Enterprise intelligent modeling platform based on large language model

Abstract

The invention relates to an enterprise intelligent modeling platform based on a big language model, and the platform comprises a model capability platform which is used for achieving the integration with a machine learning platform and a big data platform, achieving the configuration arrangement function of a big language model / big language model agent, and providing the support of data, basic algorithm capability and big language model capability for conversational modeling; and the application capability platform is used for realizing related functions of manual modeling, dialogue-type inherent model calling, dialogue-type intelligent modeling, dialogue retrieval of existing knowledge, model sharing and agent arrangement on the basis of the model capability platform. The method can be widely applied to the technical field of enterprise-level artificial intelligence application.

Description

Technical Field

[0001] This invention belongs to the field of enterprise-level artificial intelligence application technology, specifically involving an enterprise intelligent modeling platform based on Large Language Model (LLM). Background Technology

[0002] In the process of enterprise data analysis and algorithm modeling, traditional modeling methods typically adopt a collaborative model of "business side - implementation side - technical personnel", with the specific process as follows: Requirements analysis and design phase: First, the business side puts forward its requirements, such as the audit department asking whether there are risky sales contracts (such as false trade relationships) or whether there is bid rigging or collusion. Then, the product manager (R&D department) communicates with the business side in multiple rounds to break down the requirements and form the basic requirements of data model (SQL query logic) or algorithm model (such as Python machine learning process).

[0003] Technical implementation phase: Technical personnel manually write SQL or Python code based on the requirements document. When complex logic is involved (such as multi-table joins and time series forecasting), the development cycle is long. Code debugging and optimization rely on human experience. After the model is completed, it is checked with the business side. Problems such as mismatch between business requirements and missing model conditions often occur, requiring the process of iterating on the requirements.

[0004] Model maintenance phase: When business logic changes, the entire development process must be re-executed, making rapid response impossible (including requirements analysis, design, development, testing, and deployment). Moreover, when there are many fixed models, redundant development is likely to occur.

[0005] Furthermore, enterprise data environments are complex, with knowledge scattered across multiple heterogeneous systems. Regulations and process documents (such as research management regulations and bidding management regulations) are typically stored in PDF and Word formats, which traditional keyword searches (such as Elasticsearch) cannot understand. Moreover, each second- and third-tier company within an enterprise may have its own rules, regulations, and related data, making it difficult for existing search technologies to perform searches within a specific scope. For example, searching for "methods for using research funds" requires manually reviewing multiple policy documents to determine the scope of compliance. Summary of the Invention

[0006] To address the aforementioned issues, the purpose of this invention is to provide an enterprise intelligent modeling platform based on a large language model (LLM). Through LLM and Retrieval Enhanced Generation (RAG) technology, it enables automated modeling and algorithm invocation capabilities driven by natural language (data model + algorithm model). Simultaneously, the platform will support the ability to retrieve internal enterprise knowledge through natural language dialogue.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: An enterprise intelligent modeling platform based on a large language model includes: The model capability platform is used to integrate with machine learning platforms and big data platforms, as well as to configure and orchestrate large language models / large language model agents, providing data, basic algorithm capabilities, and large language model capabilities to support conversational modeling. The application capability platform is used to realize functions such as manual modeling, conversational invocation of inherent models, conversational intelligent modeling, conversational retrieval of existing knowledge, model sharing, and intelligent agent orchestration based on the model capability platform.

[0008] Furthermore, the model capability platform includes an algorithm training and deployment unit, a large language model configuration management unit, a natural language modeling unit, and an agent management unit; The algorithm training and deployment unit is used to implement data preprocessing, deployment, establishment, training of algorithm models, and management of the basic algorithm library; The large language model configuration management unit is used to manage the large language model, prompt words, knowledge base, and MCP service. The natural language modeling unit is used to implement natural language conversion, data model training, deployment, and logical verification; The agent management unit is used to realize agent registration and management, task orchestration, and task execution status monitoring.

[0009] Furthermore, the algorithm training and deployment unit includes a basic algorithm library management module, a data preprocessing module, an algorithm training module, and an algorithm deployment module; The basic algorithm library management module is used to provide the SDK and API of the basic model for users to directly call and test. The data preprocessing module is used to preprocess the data that needs to be trained and validated. The algorithm training module is used to optimize the parameters and train the established algorithm model based on the preprocessed data. The algorithm deployment module is used to deploy the algorithm model, after parameter tuning, to the model marketplace.

[0010] Furthermore, the large language model configuration management unit includes a large language model interface management module, a knowledge base configuration module, a prompt word template engine module, and an MCP configuration module; The large language model interface management module is used to manage the access methods of large language models; The knowledge base configuration module is used to inject corporate policy documents and data dictionary knowledge into the large language model to improve domain understanding capabilities, and to use NLP to parse policy documents into semantic vectors to build the RAG knowledge base. The prompt word template engine module is used to provide prompt word templates, receive user input of large language model fine-tuning requirements, and combine with the RAG knowledge base to enable the large language model to better support modeling and the needs of using the model. The MCP configuration module is used to configure the MCP interface to enable the linkage between the large language model and other systems, and to perform related operations in other systems based on the results returned by the large language model.

[0011] Furthermore, the natural language modeling unit includes a natural language conversion module, a model training module, a model deployment module, and a logic verification module; The natural language conversion module is used to convert the natural language requirements input by the user into semantic vectors; The model training module is used to train and optimize the data model generated from semantic vectors, and adjusts the data model by providing prompt word templates and multi-turn dialogue. The logic verification module is used to perform data verification on the generated data model in order to verify the correctness of the data model; The model publishing module is used to publish validated data models to the model marketplace.

[0012] Furthermore, the intelligent agent management unit includes an intelligent agent registration and management module, a task orchestration module, and a task execution status monitoring module; The agent registration and management module is used to manage agents, and users create relevant agents through multi-turn dialogues; The task orchestration module is used to support the implementation of complex intelligent agents through a workflow approach; The task execution status monitoring module is used to monitor the current usage status of each intelligent agent.

[0013] Furthermore, the application capability platform includes a manual modeling unit, a conversational invocation of inherent models unit, a conversational intelligent modeling unit, a conversational retrieval of existing knowledge unit, a model sharing unit, and an agent orchestration unit. The manual modeling unit is used for technical support based on the model capability platform, allowing users to perform manual modeling by dragging and dropping, writing Python code, and training data. The dialogic invocation of the intrinsic model unit is used to invoke intrinsic models in the model marketplace through natural language in a multi-turn dialogue manner; The conversational automatic modeling unit is used for technical support based on the model capability platform. It generates SQL or Python code through natural language in a multi-turn dialogue manner, and allows users to intuitively debug and optimize the algorithm model or data model. The dialogue retrieval unit is used to retrieve existing knowledge in the RAG knowledge base in natural language. The model sharing unit is used to provide the ability to solidify and share models, and can share completed data models and algorithm models through the model marketplace for authorized personnel to call. The agent orchestration unit is used to orchestrate agents in the form of natural language.

[0014] Furthermore, the manual modeling unit includes a data model development module and an algorithm model development module; The data model development module is used by users to write and debug SQL statements according to actual needs, as well as perform visual modeling. The algorithm model development module is used to enable users to write, train, and deploy traditional machine learning algorithms based on the model capability platform.

[0015] Furthermore, the conversational automatic modeling unit includes a conversational data model development module and a conversational algorithm model development module; The conversational data model development module includes an SQL conversion module, a BI integration module, and a multi-turn dialogue debugging module. The SQL conversion module is used to parse the natural language requirements input by the user into SQL statements using NLP. The BI integration module is used to integrate the BI system into the platform and to display data results in a chart dashboard. The multi-turn dialogue debugging module is used to provide multi-turn dialogue functionality and to debug the generated data model according to the SQL statements until it meets the user's requirements. The conversational algorithm model development module includes an algorithm statement conversion module, a debugging tool integration module, and a training and deployment platform integration module. The algorithm statement conversion module is used to parse the natural language requirements input by the user into Python language through NLP. The debugging tool integration module is used to provide basic debugging tools. The training and deployment platform integration module is used to integrate the training and deployment functions provided by the model capability platform to provide users with training and deployment functions for machine learning algorithms.

[0016] Furthermore, the model sharing unit includes a model management module and a model marketplace module; The model management module is used to provide a model interface to realize the model sharing function, and at the same time provide model descriptions for each model; The model marketplace module provides a dashboard marketplace, an agent marketplace, an analysis report marketplace, and an algorithm marketplace, enabling the sharing of data models, algorithm models, agent models, and analysis reports.

[0017] Because of the above technical solutions, this invention has significant advantages over existing technologies in terms of development efficiency, knowledge integration, security and compliance, and explainability. 1. Model development efficiency: Regarding the demand conversion cycle, this invention automatically generates executable code through natural language interaction, reducing manual coding time. Regarding model iteration speed, this invention allows users to directly adjust model logic through dialogue (e.g., "remove internal transaction types from transactions"). In terms of cross-team collaboration, this invention eliminates distortion in the transmission of requirements and reduces communication costs; 2. Knowledge integration ability In terms of policy document retrieval, this invention accurately associates with business rules, enabling retrieval at the policy content level, while also providing the source of the content; In terms of data dictionary utilization, this invention can retrieve structured data in an information system (e.g., "which is the accounts receivable / payable table") through natural language. 3. Model usage efficiency In terms of usage, this invention can understand the model's input and output parameters, usage methods, rule meanings, and usage scenarios through natural language; In terms of sharing, this invention shares models through a marketplace, improving model reusability and reducing redundant development work.

[0018] Therefore, this invention can be widely applied in the field of enterprise-level artificial intelligence application technology. Attached Figure Description

[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. In the drawings: Figure 1 This is an architecture diagram of an enterprise intelligent modeling platform based on a large language model provided in this embodiment of the invention; Figure 2 This is a screenshot of the enterprise intelligent modeling platform based on a large language model provided in this embodiment of the invention; Figure 3 This is another screenshot of the enterprise intelligent modeling platform based on a large language model provided in this embodiment of the invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] In some embodiments of the present invention, an enterprise intelligent modeling platform based on a large language model is provided, which is used to retrieve internal enterprise data using a large language model, synchronously generate executable data models (SQL) and algorithm models (Python) through natural language interaction, and retrieve existing enterprise knowledge (structured and unstructured) through natural language.

[0023] Example 1 like Figures 1-3 As shown, this invention provides an enterprise intelligent modeling platform based on a large language model, which has functions such as dialogue invocation of inherent models, dialogue intelligent modeling, dialogue retrieval of existing knowledge, model marketplace, and intelligent agent orchestration. Specifically, it includes: The model capability platform is used to integrate with machine learning platforms and big data platforms, as well as to configure and orchestrate large language models / large language model agents, providing data, basic algorithm capabilities, and large language model capabilities to support conversational modeling. The application capability platform is used to realize functions such as manual modeling, conversational invocation of inherent models, conversational intelligent modeling, conversational retrieval of existing knowledge, model sharing, and intelligent agent orchestration based on the model capability platform.

[0024] Furthermore, the model capability platform includes an algorithm training and deployment unit, a large language model configuration and management unit, a natural language modeling unit, and an agent management unit. The algorithm training and deployment unit is used for data preprocessing, deployment, establishment, and training of algorithm models, as well as management of the basic algorithm library. The large language model configuration and management unit is used for managing the large language model, prompt words, knowledge base, and MCP service. The natural language modeling unit is used for natural language conversion, training and publishing of data models, and logical verification. The agent management unit is used for agent registration and management, task orchestration, and monitoring of task execution status.

[0025] Furthermore, the algorithm training and deployment unit primarily focuses on the development and deployment of traditional machine learning models. It includes a basic algorithm library management module, a data preprocessing module, an algorithm training module, and an algorithm deployment module. The basic algorithm library management module provides SDKs and APIs for basic models, such as SVM, decision trees, and Bayesian networks, which users can directly call and test. The data preprocessing module preprocesses the data required for training and validation. The algorithm training module optimizes the parameters and trains the established algorithm model based on the preprocessed data. The algorithm deployment module deploys the optimized algorithm model.

[0026] Furthermore, the large language model configuration management unit includes a large language model interface management module, a knowledge base configuration module, a prompt word template engine module, and an MCP configuration module. The large language model interface management module manages the access methods for the large language model, such as configuring API keys for models like Deepseek and Qwen. The knowledge base configuration module injects knowledge such as enterprise policy documents and data dictionaries into the large language model to enhance domain understanding capabilities. Policy documents (PDF / Word) are parsed into semantic vectors using NLP (emphasizing BERT+SimCSE dual encoding) to construct the RAG knowledge base. The prompt word template engine module provides prompt word templates and receives user input for fine-tuning the large language model. Combined with the RAG knowledge base, this allows the large language model to better support modeling and meet the needs of model usage. The MCP configuration module configures the MCP interface to enable interaction between the large language model and other systems, allowing related operations to be performed in other systems based on the results returned by the large language model.

[0027] Furthermore, the natural language modeling unit includes a natural language conversion module, a model training module, a model deployment module, and a logic validation module. The natural language conversion module converts user-input natural language requirements into semantic vectors; the model training module trains and optimizes the data model generated from the semantic vectors, such as by generating SQL statements or Python code, providing prompt word templates, and adjusting the data model through multi-turn dialogue; the logic validation module performs data validation on the generated data model to verify its correctness; and the model deployment module publishes the validated data model to the model marketplace.

[0028] Furthermore, the agent management unit includes an agent registration and management module, a task orchestration module, and a task execution status monitoring module. The agent registration and management module manages agents, allowing users to create related agents through multi-turn dialogues. The task orchestration module supports the implementation of complex agents through workflows. The task execution status monitoring module monitors the current usage status of each agent, such as the number of users and resources used.

[0029] Furthermore, the application capability platform includes a manual modeling unit, a conversational invocation of inherent models unit, a conversational intelligent modeling unit, a conversational retrieval of existing knowledge unit, a model sharing unit, and an agent orchestration unit. The manual modeling unit provides traditional modeling methods, allowing users to model through drag-and-drop, writing Python code, and data training. The conversational invocation of inherent models unit uses natural language to invoke inherent models from the model marketplace through multi-turn dialogues. The conversational automatic modeling unit, based on the technical support of the model capability platform, generates SQL or Python code through natural language in multi-turn dialogues. The generated code can directly generate charts (data model SQL) and algorithm implementations (algorithm model Python) through the model capability platform, allowing users to intuitively debug and optimize the algorithm or data model. The conversational retrieval of existing knowledge unit retrieves existing knowledge from the RAG knowledge base using natural language. The model sharing unit provides the ability to solidify and share models, enabling the sharing of completed data models and algorithm models through the model marketplace for authorized personnel to access. The agent orchestration unit orchestrates agents using natural language.

[0030] Furthermore, the manual modeling unit includes a data model development module and an algorithm model development module. The data model development module allows users to write and debug SQL statements according to actual needs, as well as perform visual modeling. The algorithm model development module, based on the model capability platform, allows users to write, train, and deploy traditional machine learning algorithms.

[0031] Furthermore, the conversational automatic modeling unit includes a conversational data model development module and a conversational algorithm model development module.

[0032] The conversational data model development module includes an SQL conversion module, a BI integration module, and a multi-turn dialogue debugging module. The SQL conversion module uses NLP to parse user-input natural language requirements into SQL statements; the BI integration module integrates the BI system into the platform, displaying data results in a chart-based dashboard; and the multi-turn dialogue debugging module provides multi-turn dialogue functionality, debugging the generated data model based on the SQL statements until it meets user needs.

[0033] The conversational algorithm model development module includes an algorithm statement conversion module, a debugging tool integration module, and a training and deployment platform integration module. The algorithm statement conversion module uses NLP to parse user-input natural language requirements into Python; the debugging tool integration module provides basic debugging tools, such as Jupyter; and the training and deployment platform integration module integrates the training and deployment functions provided by the model capability platform, providing users with training and deployment capabilities for machine learning algorithms.

[0034] Furthermore, the model sharing unit includes a model management module and a model marketplace module. The model management module provides model interfaces to enable model sharing and also provides model descriptions for each model. The model marketplace module provides a dashboard marketplace, an agent marketplace, an analysis report marketplace, and an algorithm marketplace to enable the sharing of data models, algorithm models, agent models, and analysis reports.

[0035] Example 2 Based on the enterprise intelligent modeling platform based on a large language model provided in Embodiment 1, this embodiment takes dialogue intelligent modeling and dialogue invocation of inherent models as examples to introduce an enterprise intelligent modeling method based on a large language model provided in this embodiment, which includes the following steps: 1) Deploy and configure an enterprise intelligent modeling platform based on a large language model and the RAG knowledge base; 2) The user's natural language input request is combined with the database table information in the RAG knowledge base for similarity matching; 3) Based on the similarity matching results and the preset prompt word fine-tuning strategy, the functions of generating dialogue data models, generating dialogue algorithm models, and calling dialogue inherent models are realized.

[0036] Furthermore, in step 1) above, the deployment of the enterprise intelligent modeling platform based on a large language model includes: Hardware layer: Hybrid deployment mode adopted Production environment: NVIDIA A100 cluster (4 nodes × 8 GPUs) supporting high concurrency. Development environment: A10G single card used for fine-tuning and testing. Software stack: Basic large language models: DeepSeek-V3 / DeepSeek-R1 Platform Integration: Spring AI Furthermore, in step 1) above, the deployment and configuration of the RAG knowledge base includes: The knowledge base is built using the vector database Milvus. The structured data used to generate the data model is pulled from the data platform on a regular basis, while the training data used to generate the algorithm model can be pulled from the unstructured data imported into the knowledge base and from the data platform.

[0037] Data platform integration process: Extract the data table structure through database / table method / RestfulAPI, synchronize the table structure, field comments, and lineage relationships on a regular basis, and store the data table structure, scenario usage information, and relationship information in the RAG knowledge base in a vectorized manner.

[0038] Unstructured data acquisition: The original files are stored in MinIO / FastDFS object storage (preserving the original PDF / Word versions), the document blocks are split according to different document types, and the data is stored in a vector database.

[0039] Furthermore, in step 3) above, the prompt word fine-tuning strategy includes: Define AI roles and output specifications to focus on calling and creating data models and algorithm models, for example: #You are a professional enterprise-level data modeling assistant, proficient in SQL statement writing and Python code writing; #Retrieve relevant information about business data from the knowledge base {xxx}; #When a user's description includes "internal transaction", use the organizational structure dimension table to specify internal transactions. Avoid using `select *` when generating SQL.

[0040] Pre-set high-frequency scene templates, such as: #If the user's problem description includes "fake transaction", use the code from the "fake transaction model" to construct the partial logic. Furthermore, in step 3) above, after the user inputs their natural language requirement, if a corresponding built-in model exists in the existing model market, it is directly invoked; otherwise, if no model meets the requirement, the required data model is created based on the prompt word fine-tuning strategy, including the following steps: The user's input of natural language requests (such as "query the accounts receivable and payable information of a subsidiary since 2020") is combined with the matching database table information to form a complete text, which is then fed into the corresponding large language model. By calling the basic large language model capabilities with pre-configured prompt words, the corresponding SQL statement is generated, and the BI service is called using the MCP protocol to generate corresponding charts from the SQL for intuitive debugging of the SQL model. If there are problems with the model or the data is inaccurate, users can further refine and adjust the model through rounds of dialogue in the same session, such as "accounts receivable and payable information should exclude internal transactions of the company".

[0041] Furthermore, in step 3) above, after the user inputs their natural language requirement, if there is no model in the existing model market that meets this requirement, the required algorithm model is created based on the prompt word fine-tuning strategy, including the following steps: Obtain database table information related to the user's natural language input (e.g., "Are these suppliers involved in bid rigging or collusion?") as data for model training; By leveraging the capabilities of large language models, basic Python training code is generated, and the most suitable data source is provided as training data. After the user confirms the training code and data through dialogue, the system calls the underlying machine learning platform capabilities through the MCP protocol to deploy the corresponding training code and extract the corresponding training data from the data platform for training. After training is completed, the system provides the Python code for evaluation and prompts the user to evaluate the accuracy of the model through dialogue.

[0042] Furthermore, after the user inputs their natural language request, the system invokes the corresponding algorithm or data model through multi-turn dialogue, including: The system retrieves data from the RAG knowledge base through a dialogue mechanism and provides a list of models relevant to the user's needs for the user to choose from. Users select models through dialogue and provide the required parameters. The system uses its large language model capabilities to parse the parameters in the user's dialogue and, in conjunction with prompt word templates, calls related models from the model marketplace.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. An enterprise intelligent modeling platform based on a large language model, characterized in that, include: The model capability platform is used to integrate with machine learning platforms and big data platforms, as well as to configure and orchestrate large language models / large language model agents, providing data, basic algorithm capabilities, and large language model capabilities to support conversational modeling. The application capability platform is used to realize functions such as manual modeling, conversational invocation of inherent models, conversational intelligent modeling, conversational retrieval of existing knowledge, model sharing, and intelligent agent orchestration based on the model capability platform.

2. The enterprise intelligent modeling platform based on a large language model as described in claim 1, characterized in that, The model capability platform includes an algorithm training and deployment unit, a large language model configuration and management unit, a natural language modeling unit, and an agent management unit; The algorithm training and deployment unit is used to implement data preprocessing, deployment, establishment, training of algorithm models, and management of the basic algorithm library; The large language model configuration management unit is used to manage the large language model, prompt words, knowledge base, and MCP service. The natural language modeling unit is used to implement natural language conversion, data model training, deployment, and logical verification; The agent management unit is used to realize agent registration and management, task orchestration, and task execution status monitoring.

3. The enterprise intelligent modeling platform based on a large language model as described in claim 2, characterized in that, The algorithm training and deployment unit includes a basic algorithm library management module, a data preprocessing module, an algorithm training module, and an algorithm deployment module; The basic algorithm library management module is used to provide the SDK and API of the basic model for users to directly call and test. The data preprocessing module is used to preprocess the data that needs to be trained and validated. The algorithm training module is used to optimize the parameters and train the established algorithm model based on the preprocessed data. The algorithm deployment module is used to deploy the algorithm model, after parameter tuning, to the model marketplace.

4. The enterprise intelligent modeling platform based on a large language model as described in claim 2, characterized in that, The large language model configuration management unit includes a large language model interface management module, a knowledge base configuration module, a prompt word template engine module, and an MCP configuration module. The large language model interface management module is used to manage the access methods of large language models; The knowledge base configuration module is used to inject corporate policy documents and data dictionary knowledge into the large language model to improve domain understanding capabilities, and to use NLP to parse policy documents into semantic vectors to build the RAG knowledge base. The prompt word template engine module is used to provide prompt word templates, receive user input of large language model fine-tuning requirements, and combine with the RAG knowledge base to enable the large language model to better support modeling and the needs of using the model. The MCP configuration module is used to configure the MCP interface to enable the linkage between the large language model and other systems, and to perform related operations in other systems based on the results returned by the large language model.

5. The enterprise intelligent modeling platform based on a large language model as described in claim 2, characterized in that, The natural language modeling unit includes a natural language conversion module, a model training module, a model deployment module, and a logic verification module; The natural language conversion module is used to convert the natural language requirements input by the user into semantic vectors; The model training module is used to train and optimize the data model generated from semantic vectors, and adjusts the data model by providing prompt word templates and multi-turn dialogue. The logic verification module is used to perform data verification on the generated data model in order to verify the correctness of the data model; The model publishing module is used to publish validated data models to the model marketplace.

6. The enterprise intelligent modeling platform based on a large language model as described in claim 2, characterized in that, The agent management unit includes an agent registration and management module, a task orchestration module, and a task execution status monitoring module; The agent registration and management module is used to manage agents, and users create relevant agents through multi-turn dialogues; The task orchestration module is used to support the implementation of complex intelligent agents through a workflow approach; The task execution status monitoring module is used to monitor the current usage status of each intelligent agent.

7. The enterprise intelligent modeling platform based on a large language model as described in claim 1, characterized in that, The application capability platform includes a manual modeling unit, a dialogic invocation of inherent models unit, a dialogic intelligent modeling unit, a dialogic retrieval of existing knowledge unit, a model sharing unit, and an intelligent agent orchestration unit. The manual modeling unit is used for technical support based on the model capability platform, allowing users to perform manual modeling by dragging and dropping, writing Python code, and training data. The dialogic invocation of the intrinsic model unit is used to invoke intrinsic models in the model marketplace through natural language in a multi-turn dialogue manner; The conversational automatic modeling unit is used for technical support based on the model capability platform. It generates SQL or Python code through natural language in a multi-turn dialogue manner, and allows users to intuitively debug and optimize the algorithm model or data model. The dialogue retrieval unit is used to retrieve existing knowledge in the RAG knowledge base in natural language. The model sharing unit is used to provide the ability to solidify and share models, and can share completed data models and algorithm models through the model marketplace for authorized personnel to call. The agent orchestration unit is used to orchestrate agents in the form of natural language.

8. The enterprise intelligent modeling platform based on a large language model as described in claim 7, characterized in that, The manual modeling unit includes a data model development module and an algorithm model development module; The data model development module is used by users to write and debug SQL statements according to actual needs, as well as perform visual modeling. The algorithm model development module is used to enable users to write, train, and deploy traditional machine learning algorithms based on the model capability platform.

9. The enterprise intelligent modeling platform based on a large language model as described in claim 7, characterized in that, The conversational automatic modeling unit includes a conversational data model development module and a conversational algorithm model development module; The conversational data model development module includes an SQL conversion module, a BI integration module, and a multi-turn dialogue debugging module. The SQL conversion module is used to parse the natural language requirements input by the user into SQL statements using NLP. The BI integration module is used to integrate the BI system into the platform and to display data results in a chart dashboard. The multi-turn dialogue debugging module is used to provide multi-turn dialogue functionality and to debug the generated data model according to the SQL statements until it meets the user's requirements. The conversational algorithm model development module includes an algorithm statement conversion module, a debugging tool integration module, and a training and deployment platform integration module. The algorithm statement conversion module is used to parse the natural language requirements input by the user into Python language through NLP; the debugging tool integration module is used to provide basic debugging tools; the training and deployment platform integration module is used to integrate the training and deployment functions provided by the model capability platform, providing users with training and deployment functions for machine learning algorithms.

10. The enterprise intelligent modeling platform based on a large language model as described in claim 7, characterized in that, The model sharing unit includes a model management module and a model marketplace module; The model management module is used to provide a model interface to realize the model sharing function, and at the same time provide model descriptions for each model; The model marketplace module provides a dashboard marketplace, an agent marketplace, an analysis report marketplace, and an algorithm marketplace, enabling the sharing of data models, algorithm models, agent models, and analysis reports.

Images