows that this system can generate higher-quality code for process steps. Comparison with the ablation group also indicates that knowledge empowerment and self-reflective learning are beneficial to the system's code generation.

[0265] The execution efficiency assessment evaluated the system's step execution capability from aspects such as task completeness, user intervention, runtime, and computational resource cost. The results showed that the system could complete all omics tasks based on the generated implementation steps, while the control group's methods failed to complete most of the implementation steps. User intervention represents the frequency of user guidance. The system completed all omics tasks with minimal user intervention, highlighting a high level of automation. Notably, AutoBA had no user intervention during execution, but its task completeness was low, indicating that appropriate user intervention can prevent task interruption or failure due to misunderstanding or insufficient information during the execution of complex tasks. AutoGen had the same number of user interventions as the system in all tasks, but its tasks were not completed. This also confirms the effectiveness of the human-computer interaction mechanism designed in this system, enabling it to handle complex bioinformatics tasks. Comparison with the ablation group shows that knowledge empowerment, self-reflective learning, and literature retrieval can effectively reduce user intervention. In all tasks, the system's overall runtime was shorter than AutoGen, but its task completeness was higher. Although AutoBA had less overall runtime, its task completeness was the lowest. In terms of resource consumption, the computing resources consumed by each system are within an acceptable range. 128 CPU cores, 32GB of memory, and a GPU with 24GB of memory can meet the system's usage requirements.

[0266] In summary, the task assistance method and apparatus provided by this invention have been thoroughly evaluated from three main aspects: task planning, code generation quality, and task execution efficiency. Of course, more evaluation indicators can be adopted to more comprehensively evaluate the system's performance, such as intelligent hierarchical classification.

[0267] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 5 As shown, the electronic device may include a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, communications interface 520, and memory 530 communicate with each other via the communication bus 540. The processor 510 can execute a task-assistance method using logical instructions in the memory 530. This method includes: acquiring a user-issued omics task; determining the task implementation process required to complete the omics task using coarse-grained planning groups; sequentially traversing each step of the task implementation process and performing the following operations: determining the specific implementation steps required to complete each step using fine-grained planning groups; and executing the specific implementation steps using action execution groups to obtain the target omics result after completing the specific implementation steps.

[0268] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0269] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer is able to execute the task-assisting method provided in the above embodiments, the method including: obtaining an omics task issued by a user; determining the task implementation process required to complete the omics task through coarse-grained planning groups; sequentially traversing each process step in the task implementation process and performing the following operations: determining the specific implementation steps required to complete each process step through fine-grained planning groups; executing the specific implementation steps through action execution groups to obtain the target omics result after completing the specific implementation steps.

[0270] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the task-assisting method provided in the above embodiments. The method includes: obtaining an omics task issued by a user; determining the task implementation process required to complete the omics task through a coarse-grained planning group; sequentially traversing each process step in the task implementation process and performing the following operations: determining the specific implementation steps required to complete each process step through a fine-grained planning group; and executing the specific implementation steps through an action execution group to obtain the target omics result after completing the specific implementation steps.

[0271] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0272] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0273] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A task-assistance method, characterized in that, include: Obtain omics tasks assigned by users; The task implementation process required to complete the omics task is determined by coarse-grained planning groups; The coarse-grained planning group consists of multiple intelligent agents with different domain knowledge and planning capabilities, which can perform preliminary decomposition and planning of the omics task from a macro perspective; Iterate through each step of the task implementation process sequentially and perform the following operations: The specific implementation steps required to complete each of the process steps are determined by a fine-grained planning group; the fine-grained planning group consists of multiple agents with detailed domain knowledge and execution capabilities, which can further decompose and refine each of the process steps from a micro perspective; the specific implementation steps are executed by an action execution group to obtain the target omics results after the specific implementation steps are completed; the action execution group consists of multiple agents with practical operation and execution capabilities, which can perform relevant operations according to the requirements of the specific implementation steps; The coarse-grained planning group, the fine-grained planning group, and the action execution group are groups composed of multiple identical or different intelligent agents; Receive information call requests from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group; In response to the information retrieval request, a connection is established for any intelligent agent to retrieve knowledge from an interdisciplinary knowledge base; The interdisciplinary knowledge base is constructed by extracting knowledge from various disciplines from the environment and encoding it in a form readable by intelligent agents.

2. The task assistance method according to claim 1, characterized in that, The coarse-grained planning group includes at least a coarse-grained planning agent, a first agent agent, and a coarse-grained auditing agent; The process of determining the task implementation required to complete the omics task through coarse-grained planning groups includes: The coarse-grained planning agent generates an initial implementation process based on the omics task. Obtain the first review result of the coarse-grained review agent on the initial implementation process; If it is determined that the first review result includes process improvement suggestions, then the coarse-grained planning agent is controlled to generate a new initial implementation process based on the process improvement suggestions, and the coarse-grained review agent re-reviews the new initial implementation process until the first review result is determined to be passed; The initial implementation process when the first review result is passed is sent to the first agent intelligent agent as the task implementation process, so as to initiate the first human-computer interaction through the first agent intelligent agent; The first human-computer interaction is used to obtain the user's adjustment suggestions for the task implementation process, and the adjustment suggestions are used by the coarse-grained planning agent to readjust the initial implementation process.

3. The task assistance method according to claim 2, characterized in that, The coarse-grained planning group is temporarily constructed based on the omics task, and different coarse-grained planning groups are constructed for different omics tasks.

4. The task assistance method according to claim 1, characterized in that, The fine-grained planning group includes at least a fine-grained planning agent, a second agent agent, and a fine-grained auditing agent; The determination of the specific implementation steps required to complete each process step through fine-grained planning groups includes: The fine-grained planning agent generates initial implementation steps according to the process steps. Obtain the second audit result of the fine-grained audit agent on the initial implementation step; If it is determined that the second review result includes step improvement suggestions, then the fine-grained planning agent is controlled to generate new initial implementation steps based on the step improvement suggestions, and the fine-grained review agent is used to re-review the new initial implementation steps until the second review result is determined to be passed; The initial implementation steps when the second review result is passed are sent to the second agent intelligent agent as the specific implementation steps, so that the second agent intelligent agent can initiate the second human-computer interaction; The second human-computer interaction is used to obtain the user's adjustment suggestions for the specific implementation steps, and the adjustment suggestions are used by the fine-grained planning agent to readjust the initial implementation steps.

5. The task assistance method according to claim 4, characterized in that, The fine-grained planning group is temporarily constructed based on the process steps, and different fine-grained planning groups are constructed for different process steps.

6. The task assistance method according to claim 1, characterized in that, The action execution group includes at least a coding agent, a third-party agent, and a debugging agent; The process of executing the specific implementation steps through an action execution group to obtain the target omics results after completing the specific implementation steps includes: The coded agent generates initial omics code according to the specific implementation steps. The initial omics code is debugged by the debugging agent, so that the coding agent regenerates new initial omics code based on the debugging results, until it is determined that the debugging result of the debugging agent on the initial omics code is successful; The initial omics code obtained when the debugging result is successful is used as the target omics code; The target omics code is executed by the third agent to generate the target omics results.

7. The task assistance method according to claim 6, characterized in that, After the coded agent generates the initial omics code according to the specific implementation steps, the method further includes: A third-party human-computer interaction is initiated through the third agent intelligent agent; The third human-computer interaction is used to obtain the user's suggestions for adjusting the initial omics code.

8. The task assistance method according to claim 7, characterized in that, After debugging the initial omics code through the debugging agent, the process further includes: If the debugging agent determines that the preset triggering conditions are met, then the fourth human-computer interaction is initiated through the third agent agent; The fourth human-computer interaction is used to obtain the user's adjustment suggestions for the debugging results, and the adjustment suggestions are used by the coding agent to readjust the initial omics code.

9. The task assistance method according to claim 8, characterized in that, The preset triggering conditions include: The debugging agent generates the same code error message when running the initial omics code twice consecutively. Alternatively, the number of code error messages generated by the debugging agent when running the initial omics code exceeds a preset threshold; Alternatively, the debugging agent may generate a code error message containing preset sensitive words when running the initial omics code; Alternatively, the debugging agent may lack necessary files or parameters when running the initial omics code.

10. The task assistance method according to claim 1, characterized in that, Also includes: Receive a self-learning information call request from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group; In response to the self-learning information call request, historical files related to any of the intelligent agents are sent to any of the intelligent agents so that any of the intelligent agents can reflect and learn based on the historical files; The historical files include human-computer interaction information and / or task interaction logs.

11. The task assistance method according to claim 1, characterized in that, After obtaining the target omics results after completing the specific implementation steps, the process also includes: The target omics results corresponding to each specific implementation step are stored as objects.

12. A task assistance device, characterized in that, include: The task receiving unit is used to obtain the omics tasks issued by the user; A coarse-grained auxiliary unit is used to determine the task implementation process required to complete the omics task through coarse-grained planning groups; The coarse-grained planning group consists of multiple intelligent agents with different domain knowledge and planning capabilities, which can perform preliminary decomposition and planning of the omics task from a macro perspective; The fine-grained auxiliary unit is used to sequentially traverse each process step in the task implementation process and perform the following operations: The specific implementation steps required to complete each of the process steps are determined by a fine-grained planning group; the fine-grained planning group consists of multiple agents with detailed domain knowledge and execution capabilities, which can further decompose and refine each of the process steps from a micro perspective; the specific implementation steps are executed by an action execution group to obtain the target omics results after the specific implementation steps are completed; the action execution group consists of multiple agents with practical operation and execution capabilities, which can perform relevant operations according to the requirements of the specific implementation steps; The coarse-grained planning group, the fine-grained planning group, and the action execution group are groups composed of multiple identical or different intelligent agents. Receive information retrieval requests from any agent in the coarse-grained planning group, the fine-grained planning group, or the action execution group; respond to the information retrieval requests and establish a connection for any agent to retrieve knowledge from an interdisciplinary knowledge base; the interdisciplinary knowledge base is constructed by extracting knowledge from various disciplines in the environment and encoding it in an agent-readable form.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the task-assisting method as described in any one of claims 1 to 11.

14. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the task-assisting method as described in any one of claims 1 to 11.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the task-assisting method as described in any one of claims 1 to 11.

Images