Data development implementation method and system based on large model

Abstract

The invention provides a data development implementation system and method based on a large model. The system comprises a data element recognition agent, a data source recommendation agent, a data analysis agent and a data distribution agent. The data element recognition agent is used for constructing metadata meeting business requirements according to the sample data and user requirements, and generating a metadata knowledge base of the data source; the data source recommendation agent is used for identifying business requirements of a user and recommending a data source table which meets the business requirements of the user and has the highest association degree to the user in combination with a metadata knowledge base of a data source; the data analysis agent is used for constructing an executable business SQL text meeting the business requirements of the user according to the data source table; and the data distribution agent is used for distributing execution results of the service SQL text to different knowledge bases according to different requirements. According to the method, the service requirements can be quickly identified, the SQL statements can be quickly generated, errors can be quickly detected, the content production efficiency is improved, the application scene can be expanded in combination with the multi-modal data, and the cost of the service process is reduced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for implementing data development based on a large model. Background Art

[0002] Big model technology is in a period of rapid development and change. Big model technology is iterating rapidly, and end-side big models are being implemented at an accelerated pace. It is increasingly penetrating into various industries and has strong application potential.

[0003] However, there are also certain problems behind the development of large model technology. The domestic general large models are seriously homogenized, lack original breakthroughs, and lack unique applications. The industry large models are numerous but not refined, and the actual implementation and commercialization process are slow. The reasons are: first, the original algorithm architecture innovation is insufficient; second, the high-end computing resources are in short supply; third, there is a shortage of high-quality Chinese data, the level of public data openness and sharing is insufficient, the data volume of various industries is huge, the standards are inconsistent, and the industrialization is insufficient.

[0004] In addition, although traditional data development based on big model technology has powerful capabilities in processing and analyzing large-scale data sets, there are problems with data quality, data privacy and security, and data integration. Faced with the complexity of big model technology, the high consumption of managing computing resources, and the shortage of professional skills, how to conduct real-time data analysis, conduct data governance, achieve data scalability, and improve the interpretability of analysis results, how to keep up with the rapid updates of technology and overcome the phenomenon of data silos within the organization is also a problem that needs to be solved.

[0005] In response to the above problems, this application proposes a data development implementation method and system based on a large model. Summary of the invention

[0006] The present invention proposes the following technical solutions to address one or more technical deficiencies in the above-mentioned prior art.

[0007] Based on the first aspect of the present invention, a data development implementation system based on a large model is proposed, including a data element recognition agent, a data source recommendation agent, a data analysis agent and a data distribution agent.

[0008] The data element recognition agent is used to construct metadata that meets business needs based on sample data and user needs, and generate a metadata knowledge base of the data source;

[0009] The data source recommendation agent is used to identify the business needs of users, and recommend to users the data source table that meets the business needs of users and has the highest relevance in combination with the metadata knowledge base of the data source;

[0010] The data analysis agent is used to construct an executable business SQL text that meets the user's business needs based on the data source table;

[0011] The data distribution agent is used to distribute the execution results of the business SQL text to different knowledge bases for storage according to different requirements.

[0012] Based on the second aspect of the present invention, a method for data development according to the system as described above is also proposed, comprising:

[0013] S1: The data element recognition agent combines the ID of the data source table or the metadata information of the data source table input by the user, uses the large language model for inferring Chinese fields or calls the intelligent assistant HTTP request to infer the Chinese descriptions involving the data source table and the data source table fields, and performs array conversion to obtain array data;

[0014] S2: Output the array data to a circular queue for processing, convert it into a string and output it in Json format to obtain a metadata knowledge base of the data source;

[0015] S3: The data source recommendation agent calls the keyword extraction large language model to extract keywords from the user's business needs, converts the keywords into an array format and uses the metadata knowledge base of the data source to perform iterative keyword matching to obtain the metadata of the data source table associated with the business needs, recalls and sorts the matching results according to a preset correlation, outputs the metadata of the TopN data source tables before the matching results in Json format, and obtains the data source table with the highest correlation with the user's business needs;

[0016] S4: The data analysis agent combines the ID of the data source table or the metadata information of the data source table input by the user, and calls the SQL construction big language model to generate an SQL construction plan, inputs the SQL construction plan into the big language model for identifying user intentions to obtain the business SQL text, and outputs the business SQL text in Json format.

[0017] This application can achieve data scalability, improve the interpretability of analysis results, and overcome the data silos within the organization.

[0018] Furthermore, in S1 or S4, if the user inputs the ID of the data source table, the table metadata plug-in is directly called, and if the returned result is empty, an exception is thrown, otherwise the returned result is injected into the table metadata aggregator;

[0019] If the user inputs metadata information of the data source table, the metadata is directly injected into the table metadata aggregator;

[0020] If the parameter entered by the user is empty, an exception is thrown.

[0021] Furthermore, before recalling and sorting the matching results according to the preset relevance in S2, the method also includes filtering the successfully matched results and converting them into a string format, and identifying the user's business needs using a large language model for identifying business needs.

[0022] Calling the data source recommendation agent can identify the user's business needs and recommend the data source table with the highest correlation with the user's business needs, laying the foundation for data development and expanding application scenarios.

[0023] Furthermore, outputting the array data to a circular queue for processing includes:

[0024] Combine the tagging knowledge base and use the field tagging language model to perform tagging operations;

[0025] Combine the dictionary and value range knowledge base and use the value range dictionary large language model to perform dictionary value judgment and value range judgment;

[0026] The metadata of the data source table is supplemented according to the processing result.

[0027] Furthermore, after S4, it also includes: executing the business SQL text, and the data distribution agent distributes the business result data obtained by executing the business SQL text to different knowledge bases for storage according to different needs.

[0028] Furthermore, when the keyword in the business requirement extracted from the user input is empty or when the recognition result of the user business requirement is empty, an exception is thrown.

[0029] Through different large language models, users' business needs identification can be realized without code, business efficiency can be improved, data can be processed and analyzed quickly, and real-time insights into data can be obtained.

[0030] Based on the third aspect of the present invention, a computer program product is also proposed, which includes one or more computer programs, and when the one or more computer programs are executed by a computer processor, any of the methods described above is implemented.

[0031] The technical effect of the present invention is that the present application can not only automatically infer the Chinese description of metadata, but also combine multimodal data for processing, identify the user's business needs through different large language models without code, and assist developers in SQL generation and error detection, thereby improving content production efficiency and business efficiency, expanding application scenarios, and being able to continuously adapt to new scenarios through continuous learning, providing new directions for model optimization and innovation; while reducing maintenance costs, it can also quickly process and analyze data, provide real-time insights, and reduce overall costs through automated processes, thereby promoting innovation and efficiency improvement of the entire business process. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Other features, objects and advantages of the present application will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings.

[0033] Figure 1 It is a framework diagram of a data development implementation system based on a large model provided according to an embodiment of the present invention.

[0034] Figure 2 It is an overall framework diagram of a data element recognition agent of a large model-based data development implementation system provided according to an embodiment of the present invention.

[0035] Figure 3 It is an overall framework diagram of a data source recommendation agent of a large model-based data development implementation system provided according to an embodiment of the present invention.

[0036] Figure 4 It is an overall framework diagram of a data meta-analysis agent of a large-model-based data development implementation system provided according to an embodiment of the present invention.

[0037] Figure 5 It is a flowchart of a method for implementing data development based on a large model provided according to an embodiment of the present invention.

[0038] Figure 6 It is a structural diagram of a computer system of an electronic device suitable for implementing a specific embodiment of the present application. DETAILED DESCRIPTION

[0039] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0040] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] Figure 1 The framework diagram of a data development implementation system based on a large model of the present invention is shown, including a data element recognition agent a, a data source recommendation agent b, a data analysis agent c and a data distribution agent d.

[0042] In a specific embodiment, the data element identification agent a is configured to: construct metadata that meets business needs based on sample data and user needs, and generate a metadata knowledge base of the data source.

[0043] In a specific embodiment, the data source recommendation agent b is configured to: identify the business needs of users, and recommend to users the data source table that meets the business needs of users and has the highest relevance in combination with the metadata knowledge base of the data source.

[0044] In a specific embodiment, the data analysis agent c is configured to construct an executable business SQL text that meets the user's business needs based on the data source table.

[0045] In a specific embodiment, the data distribution agent d is configured to distribute the execution results of the business SQL text to different knowledge bases for storage according to different requirements.

[0046] It should be noted that the specific method of data element recognition agent is as follows Figure 2 As shown, the following steps are included:

[0047] Step 1: The user inputs the ID of the data source table with existing metadata information through the client, or directly inputs the metadata information of the data source table;

[0048] Determine whether the parameter passed in by the user is empty. If the ID of the data source table passed in by the user or the metadata information of the data source table passed in by the user is empty, an exception is directly thrown; if the parameter passed in by the user is not empty, determine the type of the parameter passed in by the user;

[0049] Step 2: If the user inputs the ID of the data source table, the data source table metadata plug-in (including the metadata information corresponding to the ID of the data source table) is directly called and the returned result is determined to be empty;

[0050] If the returned result is empty, an exception is thrown; if the returned result is not empty, the returned result is injected into the table metadata aggregator;

[0051] Step 3: If the user inputs metadata information of a specific data source table, the metadata is injected into the table metadata aggregator;

[0052] Step 4: Infer the Chinese description of the table and table fields by calling the intelligent assistant HTTP request built based on the large language model and combining the incoming metadata information;

[0053] Alternatively, you can build a large language model for inferring Chinese fields and combine it with the incoming metadata information to infer the Chinese descriptions of the table and table fields.

[0054] Step 5: Convert the inference results obtained in the fourth step into array data and output them to a circular queue. Combine the labeling knowledge base and use the field labeling large language model to perform labeling operations. Combine the dictionary and range knowledge base and use the range dictionary large language model to perform dictionary value judgment and range judgment, and further supplement the array data with metadata reasoning.

[0055] Step 6: Convert the supplemented metadata array into a string, and then output the final result in Json format.

[0056] It should be noted that calling the data element recognition agent can realize inference and supplement of metadata information that meets business needs based on the input data source table ID or metadata of the data source table, combined with N sample data that meet business needs. Calling the API of the agent will add metadata information that meets business needs to the metadata knowledge base by creating, appending or updating.

[0057] It should be noted that the intelligent assistant can be called by other intelligent agents as a general plug-in;

[0058] The inference Chinese description large language model is an embedded large language model, which can only be used in the data element recognition agent and is the default usage mode of the fourth step.

[0059] It should be noted that the specific method of data source recommendation agent is as follows Figure 3 As shown, the following steps are included:

[0060] Step 1: The user inputs the business requirements through the client, and the keyword extraction large language model is called to extract the keywords of the business requirements;

[0061] Step 2: If the extracted business requirement keyword is empty, an exception is thrown; otherwise, the business requirement keyword is converted into an array format;

[0062] Step 3: Call the metadata knowledge base generated by the data element recognition agent to iterate keywords, and match the metadata of the data source table associated with the business needs one by one according to the keywords;

[0063] Step 4: Filter the matching results in step 3 and convert the filtered matching results into string format;

[0064] Step 5: Call the large language model that identifies business needs, identify user intent, combine keywords to obtain metadata of possible related data source tables, recall and sort matching results according to pre-set relevance, and output the top N cloud data in JSON format for recommendation.

[0065] It should be noted that calling the data source recommendation agent can identify the user's business needs and recommend the data source table that is most relevant to the user's business needs.

[0066] It should be noted that the specific method of the data analysis agent is as follows: Figure 4 As shown, the following steps are included:

[0067] Step 1: The user inputs the ID of the data source table with existing metadata information through the client, or directly inputs the metadata information of the data source table;

[0068] Determine whether the parameter passed in by the user is empty. If the ID of the data source table passed in by the user or the metadata information of the data source table passed in by the user is empty, an exception is directly thrown; if the parameter passed in by the user is not empty, determine the type of the parameter passed in by the user;

[0069] Step 2: If the user inputs the ID of the data source table, the data source table metadata plug-in (including the metadata information corresponding to the ID of the data source table) is called and it is determined whether the returned result is empty;

[0070] If the returned result is empty, an exception is thrown; if the returned result is not empty, the returned result is injected into the table metadata aggregator;

[0071] Step 3: If the user inputs metadata information of a specific data source table, the metadata is injected into the table metadata aggregator;

[0072] Step 4: Infer the Chinese description of the table and table fields by calling the intelligent assistant HTTP request built based on the large language model and combining the incoming metadata information;

[0073] Alternatively, you can build a large language model for inferring Chinese fields and combine it with the incoming metadata information to infer the Chinese descriptions of the table and table fields.

[0074] Step 5: Call SQL to build a large language model to generate an SQL construction plan, input the SQL construction plan into the large language model for identifying user intentions to obtain business SQL text, and output the business SQL text in Json format.

[0075] It should be noted that the data analysis agent provides the business SQL text to the data distribution agent through an API, and the data analysis agent distributes the execution results of the business SQL text to different knowledge bases for persistent storage according to different requirements.

[0076] Reference below Figure 5 , which shows a data development implementation method based on a large model, including the following steps:

[0077] S1: The data element recognition agent combines the ID of the data source table or the metadata information of the data source table input by the user, uses the large language model for inferring Chinese fields or calls the intelligent assistant HTTP request to infer the Chinese descriptions involving the data source table and the data source table fields, and performs array conversion to obtain array data;

[0078] S2: Output the array data to a circular queue for processing, convert it into a string and output it in Json format to obtain a metadata knowledge base of the data source;

[0079] S3: The data source recommendation agent calls the keyword extraction large language model to extract keywords from the user's business needs, converts the keywords into an array format and uses the metadata knowledge base of the data source to perform iterative keyword matching to obtain the metadata of the data source table associated with the business needs, recalls and sorts the matching results according to a preset correlation, outputs the metadata of the TopN data source tables before the matching results in Json format, and obtains the data source table with the highest correlation with the user's business needs;

[0080] S4: The data analysis agent combines the ID of the data source table or the metadata information of the data source table input by the user, and calls the SQL construction big language model to generate an SQL construction plan, inputs the SQL construction plan into the big language model for identifying user intentions to obtain the business SQL text, and outputs the business SQL text in Json format.

[0081] It should be noted that, in S1 or S4, if the user inputs the ID of the data source table, the table metadata plug-in is directly called, and if the returned result is empty, an exception is thrown, otherwise the returned result is injected into the table metadata aggregator;

[0082] If the user inputs metadata information of the data source table, the metadata is directly injected into the table metadata aggregator;

[0083] If the parameter entered by the user is empty, an exception is thrown.

[0084] It should be noted that before recalling and sorting the matching results according to the preset relevance described in S2, the method also includes filtering the successfully matched results and converting them into a string format, and identifying the user's business needs using a large language model for identifying business needs.

[0085] It should be noted that outputting the array data to the circular queue for processing includes:

[0086] Combine the tagging knowledge base and use the field tagging language model to perform tagging operations;

[0087] Combine the dictionary and value range knowledge base and use the value range dictionary large language model to perform dictionary value judgment and value range judgment;

[0088] The metadata of the data source table is supplemented according to the processing result.

[0089] It should be noted that, after S4, the step further includes: executing the business SQL text, and the data distribution agent distributes the business result data obtained by executing the business SQL text to different knowledge bases for storage according to different requirements.

[0090] It should be noted that when the keyword in the business requirement extracted from the user input is empty or when the recognition result of the user business requirement is empty, an exception is thrown.

[0091] It should be noted that data development based on the big model constructed in this application has been implemented in the national big data operation platform. It has been verified and awarded many times in the development of enterprise digitalization, rail transit, smart cities, smart policing, and provincial practical data in recent years. It can quickly and accurately identify user intentions, construct metadata that meets business needs, recommend data source tables that best meet user needs to users, build executable SQL texts that meet business needs, and provide strong support for building data distribution intelligent entities to distribute the business result data of the executed business SQL texts to the knowledge base for persistent storage according to different needs, effectively reducing the development and maintenance costs of the entire life cycle of data development, and achieving the effects of cost reduction, quality improvement, and efficiency improvement.

[0092] It should be noted that the big language model technology has a strong processing capability for unstructured data, especially in natural language understanding and generation. It can not only automatically infer the Chinese description of metadata, but also expand application scenarios by processing data through different big language models in combination with multimodal data. It can identify users' business needs without code and assist developers in SQL generation and error detection, and improve content production efficiency and business efficiency. It can continuously adapt to new scenarios through continuous learning and provide new directions for model optimization and innovation.

[0093] It should be noted that while reducing maintenance costs, this application can also quickly process and analyze data, gain real-time insights into the data, reduce overall costs through automated processes, and promote innovation and efficiency improvements in the entire business process.

[0094] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present application. Figure 6 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0095] like Figure 6 As shown, the computer system includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. Various programs and data required for system operation are also stored in the RAM 603. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0096] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0097] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above functions defined in the method of the present application are executed. It should be noted that the computer-readable storage medium of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable storage medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wireless, wireline, optical cable, RF, etc., or any suitable combination of the foregoing.

[0098] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0099] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0100] The modules involved in the embodiments of the present application may be implemented by software or by hardware.

[0101] As another aspect, the present application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiment; or it may exist independently and not be assembled into the electronic device. The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed by the electronic device, the electronic device: the data element recognition agent combines the ID of the data source table or the metadata information of the data source table input by the user, and uses the large language model of inferring Chinese fields or calling the intelligent assistant HTTP request to infer the Chinese description involving the data source table and the data source table field and perform array conversion to obtain array data; outputs the array data to a circular queue for processing, and then converts it into a string and outputs it in Json format to obtain the metadata knowledge base of the data source; the data source recommendation agent calls the large language model for extracting keywords to extract keywords in the user's business needs, and then converts the keywords into the data source recommendation information. The keywords are converted into an array format and the metadata knowledge base of the data source is used to perform iterative keyword matching to obtain metadata of the data source table associated with the business needs, and the matching results are recalled and sorted according to a preset correlation degree, and the metadata of the TopN data source tables before the matching results are output in Json format to obtain the data source table with the highest correlation with the user's business needs; the data analysis agent combines the ID of the data source table or the metadata information of the data source table input by the user, and calls the SQL construction large language model to generate an SQL construction plan, inputs the SQL construction plan into the large language model for identifying user intentions to obtain business SQL text, and outputs the business SQL text in Json format.

[0102] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0103] Finally, it should be noted that the above embodiments are only intended to illustrate rather than limit the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the present invention can still be modified or replaced by equivalents. Any modification or partial replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A data development and implementation system based on a large model, characterized in that: It includes data element recognition agent, data source recommendation agent, data analysis agent and data distribution agent; The data element recognition agent is used to construct metadata that meets business needs based on sample data and user needs, and generate a metadata knowledge base of the data source; The data source recommendation agent is used to identify the business needs of users, and recommend to users the data source table that meets the business needs of users and has the highest relevance in combination with the metadata knowledge base of the data source; The data analysis agent is used to construct an executable business SQL text that meets the user's business needs based on the data source table; The data distribution agent is used to distribute the execution results of the business SQL text to different knowledge bases for storage according to different requirements.

2. A method for data development according to the system of claim 1, characterized in that: include: S1: The data element recognition agent combines the ID of the data source table or the metadata information of the data source table input by the user, uses the large language model for inferring Chinese fields or calls the intelligent assistant HTTP request to infer the Chinese descriptions involving the data source table and the data source table fields, and performs array conversion to obtain array data; S2: Output the array data to a circular queue for processing, convert it into a string and output it in Json format to obtain a metadata knowledge base of the data source; S3: The data source recommendation agent calls the keyword extraction large language model to extract keywords from the user's business needs, converts the keywords into an array format and uses the metadata knowledge base of the data source to perform iterative keyword matching to obtain the metadata of the data source table associated with the business needs, recalls and sorts the matching results according to a pre-set correlation, outputs the metadata of the TopN data source tables before the matching results in Json format, and obtains the data source table with the highest correlation with the user's business needs; S4: The data analysis agent combines the ID of the data source table or the metadata information of the data source table input by the user, and calls the SQL construction big language model to generate an SQL construction plan, inputs the SQL construction plan into the big language model for identifying user intentions to obtain the business SQL text, and outputs the business SQL text in Json format.

3. The method according to claim 2, characterized in that In S1 or S4, if the user inputs the ID of the data source table, the table metadata plug-in is directly called, and if the returned result is empty, an exception is thrown; otherwise, the returned result is injected into the table metadata aggregator; If the user inputs metadata information of the data source table, the metadata is directly injected into the table metadata aggregator; If the parameter entered by the user is empty, an exception is thrown.

4. The method according to claim 2, characterized in that: Before recalling and sorting the matching results according to the preset relevance in S2, the method also includes filtering the successfully matched results and converting them into a string format, and identifying the user's business needs using a large language model for identifying business needs.

5. The method according to claim 2, characterized in that: Outputting the array data to a circular queue for processing includes: Combine the tagging knowledge base and use the field tagging language model to perform tagging operations; Combine the dictionary and value range knowledge base and use the value range dictionary large language model to perform dictionary value judgment and value range judgment; The metadata of the data source table is supplemented according to the processing result.

6. The method according to claim 2, characterized in that After S4, it also includes: executing the business SQL text, and the data distribution agent distributes the business result data obtained by executing the business SQL text to different knowledge bases for storage according to different needs.

7. The method according to claim 2 or 4, characterized in that: When the keyword in the business requirement extracted from the user input is empty or when the recognition result of the user business requirement is empty, an exception is thrown.

8. A computer program product comprising one or more computer programs, characterized in that: When the one or more computer programs are executed by a computer processor, the method according to any one of claims 2 to 7 is implemented.

Images