Intelligent simulation evaluation system and method of mixed knowledge retrieval enhancement and model context protocol

Abstract

The invention discloses an intelligent simulation evaluation system and method for mixing knowledge retrieval enhancement and a model context protocol, and belongs to the technical field of simulation evaluation. The intelligent simulation evaluation system comprises a view layer, a service layer and a data layer; the view layer is provided with a natural language dialogue component used for providing an evaluation demand natural language text input interface and an evaluation result display interface for a user; the knowledge base management component is used for supporting a user to construct an evaluation priori knowledge text according to data source information and evaluation service requirements, and submitting the text to the RAG workflow system for retrieval and query; the data layer is provided with a relational simulation database, a time sequence simulation database, a non-relational simulation database and other system APIs. According to the method, reliable simulation evaluation and analysis results can be formed, evaluation users do not need to compile index codes in advance, evaluation requirements can be described only through natural languages, and related evaluation indexes and corresponding analysis results of any service can be generated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of simulation evaluation, more particularly, to an intelligent simulation evaluation system and method based on mixed knowledge retrieval enhancement and model context protocol. BACKGROUND

[0002] In the field of modern simulation evaluation, the application mode is complex, the evaluation elements are numerous and the correlation is complex, which leads to the difficulty in constructing the index system. At the same time, the index needs to be adjusted and optimized in the evaluation process to adapt to the changes of the environment and tasks. Therefore, the traditional static evaluation method cannot meet the needs, which is specifically embodied in: 1) construction of complex evaluation system: the traditional evaluation system needs to manually construct complex evaluation indexes, which requires professional knowledge and a lot of time; 2) manual coding and data source interaction: the existing method needs to manually write code to interact with the data source, which not only increases the development cost, but also reduces the flexibility and maintainability of the system; 3) limitations of natural language processing: the existing system often cannot accurately understand the user's intention when processing natural language queries, which leads to inaccurate query results; 4) evaluation personnel need to participate in evaluation coding: the current friendly evaluation system provides a graphical query interface for users to quickly construct index query algorithms, but users still need to understand the data query coding principle, which makes it difficult for non-coding professional users to participate in the construction of the evaluation system. SUMMARY

[0003] The purpose of the present application is to overcome the shortcomings of the prior art and provide an intelligent simulation evaluation system and method based on mixed knowledge retrieval enhancement and model context protocol. By introducing the model context protocol, the system provides standardized interaction capabilities for large models with simulation data sources or existing system APIs. Combined with knowledge retrieval enhancement technology, the large model can accurately understand the evaluation business and the capabilities of the tools used, enabling the large model to accurately understand user evaluation needs, correctly use tools, and query correct data. Ultimately, reliable simulation evaluation and analysis results are formed, and evaluation users do not need to pre-write index code. They only need to describe their evaluation needs in natural language, and the system can generate any business-related evaluation indexes and corresponding analysis results.

[0004] The purpose of the present application is achieved by the following scheme: An intelligent simulation evaluation system based on mixed knowledge retrieval enhancement and model context protocol, comprising a view layer, a service layer and a data layer; In the view layer, a natural language dialogue component is provided to provide a natural language text input interface for user evaluation needs and an evaluation result display interface; a knowledge base management component is used to support users to construct evaluation prior knowledge text according to data source information and evaluation business needs, and submit the text to the RAG workflow system for retrieval query; The MCP Client module and the MCP Server module are arranged in the service layer; the MCP Client module specifically comprises: an evaluation tool optimal large model service module, configured to select the most suitable tool from a list of available tools based on a user-input evaluation task natural language text using a large model; an evaluation result generation large model service module, configured to combine query result data with the user-input evaluation task natural language text, analyze and summarize the data using a large model, and form an evaluation result; a natural language processing service module, configured to extract keywords from the user-input evaluation task natural language text and optimize query requirement description, and generate evaluation requirement keywords; an RAG workflow interaction service module, configured to call an RAG workflow system using the optimized evaluation task natural language text, obtain a workflow inference result, and form a search context related to the evaluation business; and an MCP Server management service module, configured to be responsible for registration and connection maintenance of the MCP Server, and for obtaining a list of available tools in the MCP Server and calling tools in the corresponding MCP Server according to the tools selected by the large model, to realize interaction with a simulation data source; the MCP Server module specifically comprises: a relational database interaction service module, configured to execute a query operation of a relational database; a time-series database interaction service module, configured to execute a query operation of a time-series database; a non-relational database interaction service module, configured to execute a query operation of a non-relational database; and an existing system API interaction service module, configured to call an API of an existing system to obtain data. The data layer is provided with: a relational simulation database, configured to store and manage structured data; a time-series simulation database, configured to store and manage time-series data; a non-relational simulation database, configured to store and manage unstructured data; and other system API interfaces, configured to call APIs of other systems to obtain or send data.

[0005] Further, the view layer composition is further provided with: an index system evaluation component, configured to execute evaluation calculation and display evaluation results by calling a pre-encapsulated deterministic index system API interface.

[0006] Further, the view layer composition is further provided with: an evaluation report generation component, configured to receive evaluation results, call an evaluation report generation service, generate a report, and display the report.

[0007] Further, the service layer is further provided with: an MCP Host module, configured to interact with the view layer, create and manage a plurality of MCP Client services and MCP Server services.

[0008] Further, the MCP Client module further comprises an index system evaluation service module compatible with existing evaluation index calculation code for constructing a customized evaluation service for deterministic evaluation indexes.

[0009] Further, the MCP Server module specifically further comprises a data access permission management service module for controlling user access to data to ensure data security.

[0010] An intelligent evaluation method based on the MCP model context protocol, based on the hybrid knowledge retrieval enhancement and model context protocol intelligent simulation evaluation system as described above, further comprising the following steps: Step (one), register and listen to various services in the MCP Server through the MCP Server management service module; Step (two), the MCP Server management service module obtains a list of available tools from the registered MCP Server; Step (three), the evaluation tool optimization large model service module queries the list of available tools through the MCP Server management service module, and generates tool optimization and parameters for the large model in combination with the user input evaluation requirement natural language text; Step (four), initiate tool calling through the MCP Server management service module, send the request to the corresponding MCP Server through the MCP protocol, and call the tool; Step (five), the corresponding database or existing system API interaction service in the MCP Server responds to the tool calling and executes the tool parameters; Step (six), the data source executes the query statement or API calling and returns the query or calling result; Step (seven), after the MCP Server receives the query result returned by the data source, the result is encapsulated and processed according to the requirements of the MCP protocol, and is returned to the MCP Server management service module; Step (eight), the MCP Server management service module sends the query result to the evaluation result generation large model service module, which utilizes the returned query result and the evaluation requirement natural language text to summarize and analyze, forming the evaluation result.

[0011] A knowledge pre-retrieval method based on a knowledge retrieval enhancement RAG workflow, based on the hybrid knowledge retrieval enhancement and model context protocol intelligent simulation evaluation system as described above, further comprising the following steps: Step (one), knowledge vectorization process: Substep (1), the user submits the prior knowledge text through the knowledge base management component of the view layer in the interface; Sub-step (2), the submitted priori knowledge text is chunked and vectorized by the knowledge retrieval service of the RAG workflow system and stored in the vector database; Step (two), search enhanced process: Sub-step (1), the user inputs the natural language text of the query to be evaluated through the natural language evaluation component interface interaction dialog box of the view layer; Sub-step (2), the natural language processing service module in the service layer MCP Client summarizes and abstracts the natural language text input by the user through a large model to generate evaluation requirement keywords; Sub-step (3), the RAG workflow interaction service module in the service layer MCP Client takes the evaluation requirement keywords as parameters and requests the RAG workflow system service; Sub-step (4), the knowledge retrieval service module in the RAG workflow system vectorizes the evaluation requirement keywords and calculates the correlation with the vectors stored in the vector database to select the top N vectors and generate a context; Sub-step (5), the large model workflow service module in the RAG workflow system generates recommended tool types and suggested parameters in combination with the generated context and the evaluation requirement keywords input by the user, and returns the search context to the RAG workflow interaction service, wherein the tool type is the selected data source and the suggested parameter is the query statement or API call request format; Sub-step (6), the RAG workflow interaction service module in the service layer MCP Client receives the returned search context for reference by the subsequent evaluation tool optimization large model service.

[0012] A natural language evaluation method combining MCP and RAG, based on the intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol as described above, further comprising the following steps: Step (one), the user inputs the evaluation requirement natural language text through the natural language evaluation component of the view layer; Step (two), the natural language processing service module of the service layer MCP Client stores the evaluation requirement natural language text to the historical query list for multi-round dialogue and summarizes and abstracts it through a large model to generate evaluation requirement keywords; Step (three), the RAG workflow interaction service module calls the RAG workflow system to perform knowledge retrieval and large model reasoning using the evaluation requirement keywords to obtain priori vector information related to data sources and business logic and generate a search context for subsequent tool selection reference; Step (four), the MCP Server management service module queries and returns the tool list provided by the MCPServer registered in the MCP Client; Step (five), the evaluation tool preferably a large model service module to evaluate the demand keywords, RAG workflow generated search context and registered tool list as input, the large model generates the tool to be called and the corresponding tool parameters; Step (six), the various databases or existing system API interaction services of the service layer MCP Server according to the request, judge the database type; if the relational database query tool is called, connect the database and execute the generated SQL statement; if the non-relational database query tool is called, connect the database and execute the generated NoSQL statement; if the time series database query tool is called, connect the database and execute the generated InfluxQL statement; if the existing system API tool is called, connect the existing system server and execute the API call request; finally, the query statement execution result or API call result is returned to the evaluation result generation large model service in MCPClient; Step (seven), the evaluation result generation large model service module, the returned query result, combined with the user input evaluation demand natural language text and search context information, input to the large model for data analysis and summary, form the evaluation result; Step (eight), the natural language evaluation component of the view layer displays the evaluation result in the interface interaction dialog box, and the process ends. Step (nine), if multiple rounds of dialogue are needed, steps (one) to (eight) are repeated.

[0013] Further, in step (six), the data source includes a relational simulation database, a non-relational simulation database, a time series simulation database, and other system API services.

[0014] The beneficial effects of the present application include: (1) The present application allows users to query through natural language, and the system can accurately understand the user's intention and return the relevant evaluation result, improving the user experience and the usability of the system.

[0015] (2) The present application provides the ability to interact directly with simulation data for large models through the MCP model context protocol, can understand user evaluation requirements, and multi-dimensional analysis of evaluation results, improve the utilization of simulation data.

[0016] (3) The present application can more accurately retrieve and utilize relevant knowledge through knowledge retrieval enhancement technology, thereby improving the efficiency and accuracy of evaluation. This helps to reduce human errors in the evaluation process and improve the reliability of the evaluation result.

[0017] (4) The application makes the system have higher flexibility and scalability. Users can conveniently add or modify evaluation requirements according to actual requirements, adapt to different evaluation scenes, reduce the workload of manual coding and data source interaction, and reduce the development and maintenance cost of the system. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0019] Figure 1 The architecture diagram of the intelligent simulation evaluation system of the mixed knowledge retrieval enhancement and model context protocol for the embodiment of the present application; Figure 2 The running principle diagram of the MCP model context protocol in the intelligent evaluation architecture in the embodiment of the present application; Figure 3 The knowledge retrieval enhancement RAG workflow running principle diagram in the embodiment of the present application; Figure 4 The natural language evaluation principle diagram combined with MCP and RAG in the embodiment of the present application. DETAILED DESCRIPTION

[0020] All features disclosed in all embodiments of the present specification, or all steps in the methods or processes impliedly disclosed, can be combined and / or extended, replaced, unless mutually exclusive features and / or steps.

[0021] In view of the problems in the background, the inventors of the present application believe that: in recent years, with the rapid development of large model technology, the model context protocol MCP (Model Context Protocol) has emerged as the times require. The protocol is an open source protocol that can unify the communication protocol between large language models (LLM) and external data sources and tools. The core of MCP is to take "context" as an independent and standardized data transmission object, and provide a unified interface. Through the MCP protocol, different modules can work together to complete more complex tasks. However, simply through the interaction between the large model and the external data source, the large model often cannot understand the actual business scenario, resulting in incorrect use of tools or inability to correctly interact with the data source. Through retrieval augmented generation technology (RAG), the prior text knowledge of simulated scenario, database table name and field meaning, and existing system API call rules can be fully utilized. Through intelligent retrieval and generation mechanism, correct query and call of data source and existing system are realized.

[0022] Specifically, as a first aspect of the present application, refer to Figure 1 (Note that, Figure 1 The gray filled part in the middle can be directly realized by existing technology, and can participate in the overall architecture operation of the present application). The embodiment of the present application specifically designs an intelligent simulation evaluation system of mixed knowledge retrieval enhancement and model context protocol, and the architecture follows a hierarchical design mode and is divided into a view layer, a service layer and a model layer. In the view layer, unlike traditional evaluation systems, a natural language evaluation component is added to provide a highly free natural language evaluation demand interaction interface for evaluation users, and the knowledge base management component process is optimized. In the service layer, based on the MCP model context protocol, an extensible interactive query tool for evaluating large models with data sources or existing systems is provided to realize evaluation tool optimization, evaluation result generation and other services. In the model layer, various databases, other system API interfaces and the like used in the simulation process are directly accessed for query and call by the service layer large model. Specifically, the following contents are included: (I) The view layer is composed of: (1) A natural language dialogue component provides a natural language text input interface for evaluation requirements and an evaluation result display interface for users; (2) An index system evaluation component executes evaluation calculation and displays evaluation results by calling a pre-encapsulated deterministic index system API interface; (3) An evaluation report generation component receives evaluation results, calls an evaluation report generation service, generates a report and displays it; (4) Knowledge base management component, which supports users to construct evaluation priori knowledge text according to data source information and evaluation business requirements, and submits the text to the RAG workflow system for retrieval query.

[0023] (II) The service layer follows the MCP model context architecture, and the specific principle is shown in Figure 2 , and the specific service layer components are: (1) MCP Host is responsible for interacting with the view layer, creating and managing multiple MCP Client services and MCP Server services; (2) The specific components of MCP Client are: ① Evaluation tool optimization large model service, based on the evaluation task natural language text input by the user, uses the large model to select the most suitable tool from the available tool list; ② Evaluation result generation large model service, combines the query result data with the evaluation task natural language text input by the user, analyzes and summarizes the data through the large model to form the evaluation result; ③ Natural language processing service, performs keyword extraction and query requirement description optimization on the evaluation task natural language text input by the user to generate evaluation requirement keywords; ④ Index system evaluation service, compatible with existing evaluation index calculation code, constructs customized evaluation service for deterministic evaluation indexes; ⑤ RAG workflow interaction service, calls the RAG workflow system with the optimized evaluation task natural language text to obtain the workflow reasoning result to form a retrieval context related to the evaluation business, and the specific principle is shown in Figure 3 ; ⑥ MCP Server management service, which is responsible for the registration and connection maintenance of MCP Server, obtains the available tool list in MCP Server, and calls the tools in the corresponding MCP Server according to the tools selected by the large model to realize the interaction with the simulation data source.

[0024] (3) The specific components of MCP Server are: ① Relational database interaction service, which performs query operations on relational databases; ② Time series database interaction service, which performs query operations on time series databases; ③ Non-relational database interaction service, which performs query operations on non-relational databases; ④ Existing system API interaction service, which calls the API of the existing system to obtain data; ⑤ Data access permission management service, which controls the access permission of users to data to ensure data security.

[0025] (III) The data layer includes: (1) Relational simulation database, store relational data such as simulation records, and store and manage structured data such as simulation records; (2) Time series simulation database, store time series data such as simulation records, and store and manage time series data such as simulation records; (3) Non-relational simulation database, store non-relational data such as simulation records, and store and manage unstructured data such as simulation records; (4) Other system API interface, call the API of other systems to obtain or send data.

[0026] It should be noted that within the scope of protection defined in the claims of the present application, the following embodiments can be combined and / or extended, replaced, in any logical manner from the above specific embodiments, such as disclosed technical principles, disclosed technical features or implied disclosed technical features.

[0027] As a second aspect of the present application, refer to Figure 2 Based on the above-mentioned example system, the embodiment of the present application specifically designs an intelligent evaluation method based on the MCP model context protocol, so that the large model can interact with the simulation data source by using the tool, and analyze and summarize. Through the standardized interaction process design of the MCP protocol, the scalability of the large model using tool services is enhanced. Specifically, the running steps are as follows: (I), the evaluation MCP Client first registers to listen to various types of MCP Server through the MCP Server management service, and the MCP Server type is various types of services in the MCP Server; (II), the MCP Server management service obtains the available tool list such as data source query and API call from the registered MCP Server; (III), the evaluation tool in the MCP Client optimizes the large model service, queries the available tool list through the MCP Server management service, and generates tool optimization and parameters for the large model in combination with the user input evaluation requirement natural language text; (IV), initiate tool calling through the MCP Server management service, send the request to the corresponding MCP Server through the MCP protocol, and call the tool; (V), the corresponding database or existing system API interaction service in the MCP Server responds to the tool calling and executes the tool parameters, wherein the tool parameters are query statements or API calling requests; (Six), relational simulation database, non-relational simulation database, time series simulation database and other system API services and other data sources, execute query statements or API calls, and return query or call results; (Seven), after the MCP Server receives the query results returned by the data source, the results are encapsulated and processed according to the requirements of the MCP protocol, and returned to the MCP Server management service of the evaluation MCP Client; (Eight) The MCP Server management service sends the query results to the evaluation result generation large model service, which uses the returned query results and evaluation requirement natural language text to summarize and analyze, forming the evaluation result.

[0028] Through the MCP protocol, the MCP Server can be flexibly registered, and the scalability of the large model using the tools provided by the MCP Server can be enhanced.

[0029] As a third aspect of the present application, refer to Figure 3 In order for the large model to accurately understand user requirements and correctly use tools, the embodiment of the present application specifically designs a knowledge pre-retrieval method based on knowledge retrieval enhanced RAG workflow, which uses the RAG workflow system to assist in enhancing the understanding ability of the large model to the evaluation business. Specifically, the running steps are as follows: (One), knowledge vectorization process (1) The user submits data source information, evaluation business information and other prior knowledge text through the knowledge base management component of the view layer interface; (2) The submitted prior knowledge text is blocked and vectorized by the knowledge retrieval service of the RAG workflow system, and stored in the vector database.

[0030] (Two), retrieval enhancement process (1) The user inputs the natural language text that wants to evaluate through the natural language evaluation component interface interactive dialog box of the view layer; (2) The natural language processing service in the service layer MCP Client summarizes and refines the natural language text input by the user through the large model to generate evaluation requirement keywords; (3) The RAG workflow interaction service in the service layer MCP Client takes the evaluation requirement keywords as parameters and requests the RAG workflow system service; (4) The knowledge retrieval service in the RAG workflow system vectorizes the evaluation requirement keywords and calculates the correlation with the vectors stored in the vector database to select the top N vectors and generate the context; (5) The large model workflow service in the RAG workflow system generates recommended tool types and suggested parameters in combination with the generated context and the evaluation requirement keywords input by the user, and returns the search context to the RAG workflow interactive service as the tool types are selected data sources and the suggested parameters are query statements or API call request formats; (6) The RAG workflow interactive service in the service layer MCP Client receives the returned search context for reference by the subsequent evaluation tool optimization large model service.

[0031] As a fourth aspect of the present application, refer to Figure 4 The embodiment of the present application specifically designs a natural language evaluation method combining MCP and RAG, which relies on the above-mentioned intelligent simulation evaluation architecture of mixed knowledge retrieval enhancement and model context protocol, realizes evaluation result generation and evaluation result analysis by only natural language description of evaluation requirements by the user. Specifically, the running steps are as follows: (I) The user inputs the evaluation requirement natural language text through the natural language evaluation component of the view layer.

[0032] (II) The natural language processing service of the service layer MCP Client stores the evaluation requirement natural language text to the historical query list for multi-round dialogue, and generates evaluation requirement keywords through summarization and refinement by the large model.

[0033] (III) The RAG workflow interactive service calls the RAG workflow system to perform knowledge retrieval and large model reasoning by using the evaluation requirement keywords, acquires prior vector information related to data sources and business logic, and generates search context for subsequent tool selection reference.

[0034] (IV) The MCP Server management service queries and returns the tool list provided by the MCP Server registered in the MCP Client.

[0035] (V) The evaluation tool optimization large model service takes the evaluation requirement keywords, the search context generated by the RAG workflow, and the registered tool list as inputs, and generates the tool to be called and the corresponding tool parameters by the large model.

[0036] (Six), the various databases of the service layer MCP Server or the API interaction service of the existing system service judges the database type according to the request. If the relational database query tool is called, the database is connected and the generated SQL statement is executed; if the non-relational database query tool is called, the database is connected and the generated NoSQL statement is executed; if the time series database query tool is called, the database is connected and the generated InfluxQL statement is executed; if the existing system API tool is called, the existing system server is connected and the API call request is executed. Finally, the query statement execution result or the API call result is returned to the evaluation result generation large model service in the MCP Client.

[0037] (Seven), the evaluation result generation large model service inputs the returned query result, the evaluation requirement natural language text input by the user and the retrieval context information into the large model to perform data analysis and summary, and forms an evaluation result.

[0038] (Eight), the natural language evaluation component of the view layer displays the evaluation result in the interface interaction dialog box, and the process ends.

[0039] (Nine), if multiple rounds of dialogues are required, steps (one) to (eight) are repeatedly executed.

[0040] The units described in the embodiments of the present application can be implemented in the form of software or hardware, and the described units can also be arranged in a processor. In some cases, the names of these units do not constitute a limitation on the units themselves.

[0041] According to an aspect of an embodiment of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the method provided in the various optional implementation manners described above.

[0042] As another aspect, the embodiment of the present application also provides a computer readable medium, which can be included in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The computer readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device implements the method described in the above embodiments.

Claims

1. An intelligent simulation evaluation system that combines knowledge retrieval enhancement and model context protocol, characterized by: Includes view layer, service layer and data layer; The view layer includes: a natural language dialogue component that provides users with a natural language text input interface for evaluation requirements and an evaluation result display interface; a knowledge base management component that supports users in constructing evaluation prior knowledge text based on data source information and evaluation business requirements, and submitting the text to the RAG workflow system for retrieval and query; The service layer is equipped with an MCP Client module and an MCP Server module; wherein the MCP Client module is specifically equipped with: an evaluation tool optimization large model service module, which is used to select the most suitable tool from the list of available tools based on the natural language text of the evaluation task input by the user using the large model; an evaluation result generation large model service module, which is used to combine the query result data with the natural language text of the evaluation task input by the user, analyze and summarize the data through the large model to form an evaluation result; a natural language processing service module, which is used to extract keywords and optimize the query requirement description of the natural language text of the evaluation task input by the user, and generate evaluation requirement keywords; a RAG workflow interaction service module, which is used to call the RAG workflow system with the optimized evaluation task natural language text to obtain the workflow reasoning results to form a retrieval context related to the evaluation business; an MCP Server management service module, which is responsible for the registration and connection maintenance of the MCP Server, as well as obtaining the list of available tools in the MCP Server and calling the corresponding tools in the MCP Server according to the tools selected by the large model to realize interaction with the simulation data source; wherein, the MCP The server module specifically includes: a relational database interaction service module for executing query operations on relational databases; a time series database interaction service module for executing query operations on time series databases; a non-relational database interaction service module for executing query operations on non-relational databases; and an existing system API interaction service module for calling the API of an existing system to obtain data. The data layer is equipped with: a relational simulation database for storing and managing structured data; a time series simulation database for storing and managing time series data; a non-relational simulation database for storing and managing unstructured data; and other system API interfaces for calling APIs of other systems to obtain or send data.

2. The intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol according to claim 1 is characterized in that: The view layer also includes: an indicator system evaluation component, which is used to perform evaluation calculations and display evaluation results by calling a pre-packaged deterministic indicator system API interface.

3. The intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol according to claim 1 is characterized in that: The view layer also includes: an assessment report generation component, which is used to receive assessment results, call the assessment report generation service, generate a report and display it.

4. The intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol according to claim 1 is characterized in that: The service layer also has the MCP Host module, which is used to interact with the view layer and create and manage multiple MCPClient services and MCP Server services.

5. The intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol according to claim 1 is characterized in that: The MCP Client module also includes: an indicator system evaluation service module, which is compatible with existing evaluation indicator calculation codes and is used to build customized evaluation services for deterministic evaluation indicators.

6. The intelligent simulation evaluation system of hybrid knowledge retrieval enhancement and model context protocol according to claim 1 is characterized in that: The MCP Server module specifically also includes: a data access permission management service module, which is used to control user access permissions to data and ensure data security.

7. An intelligent evaluation method based on MCP model context protocol, characterized in that: The intelligent simulation evaluation system based on the hybrid knowledge retrieval enhancement and model context protocol according to claim 1 further includes the following steps: Step 1: Register and monitor various services in the MCP Server through the MCP Server management service module; Step (2), the MCP Server management service module obtains a list of available tools from the registered MCP Server; Step (3): The evaluation tool optimization large model service module queries the list of available tools through the MCP Server management service module, and combines the user-entered evaluation requirements natural language text to optimize the tool and generate parameters for the large model; Step 4: Initiate a tool call through the MCP Server management service module, send the request to the corresponding MCP Server through the MCP protocol, and call the tool; Step 5: The corresponding database or existing system API interaction service in the MCP Server responds to the tool call and executes the tool parameters; Step (6): The data source executes the query statement or API call and returns the query or call result; Step (7): After receiving the query results returned by the data source, the MCP Server encapsulates and processes the results according to the requirements of the MCP protocol and returns them to the MCP Server management service module; Step (8): The MCP Server management service module sends the query results to the evaluation result generation model service module, which uses the returned query results and the natural language text of the evaluation requirements to summarize and analyze and form the evaluation results.

8. A knowledge pre-retrieval method based on knowledge retrieval to enhance RAG workflow, characterized in that: The intelligent simulation evaluation system based on the hybrid knowledge retrieval enhancement and model context protocol according to claim 1 further includes the following steps: Step (1), knowledge vectorization process: Sub-step (1), the user submits the prior knowledge text on the interface through the knowledge base management component of the view layer; Sub-step (2), through the knowledge retrieval service of the RAG workflow system, the submitted prior knowledge text is divided into blocks and vectorized and stored in the vector database; Step 2: Search Enhancement Process: In sub-step (1), the user enters the natural language text of the query to be evaluated through the natural language evaluation component interface interactive dialog box of the view layer; In sub-step (2), the natural language processing service module in the MCP Client of the service layer summarizes and refines the natural language text input by the user through the large model to generate evaluation demand keywords; In sub-step (3), the RAG workflow interaction service module in the service layer MCP Client takes the evaluation requirement keyword as a parameter and requests the RAG workflow system service; In sub-step (4), the knowledge retrieval service module in the RAG workflow system vectorizes the evaluation requirement keywords, calculates the correlation with the vectors stored in the vector database, selects the top N vectors, and generates the context; In sub-step (5), the large model workflow service module in the RAG workflow system combines the generated context with the evaluation requirement keywords input by the user to generate a recommended tool type and suggested parameters, which are returned as search context to the RAG workflow interaction service, where the tool type is the selected data source and the suggested parameters are the query statement or API call request format; In sub-step (6), the RAG workflow interaction service module in the service layer MCP Client receives the returned search context for reference by the subsequent evaluation tool to select the large model service.

9. A natural language evaluation method combining MCP and RAG, characterized in that: The intelligent simulation evaluation system based on the hybrid knowledge retrieval enhancement and model context protocol according to claim 1 further includes the following steps: Step (1): The user enters the natural language text of the evaluation requirement through the natural language evaluation component of the view layer; Step 2: The natural language processing service module of the MCP Client in the service layer stores the natural language text of the assessment requirements in a historical query list for multiple rounds of dialogue. It also summarizes and refines it through a large model to generate assessment requirement keywords. Step (3): The RAG workflow interaction service module calls the RAG workflow system, uses the evaluation requirement keywords to perform knowledge retrieval and large model reasoning, obtains prior vector information related to data sources and business logic, and generates retrieval context for subsequent tool selection reference; Step (4): The MCP Server management service module queries and returns a list of tools provided by the MCP Server that has been registered with the MCP Client. Step (5): The evaluation tool optimization big model service module takes the evaluation requirement keywords, the search context generated by the RAG workflow, and the registered tool list as input, and the big model generates the tools to be called and the corresponding tool parameters; Step (six), the various databases or existing system API interaction services of the service layer MCP Server determine the database type based on the request; If a relational database query tool is called, it connects to the database and executes the generated SQL statement. If a non-relational database query tool is called, it connects to the database and executes the generated NoSQL statement. If a time series database query tool is called, it connects to the database and executes the generated InfluxQL statement. If an existing system API tool is called, it connects to the existing system server and executes the API call request. Finally, the query statement execution result or API call result is returned to the evaluation result generation large model service in the MCP Client. Step (7): The evaluation result generates a large model service module, which combines the returned query results with the natural language text of the evaluation requirements entered by the user and the search context information, and inputs them into the large model for data analysis and summary to form the evaluation results; Step (8): The natural language evaluation component of the view layer displays the evaluation results in the interface interaction dialog box, and the process ends. Step (9): If multiple rounds of dialogue are required, repeat steps (1) to (8).

10. The intelligent evaluation method based on the MCP model context protocol according to claim 7, characterized in that: In step (six), the data sources include relational simulation databases, non-relational simulation databases, time series simulation databases, and other system API services.

Images