Scheduling methods, devices, equipment, media, and products based on large models

By using a large model to uniformly call and analyze the small business models of water plants, an efficient and accurate scheduling scheme is generated, which solves the problem that independent small models are difficult to directly transform into scheduling schemes, and realizes the generation of efficient and reliable scheduling schemes.

CN123067967APending Publication Date: 2026-11-10BEIJING BAIDU NETCOM SCI & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611160290.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-31
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing technologies, various business models of water plants exist as independent tools, independent pages, or independent services, making it difficult to directly transform model capabilities into executable scheduling schemes, resulting in low efficiency and insufficient accuracy.

Method used

The scheduling is performed using a large model, and the target small model is encapsulated through a unified calling protocol to generate a scheduling scheme, including parsing the scheduling problem, calling the target small model, and analyzing the processing results.

Benefits of technology

It improves the efficiency and accuracy of scheduling scheme generation, reduces the complexity of system operation, and enhances the reliability and automation of model invocation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067967A_ABST
    Figure CN123067967A_ABST
Patent Text Reader

Abstract

This disclosure provides a scheduling method, apparatus, device, medium, and product based on a large model, relating to the field of artificial intelligence technology, particularly to the fields of large models and intelligent agents. The scheduling method based on a large model includes: analyzing a scheduling problem based on a large model to obtain an analysis result; the analysis result includes a target skill, which contains invocation information of a target small model; based on the large model, invoking a target tool corresponding to the target small model according to the invocation information using a unified invocation protocol, and obtaining the target processing result of the target small model; the target tool is obtained by pre-encapsulating the target small model using the unified invocation protocol; and analyzing the target processing result based on the large model to generate a scheduling scheme. This disclosure can improve the efficiency and accuracy of scheduling scheme generation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, particularly to the fields of large models and intelligent agents, and specifically to a scheduling method, apparatus, device, medium, and product based on a large model. Background Technology

[0002] With the development of technology, taking water plants as an example, the operation of water plants has gradually developed various small business models, such as water distribution analysis models, production line water volume analysis models, and process simulation models.

[0003] These small models typically exist as standalone tools, standalone pages, or standalone services, making it difficult to directly translate model capabilities into executable scheduling solutions. Summary of the Invention

[0004] This disclosure provides a scheduling method, apparatus, scheduling agent, device, medium, and product based on a large model.

[0005] According to one aspect of this disclosure, a scheduling method based on a large model is provided, comprising: analyzing a scheduling problem based on the large model to obtain an analysis result; the analysis result including a target skill, wherein the target skill contains invocation information of a target small model; based on the large model, invoking a target tool corresponding to the target small model according to the invocation information through a unified invocation protocol, and obtaining a target processing result of the target small model; the target tool is obtained by pre-encapsulating the target small model using the unified invocation protocol; and analyzing the target processing result based on the large model to generate a scheduling scheme.

[0006] According to another aspect of this disclosure, a scheduling device based on a large model is provided, comprising: a parsing module, used to parse a scheduling problem based on the large model to obtain a parsing result; the parsing result includes a target skill, the target skill containing invocation information of a target sub-model; an acquisition module, used to, based on the large model, invoke a target tool corresponding to the target sub-model according to the invocation information via a unified invocation protocol, and obtain a target processing result of the target sub-model; the target tool is obtained by pre-encapsulating the target sub-model using the unified invocation protocol; and a generation module, used to analyze the target processing result based on the large model to generate a scheduling scheme.

[0007] According to another aspect of this disclosure, a scheduling agent is provided, comprising: an input module for inputting a scheduling problem; a processing module for parsing the scheduling problem based on a large model to obtain a parsing result; the parsing result including a target skill, the target skill containing invocation information of a target small model; based on the large model, invoking a target tool corresponding to the target small model according to the invocation information via a unified invocation protocol, and obtaining a target processing result of the target small model; the target tool being obtained by pre-encapsulating the target small model using the unified invocation protocol; analyzing the target processing result based on the large model to generate a scheduling scheme; and an output module for outputting the scheduling scheme.

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to said at least one processor; wherein the memory stores instructions executable by said at least one processor, said instructions being executed by said at least one processor to enable said at least one processor to perform the method as described in any of the foregoing aspects.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method according to any of the preceding aspects.

[0010] According to another aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method according to any of the preceding aspects.

[0011] According to embodiments of this disclosure, the efficiency and accuracy of generating scheduling schemes can be improved.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0013] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0014] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0015] Figure 2 This is an overall architecture diagram of the scheduling system provided according to embodiments of this disclosure;

[0016] Figure 3 This is a schematic diagram according to the second embodiment of the present disclosure;

[0017] Figure 4 This is a schematic diagram according to the third embodiment of the present disclosure;

[0018] Figure 5 This is a schematic diagram according to the fourth embodiment of the present disclosure;

[0019] Figure 6 This is a schematic diagram of an electronic device used to implement the large-model-based scheduling method of the embodiments of this disclosure. Detailed Implementation

[0020] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0021] An intelligent agent is an entity capable of perceiving its environment and autonomously taking actions to achieve specific goals. It can be software, hardware, or a system. An intelligent agent is not merely a simple tool, but an intelligent agent that proactively adapts to its environment and learns to optimize its strategies. Current development of intelligent agents is based on large models, such as the Large Language Model (LLM), as its core intelligent foundation. Its architecture typically includes modules for memory, planning, tool usage, and action execution.

[0022] Large Language Models (LLMs) are deep learning models trained on large amounts of text data. They are capable of generating natural language text or understanding the meaning of language text, and their parameter scale typically reaches billions to trillions.

[0023] Prompts are textual instructions input into an LLM (Low-Level Model) to guide the model in generating specific, accurate, or expected outputs. They are the core way humans interact with large models, determining the direction of the model's understanding of tasks and the content it generates. Specifically, prompts can include tasks, which explicitly tell the model what to do; context, which provides relevant information to help the model understand the scenario or conditions in which the task occurs; examples, which indicate the expected format or style of input and output; and roles, which indicate the role the model should play.

[0024] In related technologies, task scheduling is usually done manually. For example, schedulers manually compile information such as water volume, water quality, and pressure, and switch between multiple business mini-models to obtain the output results of the mini-models. Based on this information and the output results of the mini-models, they manually arrange the data to obtain a scheduling plan.

[0025] However, the above methods are inefficient. For example, dispatchers need to switch systems repeatedly and manually organize parameters and results. In addition, the accuracy needs to be improved due to limitations in human experience.

[0026] To improve the efficiency and accuracy of scheduling scheme generation, this disclosure provides the following embodiments.

[0027] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure. This embodiment provides a scheduling method based on a large model. Figure 1 As shown, the method includes:

[0028] 101. Based on the large model, the scheduling problem is analyzed to obtain the analysis results; the analysis results include: target skills, which contain the calling information of target small models.

[0029] 102. Based on the large model, the target tool corresponding to the target small model is invoked according to the invocation information through the unified invocation protocol, and the target processing result of the target small model is obtained; the target tool is obtained by pre-encapsulating the target small model using the unified invocation protocol.

[0030] 103. Based on the large model, analyze the target processing results to generate a scheduling scheme.

[0031] The scheduling problem can be natural language text input by the user, such as "I want to increase the water supply of the water plant to x tons / day, what adjustments are needed?"

[0032] A skill is a pre-packaged, reusable object used to represent a workflow corresponding to a specific ability. Specifically, it can be encapsulated from expert experience, known knowledge, and rules to obtain a skill.

[0033] Each Skill can contain the following:

[0034] Triggering conditions: Define the problem type, key intent and necessary parameters to which this skill applies, such as water supply target, effluent turbidity, influent turbidity, pump status, lower pressure limit, valve status, chemical inventory, substation operation mode, etc.

[0035] Data preparation logic: Define which data to obtain from real-time databases, historical databases, equipment ledgers, knowledge bases, or manual input, and handle missing data by asking follow-up questions or using default strategies.

[0036] Model call chain: Defines which models (small models) are called, the order of calls, the input-output mapping relationship, and the degradation strategy when the model returns an exception or has insufficient credibility.

[0037] Rule verification logic: Introduce process boundary, equipment safety boundary, water supply service boundary, electrical safety operation boundary and emergency plan constraints to check the legality and feasibility of candidate solutions.

[0038] Results presentation template: Convert the model output into structured content that schedulers can understand, including recommended solutions, adjustment amounts, expected effects, risk points, observation indicators, pre-execution confirmation items, and alternative solutions.

[0039] By using the Skill encapsulation method described above, scattered expert experience and model invocation experience can be solidified into reusable assets, enabling large models not only to "answer questions" but also to complete reliable task orchestration according to the water plant's business processes.

[0040] The target skill refers to the skill that corresponds to the current scheduling problem. For example, multiple candidate skills can be pre-packaged, and one or more of them can be selected as the target skill.

[0041] The target skill contains the calling information of the target mini-model. The target mini-model is the mini-model to be called. The calling information can be a model calling chain, which is used to represent which models need to be called and the calling order, such as calling model A first and then calling model B, etc.

[0042] Small models are models that are smaller in scale than large models. They are existing small business models for various scenarios. For example, in the water plant scenario, small models can include: water distribution analysis model, production line water volume analysis model, process simulation model, turbidity prediction model, alum dosing optimization model, flow prediction model, pressure prediction model, pump set matching optimization model, and other types of small business models.

[0043] The target tool is obtained by encapsulating the business mini-model using a unified calling protocol, such as the Model Context Protocol (MCP). For example, using MCP, the water distribution analysis model is encapsulated as tool 1, and the production line water volume analysis model is encapsulated as tool 2, etc.

[0044] After the target tool is invoked, the target tool uses its encapsulated business model to process the data. For example, after invoking tool 1, the water distribution analysis results, such as the water output of each device, are obtained through its encapsulated water distribution analysis model.

[0045] There can be one or more target tools. The large model comprehensively analyzes the target processing results obtained from the business sub-models encapsulated by each target tool to obtain a scheduling plan, such as the increase or decrease of water volume for each production line, pump start-up and shutdown combinations, pressure control range, expected arrival time, and key indicators that need to be observed.

[0046] In this embodiment, parsing, invocation, and scheduling scheme generation based on a large model improves efficiency and accuracy compared to manual methods. By including the invocation information of the target sub-models in the target skill, scheduling information can be directly obtained from the target skill without real-time orchestration of invocations of each business sub-model, thus improving processing efficiency. By encapsulating the business sub-models into tools using a unified invocation protocol, instead of the large model directly invoking the sub-models, the uncontrollable problem caused by the need to generate invocation code for each sub-model can be solved, improving invocation efficiency and reliability. Therefore, the efficiency and accuracy of scheduling scheme generation can be improved.

[0047] The following section uses a water plant process scenario as an example to explain the relevant content.

[0048] Figure 2 This is an overall architecture diagram of the scheduling system provided according to embodiments of this disclosure.

[0049] This embodiment takes a small business model encapsulated using MCP as an example, such as... Figure 2 As shown, the system 200 mainly includes: a business scenario routing module 201, a skill library 202, and an MCP tool library 203.

[0050] The business scenario routing module 201 is used to determine the target business scenario corresponding to the scheduling problem based on the large model. For example, the target business scenario may be a water supply and demand balance scenario, a water quality adjustment scenario, a control optimization scenario, or an emergency event handling scenario.

[0051] Skill library 202 is used to store at least one skill and provide the target skill corresponding to the target business scenario. For example, the target skill corresponding to the horizontal supply and demand balance scenario is skill 1, and the target skill corresponding to the water quality control scenario is skill 2, etc.

[0052] Each skill predefines a model call chain, which specifies the models to be called and their order. For example, skill 1 predefines: first call the configuration analysis model, then the production line water volume analysis model, and finally the process simulation model. The order of model calls is also not important; for example, skill 2 predefines: call the turbidity prediction model and the alum dosing optimization model.

[0053] MCP Tool Library 203 is used to store at least one model tool. Each business mini-model can be pre-packaged into one or more MCP tools and provide tool description information. The tool description information defines the mini-model name, input fields, field units, constraint range, required conditions, output structure, exception codes, and calling permissions, etc.

[0054] In conjunction with the above overall architecture, this disclosure also provides the following embodiments.

[0055] Figure 3 This is a schematic diagram based on the second embodiment of the present disclosure. This embodiment provides a scheduling method based on a large model. Figure 3 As shown, the method includes:

[0056] 301. Based on the large model, perform scenario identification on the scheduling problem to obtain the target business scenario, and extract the initial input parameters in the scheduling problem.

[0057] 302. Obtain the target skill corresponding to the target business scenario, wherein the target skill contains the calling information of the target mini-model.

[0058] When the system receives a scheduling question from the user, it sends it to the large model via prompts and instructs the large model to select the target business scenario from the existing business scenarios, thereby obtaining the target skill corresponding to the target business scenario.

[0059] For example, existing business scenarios include: water supply and demand balance, water quality regulation, control optimization and consumption reduction, substation switching and emergency event handling, etc. The target business scenario is selected from the above scenarios.

[0060] Based on this, assuming the target business scenario is a water supply and demand balance scenario, the target skill can be skill 1; assuming the target business scenario is a water quality adjustment scenario, the target skill can be skill 2, etc.

[0061] A particular scenario may involve one or more target skills, and different scenarios may involve the same or different skills.

[0062] In this way, by obtaining the target business scenario and then the target skill based on a large model, the search volume can be reduced and the processing efficiency can be improved compared to selecting from all candidate skills corresponding to all scenarios.

[0063] In some embodiments, candidate skills corresponding to the target business scenario can be obtained based on a pre-configured mapping relationship between scenarios and candidate skills; and the target skill can be obtained from the candidate skills based on the scheduling problem.

[0064] Among them, the applicable scenarios can be predefined in the skill, so after the target business scenario is determined, candidate skills are obtained according to the scenario information defined in the skill. Then, according to the scheduling problem, one or more of the candidate skills are selected as the target skill.

[0065] For example, for the same "water balance" scenario, the candidate skills that can be adapted to the scenario include skill 1 and skill 2. Skill 1 is used for water demand analysis, and skill 2 is used for pump control scheme generation. If the scheduling problem is to query water demand, then the target skill is skill 1. If the scheduling problem involves pump control, then the target skill can include skill 2, or skill 1 and skill 2.

[0066] Each skill also includes descriptive information, such as the type of problem to be adapted and the function. Semantic similarity is calculated based on the scheduling problem and descriptive information, and one or more target skills are selected based on semantic similarity.

[0067] In this way, by first obtaining candidate skills based on mapping relationships and then obtaining target skills based on semantics, a balance between processing speed and accuracy can be achieved.

[0068] In addition, prompt words can be used to instruct the large model to extract input parameters. The input parameters extracted by the large model from the scheduling problem can be called initial input parameters, which may include one or more of the following: scheduling objectives (such as water supply to x tons / day), time range (such as within 24 hours), constraints (which can be set by the user as needed), key equipment (such as equipment that the user needs to schedule), key indicators (which the user can specify), and expected output format (which the user can specify).

[0069] 303. Based on the target skill, process the initial input parameters to obtain standard input parameters.

[0070] Once the target skill is acquired, the initial input parameters can be processed based on the predefined information within the target skill to obtain standard input parameters. For example, the target skill predefines data preparation logic and the input format of the target tool. Based on this, the completeness of the input parameters can be checked through the target skill, missing information can be investigated, or the missing information can be automatically filled in from real-time databases, historical databases, equipment ledgers, and process rule bases to obtain standard input parameters that conform to the input format of the target tool.

[0071] In this way, the initial input parameters can be automatically completed and standardized by using the target skills, thereby improving the degree of automation and accuracy.

[0072] 304. Based on the large model, through the unified calling protocol and according to the calling information, the standard input parameters are transmitted to the target small model in the target tool, and the target processing result is obtained.

[0073] Among them, the target skills predefine the scheduling information between target tools, such as the tool scheduling chain. The large model can call each target tool in sequence according to the call chain and obtain the target processing results obtained by the business mini-models encapsulated in each target tool.

[0074] 305. Based on the large model, analyze the target processing results to generate a scheduling scheme.

[0075] In some embodiments, the target processing result can be verified based on the target skill; in response to the target processing result passing the verification, the target processing result is analyzed to generate a scheduling scheme.

[0076] After the large model obtains the processing results of each business sub-model, it can first perform verification, such as verification based on the predefined verification rules in the target skill. Specifically, it can perform verifications such as credibility checks, boundary checks, and conflict resolution. Then, it can generate a scheduling scheme based on the processing results that have passed the verification.

[0077] In this way, by verifying the target processing results obtained from the small model, the reliability of the target processing results can be improved, thereby improving the reliability of the scheduling scheme.

[0078] In addition, after generating the scheduling scheme, subsequent processing can be carried out according to the predefined processing rules in the target skill, such as sorting multiple schemes when there are multiple scheduling schemes, and eliminating scheduling schemes that violate process, safety or operational constraints.

[0079] When generating scheduling schemes, large models can also provide reasons for recommendation, expected results, execution steps, risk warnings, and alternative schemes.

[0080] Then, the above scheduling scheme and related information will be displayed to the user.

[0081] In addition, the system can also record relevant data, such as scheduling schemes, user adoption status, execution feedback, and subsequent indicator changes. This relevant data can be used to optimize skills, knowledge bases, and business mini-models.

[0082] For specific applications, please refer to the following scenario descriptions:

[0083] (I) Water Supply and Demand Balance Scenario. When users raise questions such as increasing water supply, adjusting production line load, meeting regional pressure, or predicting future outflow of water, the plugin first extracts parameters such as target water supply, target time period, pressure constraints, water tank level, pump status, and pipeline boundaries. Then, it calls water distribution analysis algorithms, production line water volume analysis algorithms, flow prediction algorithms, and pressure prediction algorithms to calculate the adjustment space for different production lines, pump stations, valves, and water distribution areas. If the user's target may touch upon process or equipment boundaries, the system calls process simulation tools for verification and finally outputs step-by-step adjustment suggestions, such as increasing or decreasing water supply for each production line, pump start / stop combinations, pressure control ranges, expected arrival times, and key indicators requiring close monitoring.

[0084] (II) Water Quality Control Scenarios. When users raise questions such as turbidity prediction, alum dosing recommendations, filter sand replacement impact assessment, and changes in influent turbidity caused by upstream rainstorms, the plugin automatically invokes the turbidity prediction algorithm, alum dosing optimization algorithm, process simulation system, and relevant historical operating condition database. The system uses influent turbidity, effluent turbidity target, decline trend, influent flow rate, sedimentation tank operating status, upper limit of chemical dosing, and historical similar operating conditions as context to generate recommended chemical dosing, adjustment frequency, expected turbidity change curve, and risk warning. For example, for the question "The effluent turbidity will gradually decline in the next 30 minutes, the influent flow rate is stable, and the current chemical dosing is 24 mg / L," the system can output whether to reduce the chemical dosing, the reduction range, the observation window, and conditions under which adjustment is not recommended.

[0085] (III) Control Optimization and Energy Reduction Scenarios. When users raise questions such as ensuring water supply pressure after pump set failure, selecting the lowest energy consumption combination under a given water supply volume, and predicting pressure and pump set matching for the next 24 hours, the plug-in calls the flow prediction algorithm, pressure prediction algorithm, pump set matching optimization algorithm, and equipment status data. Under the premise of meeting the target pressure, target water volume, pump set availability, number of start-stop cycles, electricity price periods, and safety redundancy, the system generates recommended pump set combinations, operating frequencies, switching times, energy consumption estimates, and alternative solutions, thereby achieving control optimization and energy reduction.

[0086] (iv) Substation switching scenarios. For switching operations in water plant power distribution systems, this invention can set a switching operation verification Skill. This Skill receives the current primary wiring method, load status, equipment to be energized / discontinued, protection status, operation ticket rules, and operating constraints. It connects to the electrical ledger, real-time load data, and switching rule library through MCP, automatically generating or verifying the switching operation sequence, and prompting interlock conditions, load transfer impact, prohibited operation conditions, and safety items requiring manual confirmation, avoiding the risks caused by relying solely on manual memory.

[0087] (V) Emergency Situation Simulation. In the event of emergencies such as torrential rain, upstream pollution, sudden pump failure, abnormal pipeline pressure, filter malfunction, insufficient reagents, or water source switching, the plugin invokes the emergency response plan knowledge base, process simulation tools, and relevant prediction algorithms to simulate changes in water volume, water quality, pressure, energy consumption, and risk under different treatment options. The system outputs recommended treatment plans, critical time windows, resource requirements, risk levels, and the positions requiring reporting or coordination to support emergency dispatch decisions.

[0088] In this embodiment, MCP (Multi-Channel Programming) unifies the access of various business models, enabling previously isolated prediction, simulation, and optimization tools to perform joint calculations and result fusion around the same business objective, thereby enhancing the collaborative capabilities of the small models. Schedulers no longer need to switch between multiple systems; they can obtain cross-system and cross-model aggregated scheduling suggestions simply by describing the objective and constraints in natural language, thus reducing system operational complexity. By accumulating expert experience, rule boundaries, and operational templates, complex scheduling tasks can be executed in a standardized manner, and new users can also complete analysis and handling according to system suggestions, thereby reducing reliance on expert experience. Before outputting suggestions, the system verifies them in conjunction with real-time equipment status, process constraints, safety boundaries, and emergency plans to avoid a disconnect between single model output and on-site operation, thereby improving the feasibility of the solution. Through capabilities such as pump group matching optimization, pressure prediction, chemical dosing optimization, and emergency simulation, energy consumption, chemical consumption, and the risk of misoperation are reduced while ensuring water supply quantity and quality safety, thus supporting energy conservation, consumption reduction, and risk control. The system records each problem, model call, recommended solution, manual adoption, and execution effect, providing a basis for subsequent process model optimization, knowledge base updates, and scheduling strategy skill optimization, thus forming a reviewable intelligent scheduling closed loop.

[0089] Figure 4 The diagram is based on a third embodiment of the present disclosure. This embodiment provides a scheduling device based on a large model. The device 400 includes: a parsing module 401, an acquisition module 402, and a generation module 403.

[0090] The parsing module 401 is used to parse the scheduling problem based on the large model to obtain the parsing result; the parsing result includes: target skill, which contains the calling information of target small model; the acquisition module 402 is used to, based on the large model, call the target tool corresponding to the target small model according to the calling information through a unified calling protocol, and obtain the target processing result of the target small model; the target tool is obtained by pre-encapsulating the target small model using the unified calling protocol; the generation module 403 is used to analyze the target processing result based on the large model to generate a scheduling scheme.

[0091] In this embodiment, parsing, invocation, and scheduling scheme generation based on a large model improves efficiency and accuracy compared to manual methods. By including the invocation information of the target sub-models in the target skill, scheduling information can be directly obtained from the target skill without real-time orchestration of invocations of each business sub-model, thus improving processing efficiency. By encapsulating the business sub-models into tools using a unified invocation protocol, instead of the large model directly invoking the sub-models, the uncontrollable problem caused by the need to generate invocation code for each sub-model can be solved, improving invocation efficiency and reliability. Therefore, the efficiency and accuracy of scheduling scheme generation can be improved.

[0092] In some embodiments, the parsing module 401 is further configured to:

[0093] Based on the large model, scenario identification is performed on the scheduling problem to obtain the target business scenario;

[0094] Obtain the target skills corresponding to the target business scenario.

[0095] In this embodiment, by obtaining the target business scenario based on a large model and then obtaining the target skill, the retrieval volume can be reduced and the processing efficiency can be improved compared to selecting from all candidate skills corresponding to all scenarios.

[0096] In some embodiments, the parsing module 401 is further configured to:

[0097] Based on the pre-configured mapping relationship between scenarios and candidate skills, obtain the candidate skills corresponding to the target business scenario;

[0098] Based on the scheduling problem, the target skill is obtained from the candidate skills.

[0099] In this embodiment, candidate skills are obtained first based on mapping relationships, and then target skills are obtained based on semantics, which can balance processing speed and accuracy.

[0100] In some embodiments, the parsing result further includes: initial input parameters; the acquisition module 402 is further used for:

[0101] Based on the target skill, the initial input parameters are processed to obtain standard input parameters;

[0102] Based on the large model, the standard input parameters are transmitted to the target small model in the target tool according to the call information through the unified calling protocol, and the target processing result is obtained.

[0103] In this embodiment, the initial input parameters can be automatically completed and standardized by using the target skill, thereby improving the degree of automation and accuracy.

[0104] In some embodiments, the generation module 403 is further configured to:

[0105] The target processing result is verified based on the target skill;

[0106] In response to the verification of the target processing result, the target processing result is analyzed to generate a scheduling scheme.

[0107] In this embodiment, by verifying the target processing results obtained from the small model, the reliability of the target processing results can be improved, thereby improving the reliability of the scheduling scheme.

[0108] Figure 5 This is a schematic diagram based on the fourth embodiment of the present disclosure. This embodiment provides an intelligent agent 500, which includes an input module 501, a processing module 502, and an output module 503.

[0109] Input module 501 is used to input a scheduling problem; processing module 502 is used to parse the scheduling problem based on a large model to obtain a parsing result; the parsing result includes: a target skill, which contains the calling information of a target sub-model; based on the large model, using a unified calling protocol, calling the target tool corresponding to the target sub-model according to the calling information, and obtaining the target processing result of the target sub-model; the target tool is obtained by pre-encapsulating the target sub-model using the unified calling protocol; based on the large model, the target processing result is analyzed to generate a scheduling scheme; output module 503 is used to output the scheduling scheme.

[0110] In this embodiment, parsing, invocation, and scheduling scheme generation based on a large model improves efficiency and accuracy compared to manual methods. By including the invocation information of the target sub-models in the target skill, scheduling information can be directly obtained from the target skill without real-time orchestration of invocations of each business sub-model, thus improving processing efficiency. By encapsulating the business sub-models into tools using a unified invocation protocol, instead of the large model directly invoking the sub-models, the uncontrollable problem caused by the need to generate invocation code for each sub-model can be solved, improving invocation efficiency and reliability. Therefore, the efficiency and accuracy of scheduling scheme generation can be improved.

[0111] In some embodiments, the processing module 502 is further configured to:

[0112] Based on the large model, scenario identification is performed on the scheduling problem to obtain the target business scenario;

[0113] Obtain the target skills corresponding to the target business scenario.

[0114] In this embodiment, by obtaining the target business scenario based on a large model and then obtaining the target skill, the retrieval volume can be reduced and the processing efficiency can be improved compared to selecting from all candidate skills corresponding to all scenarios.

[0115] In some embodiments, the processing module 502 is further configured to:

[0116] Based on the pre-configured mapping relationship between scenarios and candidate skills, obtain the candidate skills corresponding to the target business scenario;

[0117] Based on the scheduling problem, the target skill is obtained from the candidate skills.

[0118] In this embodiment, candidate skills are obtained first based on mapping relationships, and then target skills are obtained based on semantics, which can balance processing speed and accuracy.

[0119] In some embodiments, the parsing result further includes: initial input parameters; the processing module 502 is further used for:

[0120] Based on the target skill, the initial input parameters are processed to obtain standard input parameters;

[0121] Based on the large model, the standard input parameters are transmitted to the target small model in the target tool according to the call information through the unified calling protocol, and the target processing result is obtained.

[0122] In this embodiment, the initial input parameters can be automatically completed and standardized by using the target skill, thereby improving the degree of automation and accuracy.

[0123] In some embodiments, the processing module 502 is further configured to:

[0124] The target processing result is verified based on the target skill;

[0125] In response to the verification of the target processing result, the target processing result is analyzed to generate a scheduling scheme.

[0126] In this embodiment, by verifying the target processing results obtained from the small model, the reliability of the target processing results can be improved, thereby improving the reliability of the scheduling scheme.

[0127] It is understood that the same or similar content in different embodiments of this disclosure can be referred to each other.

[0128] It is understood that the terms "first" and "second" in the embodiments of this disclosure are only used for distinction and do not indicate the degree of importance or the order of events.

[0129] It is understandable that, unless otherwise specified, the order of steps in the process indicates that the temporal relationship between these steps is not limited.

[0130] The technical solutions disclosed herein involve the collection, storage, use, processing, transmission, provision, and disclosure of various types of information, such as user personal information, in accordance with relevant laws and regulations and do not violate public order and good morals.

[0131] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0132] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device 600 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0133] like Figure 6 As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0134] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of displays, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0135] The computing unit 601 can be a variety of general-purpose and / or proprietary processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various proprietary artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the large model-based scheduling method. For example, in some embodiments, the large model-based scheduling method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the large model-based scheduling method described above may be performed. Alternatively, in other embodiments, computing unit 601 may be configured to perform a large-model-based scheduling method by any other suitable means (e.g., by means of firmware).

[0136] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), proprietary integrated circuits (ASICs), proprietary standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a proprietary or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0137] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable task processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0138] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0139] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0140] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0141] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0142] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0143] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A scheduling method based on a large model, comprising: Based on a large model, the scheduling problem is analyzed to obtain analytical results; The analysis results include: target skills, which contain the calling information of the target mini-model; Based on the large model, the target tool corresponding to the target small model is invoked according to the invocation information through a unified invocation protocol, and the target processing result of the target small model is obtained; the target tool is obtained by pre-encapsulating the target small model using the unified invocation protocol. Based on the large model, the target processing results are analyzed to generate a scheduling scheme.

2. The method according to claim 1, wherein, The method based on a large model analyzes the scheduling problem to obtain analytical results, including: Based on the large model, scenario identification is performed on the scheduling problem to obtain the target business scenario; Obtain the target skills corresponding to the target business scenario.

3. The method according to claim 2, wherein, The acquisition of the target skill corresponding to the target business scenario includes: Based on the pre-configured mapping relationship between scenarios and candidate skills, obtain the candidate skills corresponding to the target business scenario; Based on the scheduling problem, the target skill is obtained from the candidate skills.

4. The method according to claim 1, wherein, The analysis results also include: initial input parameters; Based on the large model, the process involves using a unified invocation protocol to invoke the target tool corresponding to the target small model according to the invocation information, and obtaining the target processing result of the target small model, including: Based on the target skill, the initial input parameters are processed to obtain standard input parameters; Based on the large model, the standard input parameters are transmitted to the target small model in the target tool according to the call information through the unified calling protocol, and the target processing result is obtained.

5. The method according to any one of claims 1-4, wherein, The step of analyzing the target processing results based on the large model to generate a scheduling scheme includes: The target processing result is verified based on the target skill; In response to the verification of the target processing result, the target processing result is analyzed to generate a scheduling scheme.

6. A scheduling device based on a large model, comprising: The parsing module is used to analyze the scheduling problem based on a large model in order to obtain the analysis results; The analysis results include: target skills, which contain the calling information of the target mini-model; The acquisition module is used to, based on the large model, invoke the target tool corresponding to the target small model according to the invocation information through a unified invocation protocol, and obtain the target processing result of the target small model; the target tool is obtained by pre-encapsulating the target small model using the unified invocation protocol; The generation module is used to analyze the target processing results based on the large model in order to generate a scheduling scheme.

7. A scheduling agent, comprising: The input module is used to input scheduling problems; The processing module is used to analyze the scheduling problem based on a large model in order to obtain the analysis results; The analysis results include: target skills, which contain call information for target tools, which are obtained by encapsulating a business mini-model using a unified call protocol; based on the large model, the target tool is called according to the call information through the unified call protocol to obtain the target processing result of the corresponding business mini-model; based on the large model, the target processing result is analyzed to generate a scheduling scheme. The output module is used to output the scheduling scheme.

8. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-5.

10. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-5.