Cloud resource arrangement method, system and device based on LLM model, and medium

Abstract

The invention discloses a cloud resource arrangement method, system and device based on an LLM model, and a medium, and the method comprises the following steps: employing a natural language processing technology to process the demands of business personnel, extracting key information, inputting the key information into the LLM model to understand semantics and intentions, and carrying out the structural processing, and obtaining the structural demands; matching and mapping the structured demand with a predefined template library, generating a new template by an LLM model if the structured demand cannot be directly matched, and adjusting and optimizing parameters; establishing an interface connection with each cloud platform, performing identity verification and authorization, sending a resource opening request according to the resource opening template, and monitoring a state and processing an exception at the same time; and integrating common tools, triggering execution of an initialization script configuration task according to a resource opening result, and monitoring an execution process and result to ensure that the execution process and result conform to expectation. According to the scheme, efficient arrangement of cloud resources can be realized.

Description

Technical Field

[0001] The present invention relates to the field of multi-cloud management and cloud computing technology, and in particular to a cloud resource orchestration method, system, device and medium based on an LLM model. Background Art

[0002] In the existing enterprise environment, there are generally multiple clouds, such as Alibaba Cloud, Huawei Cloud, Tencent Cloud, etc., and business systems are usually distributed on multiple cloud platforms, using dozens of cloud resources. The deployment of business systems in test environments, development environments, and production environments often requires business personnel to raise requirements, fill out resource forms, and operation and maintenance personnel to allocate resources and complete resource information, open resources, and then initialize the configuration of resources. This involves multiple scripts. The entire process is complex and lengthy, and the final resources may not meet the requirements and need to be started again. This results in slow and inefficient construction of the business environment, and operation and maintenance personnel need to repeatedly configure scripts for the same environment.

[0003] The original process of multi-cloud management in the existing enterprise environment is as follows: Figure 1 As shown, the following problems may exist under the current process:

[0004] 1) Business personnel are not clear about the specific configuration of resource activation and cannot describe in detail the resources required for the business;

[0005] 2) Operation and maintenance personnel often need to repeatedly build, modify, and upgrade different environments of the same system, which is time-consuming and labor-intensive;

[0006] 3) For multi-cloud scenarios, operation and maintenance personnel need to switch between multiple platforms multiple times, and inconsistent operation logic may lead to operational errors. Summary of the invention

[0007] The purpose of the present invention is to provide a cloud resource orchestration method, system, device and medium based on the LLM model to improve the processing efficiency of operation and maintenance personnel.

[0008] The solution of the present invention can solve the problem that business personnel cannot accurately propose business system resource activation requirements, resulting in the mismatch between resources and requirements and the need to modify and rebuild resources back and forth, and accurately analyze business system requirements through natural language models;

[0009] The solution of the present invention can solve the problem that operation and maintenance personnel need to repeatedly configure different resource environments, and can deploy and copy the same business system to multiple environments by modifying some parameters in different environments.

[0010] The solution of the present invention can solve the problem of operation and maintenance personnel constantly switching platforms to process needs across multi-cloud environments. It uses models to connect with cloud platform interfaces, automated operation and maintenance tools, and process engines to achieve fully automated generation of needs to resources, thereby improving the processing efficiency of operation and maintenance personnel.

[0011] To achieve the above purpose, the technical solution provided by the present invention is:

[0012] The first aspect of the present application provides a cloud resource orchestration method based on the LLM model, comprising the following steps:

[0013] Use natural language processing technology to process the needs of business personnel, extract key information, input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured needs;

[0014] Match and map structured requirements with pre-defined template libraries. If they cannot be matched directly, new templates are generated by the LLM model and parameter adjustment and optimization are performed.

[0015] Establish interface connections with each cloud platform, perform identity authentication and authorization, send resource activation requests based on resource activation templates, monitor status and handle exceptions;

[0016] Integrate common tools to trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations.

[0017] To optimize the above technical solutions, the specific measures taken also include:

[0018] Furthermore, the use of natural language processing technology to process the needs of business personnel includes:

[0019] Data preprocessing: Clean the demand text entered by business personnel to remove noise and irrelevant characters;

[0020] Word segmentation and part-of-speech tagging: Use appropriate word segmentation algorithms to segment the demand text into words or phrases, and perform part-of-speech tagging for each word;

[0021] Named Entity Recognition: Identify specific entities in the requirements.

[0022] Furthermore, the extraction of key information and inputting it into the LLM model to understand semantics and intent includes:

[0023] Feature extraction: extract key features from word segmentation and named entities;

[0024] Input the LLM model: Input the extracted feature vector into the LLM model so that the LLM model can identify the business scenario;

[0025] Semantic understanding and intent analysis: With the help of the LLM model, the input features are learned and inferred to understand the semantics and intent of the requirements.

[0026] The structured processing to obtain structured requirements includes: converting the understood requirements into a structured data format, where the data format contains key information: application type, resource scale, performance requirements, and time constraints.

[0027] Furthermore, the structured requirements are matched and mapped with the predefined template library, new templates are generated by the LLM model if they cannot be directly matched, and parameter adjustment and optimization are performed, including:

[0028] Template library establishment and maintenance: collect and organize various common resource activation templates, covering different application scenarios and demand types; regularly update and optimize templates to adapt to new technologies and business needs;

[0029] Demand and template matching: Compare and match structured requirements with templates in the template library one by one, calculate the similarity between structured requirements and templates, and deploy templates that meet the similarity requirements;

[0030] Template generation and adjustment: For structured requirements that do not meet similarity requirements and cannot be directly matched, the generation capability of the LLM model is used to refer to similar historical requirements and templates to generate preliminary resource activation templates; the generated templates are adjusted and optimized according to the specific parameters in the structured requirements;

[0031] Template verification and confirmation: Verify the generated or adjusted template to check its rationality and completeness.

[0032] Furthermore, the establishment of interface connections with each cloud platform, identity authentication and authorization, sending resource activation requests according to resource activation templates, and monitoring status and handling exceptions include:

[0033] Interface integration and configuration: Obtain interface documents and development guides for different cloud platforms, and integrate and configure interfaces as required;

[0034] Send resource activation request: According to the converted resource activation template, construct the request parameters that meet the requirements of the cloud platform interface; use the corresponding API call method to send the resource activation request to the cloud platform;

[0035] Call status monitoring: Establish a real-time monitoring mechanism, regularly poll the interface to obtain call status information; handle possible request timeouts and error response exceptions;

[0036] Error handling and retry: When an error occurs in an interface call, classify and handle it according to the error code and error message; for errors that can be retried, set a reasonable retry strategy and interval time.

[0037] Furthermore, the integrated common tool triggers the execution of the initialization script configuration task according to the resource activation result, and monitors the execution process and results to ensure that they meet expectations, including:

[0038] Tool selection and integration: Evaluate and select appropriate automated operation and maintenance tools, complete tool installation and configuration, and ensure compatibility with the system environment;

[0039] Script writing and management: Write corresponding scripts for different resource initialization requirements, and perform version control and management on the scripts;

[0040] Trigger execution and monitoring: When cloud platform resources are successfully activated, the automated operation and maintenance tools are triggered in time to execute the corresponding scripts; the execution progress and results of the scripts are monitored, and the execution logs are output for viewing and analysis;

[0041] Result feedback and adjustment: Feedback is given to the resource template conversion module and the demand understanding module based on the results of script execution; for unsatisfactory execution, the reasons are analyzed and subsequent processes are adjusted and optimized.

[0042] The second aspect of the present application provides a cloud resource orchestration system based on LLM, including:

[0043] The demand understanding and analysis module is used to process the needs of business personnel using natural language processing technology, extract key information and input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured requirements;

[0044] Resource template conversion module, which is used to match and map structured requirements with pre-defined template libraries. If there is no direct match, a new template is generated by the LLM model, and parameter adjustment and optimization are performed;

[0045] The cloud platform interface calling module is used to establish interface connections with various cloud platforms, perform identity authentication and authorization, send resource activation requests based on resource activation templates, and monitor status and handle exceptions.

[0046] The automated operation and maintenance tool calling module is used to integrate common tools, trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations;

[0047] The LLM-based cloud resource orchestration system is used to execute the LLM model-based cloud resource orchestration method as described in any one of claims 1 to 7.

[0048] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the cloud resource orchestration method based on the LLM model as described above is implemented.

[0049] A fourth aspect of the present application provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the cloud resource orchestration method based on the LLM model as described above.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The present invention utilizes LLM to understand and parse requirements. Compared with the traditional requirement parsing method based on rules and templates, the LLM model adopted by the present invention has stronger semantic understanding and generalization capabilities, can handle complex and diverse natural language expressions, accurately extract key information in the requirements of business personnel, and reduce the understanding deviation caused by vague or non-standard requirement expressions; the present invention adopts deep learning technology to learn language patterns and semantic relationships from a large amount of text data, so as to more accurately understand the intention of business requirements and improve the accuracy and comprehensiveness of requirement parsing.

[0052] The present invention performs intelligent resource template conversion, which is different from the traditional fixed template matching method. It can flexibly generate or adjust resource activation templates according to business needs, can better adapt to different business scenarios and personalized needs, and improve the rationality and efficiency of resource allocation; the present invention uses the generation ability of LLM and the learning of historical needs to provide innovative template solutions for emerging or special needs, breaking through the limitations of template libraries in the prior art.

[0053] The present invention includes efficient cloud platform interface calls, and realizes integration and unified management with multiple cloud platforms through Terraform, avoiding the complexity and repetitive work of developing interface call logic for each cloud platform separately; the present invention can timely discover and solve problems in the resource activation process by real-time monitoring of interface call status and processing errors, thereby improving the success rate and stability of resource activation, and reducing time waste and business delays caused by interface call failures compared with the prior art.

[0054] The present invention includes automated operation and maintenance tool calls, which replaces the manual resource initialization configuration process, greatly improves work efficiency and reduces the possibility of human error; the present invention can trigger corresponding script configurations in real time according to the results of cloud resource activation, and realizes the full process automation of resource orchestration. Compared with the step-by-step, manual operation method in the prior art, it greatly shortens the time for resources to go online and improves business response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 : Schematic diagram of the original process of multi-cloud management in the existing enterprise environment.

[0056] Figure 2: Schematic diagram of the cloud resource orchestration method based on the LLM model of the present invention. DETAILED DESCRIPTION

[0057] The above contents of the present invention are further described in detail below in the form of embodiments, but this should not be understood as the scope of the above subject matter of the present invention being limited to the following embodiments, and all technologies realized based on the above contents of the present invention belong to the scope of the present invention.

[0058] Some of the terms involved in the solution of the present invention are explained as follows:

[0059] LLM (Large Language Model) refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text.

[0060] Terraform, Terraform is an open source "infrastructure as code" tool that can describe and automate the provisioning of infrastructure.

[0061] Provider, Provider is a module of Terraform, which is used to connect to Terraform to call the cloud platform interface, generally provided by various cloud vendors.

[0062] Ansible is an extremely powerful and widely used open source automated operation and maintenance tool that works through the SSH protocol without installing agent programs on the managed nodes.

[0063] Puppet, Puppet is an open source tool for server configuration management that can manage various system resources such as files, services, users, and software packages.

[0064] Chef is also an automated configuration management tool with good flexibility and scalability, and can adapt to complex environments and special needs.

[0065] In one embodiment of the present invention, a cloud resource orchestration method based on the LLM model is proposed, and the flow chart is as follows: Figure 2 As shown in the figure, the entire cloud resource orchestration method based on the LLM model includes the following steps:

[0066] Use natural language processing technology to process the needs of business personnel, extract key information, input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured needs;

[0067] Match and map structured requirements with pre-defined template libraries. If they cannot be matched directly, new templates are generated by the LLM model and parameter adjustment and optimization are performed.

[0068] Establish interface connections with each cloud platform, perform identity authentication and authorization, send resource activation requests based on resource activation templates, monitor status and handle exceptions;

[0069] Integrate common tools to trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations.

[0070] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods:

[0071] 1. Understanding and analyzing requirements

[0072] Step 1: Data Preprocessing

[0073] Clean the demand text entered by the business personnel to remove noise and irrelevant characters. For example, remove special symbols, extra spaces, etc. For example, if you enter "I need to deploy a high-performance live broadcast platform, including 10 virtual machines with specifications of 8 cores and 16G; 2 load balancers; 1 high-performance cloud database RDS; 1 message queue", the processing result is "I need to deploy a high-performance live broadcast platform, including 10 virtual machines with specifications of 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue".

[0074] Step 2: Word segmentation and part-of-speech tagging

[0075] Use appropriate word segmentation algorithms to segment the demand text into words or phrases. Perform part-of-speech tagging for each word, such as nouns, verbs, adjectives, etc. For example, "I need to deploy a high-performance live broadcast platform, 10 virtual machines with specifications of 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue." The word segmentation is "I" (noun), "need" (verb), "deploy" (verb), "a set" (quantifier), "high performance" (adjective), "of" (particle), "live broadcast platform" (noun), "virtual machine" (noun), "10" (quantifier), "8 cores and 16G" (noun), "load balancing" (noun), "2" (quantifier), "high performance" (adjective), "of" (particle), "cloud database RDS" (noun), "1" (quantifier), "message queue" (noun), "1" (quantifier).

[0076] Step 3: Named Entity Recognition

[0077] Identify specific entities in the demand, such as application type (e-commerce, social networking, etc.), resource scale (such as the number of servers, storage capacity), performance indicators (such as concurrency, response time), etc. For example, from "I need to deploy a high-performance live broadcast platform, 10 virtual machines with specifications of 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue", identify entities such as "live broadcast platform", "virtual machine", "10", "8 cores and 16G".

[0078] Step 4: Feature Extraction

[0079] Extract key features from word segmentation and named entities. You can use methods such as bag-of-words model and TF-IDF for feature representation. Convert features into feature vectors. For example, compare "I need to deploy a high-performance live broadcast platform, 10 virtual machines with 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue" with the vector vocabulary and output the vector [1,1,1,1,1,0,1,1,0,0,0,...,0,0,0,0,0,1,0,0,1].

[0080] Step 5: Input LLM model

[0081] Input the extracted feature vectors into the LLM model. Enable the model to recognize business scenarios. Perform word segmentation, prediction, probability distribution and selection on the text. For example, group the feature vectors of "[1,1,1,0,0,0,...,0,0,1,0,0,1]" and change it to resource type: virtual machine, parameters: 8 cores 16G, quantity: 10 units.

[0082] Step 6: Semantic Understanding and Intent Analysis

[0083] With the help of the LLM model, the input features are learned and inferred to understand the semantics and intent of the requirements. For example, the model can determine that the core requirements in "I need to deploy a high-performance live broadcast platform, including 10 virtual machines with 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue" are 10 8-core 16G virtual machines, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue.

[0084] Step 7: Structuring

[0085] Convert the understood requirements into a structured data format, such as JSON or XML. Include key information such as application type, resource scale, performance requirements, time limit, etc. Convert the input text information, such as "10 virtual machines with 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue", into

[0086] "{

[0087] "type": virtual machine,

[0088] "intance type":8core 16G

[0089] "num":10

[0090] "type": load balance,

[0091] "num":2

[0092] "type":Relational Database,

[0093] "intance type":32core 64G

[0094] "num":1

[0095] "type":kafka,

[0096] "num":1

[0097] }”

[0098] 2. Resource Template Conversion

[0099] Step 1: Template library creation and maintenance

[0100] Collect and organize various common resource activation templates, covering different application scenarios and demand types. Regularly update and optimize templates to adapt to new technologies and business needs.

[0101] The template library supports resource types, such as built-in virtual machine templates, cloud hard disk templates, etc., and supports high-performance cluster templates and database cluster versions based on business scenarios.

[0102] Step 2: Match requirements with templates

[0103] Compare and match the structured demand data with the templates in the template library one by one. Calculate the similarity between the structured demand and the template. If the similarity exceeds 0.9, use the template for deployment. For example, if the matching degree of "10 virtual machines with 8 cores and 16G, 2 load balancers, 1 high-performance cloud database RDS, and 1 message queue" is 0.92 with the template "10 virtual machines with 8 cores and 16G, 1 application-based load balancer, 1 cloud database with 16 cores and 32G, and 1 message queue Kafka", the template can be used for direct deployment.

[0104] Step 3: Template generation and adjustment

[0105] For requirements that cannot be directly matched, use the generation capability of LLM, refer to similar historical requirements and templates, and generate preliminary resource activation templates. Adjust and optimize the generated templates according to the specific parameters in the structured requirements. For example, adjust the configuration parameters of the server according to the actual resource scale requirements. If "10 virtual machines with specifications of 8 cores and 16G, 10 virtual machines with 4 cores and 8G, 10 cloud hard disks with 100G blocks, 2 load balancers, 2 high-performance cloud databases RDS, 1 message queue, and 2 elastic public network IPs of 5Mbps" cannot match the system template, the system will automatically generate the template "Public network high-performance business system deployment".

[0106] Step 4: Template Verification and Validation

[0107] Verify the generated or adjusted template to check its rationality and completeness. The accuracy of the template can be confirmed by simulation or comparison with expert experience.

[0108] 3. Cloud Platform Interface Call

[0109] Step 1: Interface integration and configuration

[0110] Obtain the interface documentation and development guides for different cloud platforms. Integrate and configure the interfaces as required, including setting access keys, configuring permissions, etc.

[0111] Or you can directly choose to integrate the Terraform module and call the corresponding cloud vendor's Provider to enable and configure resources.

[0112] Step 2: Send resource activation request

[0113] According to the converted resource activation template, construct the request parameters that meet the cloud platform interface requirements. Use the corresponding API call method to send the resource activation request to the cloud platform.

[0114] Step 3: Call status monitoring

[0115] Establish a real-time monitoring mechanism and regularly poll the interface to obtain the status information of the call. Handle possible abnormal situations such as request timeouts and error responses.

[0116] Step 4: Error handling and retrying

[0117] When an error occurs during an interface call, it is classified and processed according to the error code and error message. For errors that can be retried, a reasonable retry strategy and interval time are set.

[0118] 4. Calling Automated Operation and Maintenance Tools

[0119] Step 1: Tool selection and integration

[0120] Evaluate and select appropriate automated operation and maintenance tools, such as Ansible, Puppet, Chef, etc. Complete the installation and configuration of the tools to ensure compatibility with the system environment. If shell scripts need to be executed, automatically call the Ansible tool to execute for the selected resources.

[0121] Step 2: Script writing and management

[0122] Write corresponding scripts for different resource initialization requirements. Perform version control and management on the scripts to facilitate subsequent updates and maintenance. For example, after the virtual machine mounts the disk, the disk partition needs to be initialized.

[0123] Step 3: Trigger execution and monitoring

[0124] When cloud platform resources are successfully activated, the automated operation and maintenance tools are triggered in time to execute the corresponding scripts. The execution progress and results of the scripts are monitored, and the execution logs are output for viewing and analysis.

[0125] Step 4: Result feedback and adjustment

[0126] Feedback is provided to the resource template conversion module and the demand understanding module based on the results of the script execution. If the execution is not ideal, the reasons are analyzed and the subsequent processes are adjusted and optimized.

[0127] In another embodiment, the present invention proposes a cloud resource orchestration system based on LLM, comprising:

[0128] The demand understanding and analysis module is used to process the needs of business personnel using natural language processing technology, extract key information and input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured requirements;

[0129] Resource template conversion module, which is used to match and map structured requirements with pre-defined template libraries. If there is no direct match, a new template is generated by the LLM model, and parameter adjustment and optimization are performed;

[0130] The cloud platform interface calling module is used to establish interface connections with various cloud platforms, perform identity authentication and authorization, send resource activation requests based on resource activation templates, and monitor status and handle exceptions.

[0131] The automated operation and maintenance tool calling module is used to integrate common tools, trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations;

[0132] Through the collaborative work of these modules, efficient orchestration of cloud resources can be achieved.

[0133] In another embodiment of the present invention, an electronic device is proposed, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the LLM-based cloud resource orchestration method as described above is implemented.

[0134] In another embodiment, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the above-mentioned LLM-based cloud resource orchestration method.

[0135] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0136] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any technician familiar with the profession, without departing from the scope of the technical solution of the present invention, according to the technical essence of the present invention, any simple modification, equivalent replacement and improvement made to the above embodiment still falls within the protection scope of the technical solution of the present invention.

Claims

1. A cloud resource orchestration method based on the LLM model, characterized in that: The following steps are involved: Use natural language processing technology to process the needs of business personnel, extract key information and input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured needs; Match and map structured requirements with pre-defined template libraries. If they cannot be matched directly, new templates are generated by the LLM model and parameter adjustment and optimization are performed. Establish interface connections with each cloud platform, perform identity authentication and authorization, send resource activation requests based on resource activation templates, monitor status and handle exceptions; Integrate common tools to trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations.

2. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The use of natural language processing technology to process the needs of business personnel includes: Perform data preprocessing on the demand text entered by business personnel, including: cleaning, removing noise and irrelevant characters; Use appropriate word segmentation algorithms to segment the demand text into words or phrases, and perform part-of-speech tagging for each word; Identify the named entities of the requirement.

3. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The key information is extracted and then input into the LLM model to understand the semantics and intent, including: Extract key features from tokens and named entities; The extracted feature vector is input into the LLM model so that the LLM model can identify the business scenario; With the help of the LLM model, the input features are learned and reasoned to understand the semantics and intent of the requirements.

4. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The structured processing to obtain structured requirements includes: converting the understood requirements into a structured data format, wherein the data format includes key information: application type, resource scale, performance requirements, and time constraints.

5. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The structured requirements are matched and mapped with the predefined template library. If they cannot be matched directly, new templates are generated by the LLM model, and parameter adjustment and optimization are performed, including: Collect and organize various common resource activation templates, covering different application scenarios and demand types; regularly update and optimize templates to adapt to new technical and business needs; Compare and match the structured requirements with the templates in the template library one by one, calculate the similarity between the structured requirements and the templates, and deploy the templates that meet the similarity requirements; For structured requirements that do not meet the similarity requirements and cannot be directly matched, the generation capability of the LLM model is used to refer to similar historical requirements and templates to generate preliminary resource activation templates; the generated templates are adjusted and optimized according to the specific parameters in the structured requirements; Verify the generated or adjusted template to check its rationality and completeness.

6. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The establishment of interface connections with each cloud platform, identity authentication and authorization, sending resource activation requests according to resource activation templates, and monitoring status and handling exceptions include: Obtain interface documentation and development guides for different cloud platforms, and integrate and configure interfaces as required; According to the converted resource activation template, construct the request parameters that meet the cloud platform interface requirements; use the corresponding API call method to send the resource activation request to the cloud platform; Establish a real-time monitoring mechanism, regularly poll the interface to obtain call status information; handle possible request timeouts and error response exceptions; When an error occurs in an interface call, it is classified and processed according to the error code and error message; for errors that can be retried, a reasonable retry strategy and interval time are set.

7. The cloud resource orchestration method based on the LLM model according to claim 1, characterized in that: The integrated common tools described above trigger the execution of initialization script configuration tasks according to the resource provisioning results, and monitor the execution process and results to ensure that they meet expectations, including: Evaluate and select appropriate automated operation and maintenance tools, complete tool installation and configuration, and ensure compatibility with the system environment; Write corresponding scripts for different resource initialization requirements, and perform version control and management on the scripts; When cloud platform resources are successfully activated, the automated operation and maintenance tools are triggered in time to execute the corresponding scripts; the execution progress and results of the scripts are monitored, and the execution logs are output for viewing and analysis; Based on the results of script execution, feedback is given to the resource template conversion module and the demand understanding module; for unsatisfactory execution, the reasons are analyzed and the subsequent processes are adjusted and optimized.

8. A cloud resource orchestration system based on LLM, characterized in that: include: The demand understanding and analysis module is used to process the needs of business personnel using natural language processing technology, extract key information and input it into the LLM model to understand the semantics and intentions, and then perform structural processing to obtain structured requirements; Resource template conversion module, which is used to match and map structured requirements with pre-defined template libraries. If there is no direct match, a new template is generated by the LLM model, and parameter adjustment and optimization are performed; The cloud platform interface calling module is used to establish interface connections with various cloud platforms, perform identity authentication and authorization, send resource activation requests based on resource activation templates, and monitor status and handle exceptions. The automated operation and maintenance tool calling module is used to integrate common tools, trigger the execution of initialization script configuration tasks based on resource provisioning results, and monitor the execution process and results to ensure that they meet expectations; The LLM-based cloud resource orchestration system is used to execute the LLM model-based cloud resource orchestration method as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the cloud resource orchestration method based on the LLM model as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the cloud resource orchestration method based on the LLM model according to any one of claims 1 to 7.

Images