CN-121981665-A - Platform for realizing management and control of intelligent rules of accounting

CN121981665ACN 121981665 ACN121981665 ACN 121981665ACN-121981665-A

Abstract

The invention relates to a platform for realizing management and control of an intelligent rule of accounting, and belongs to the technical field of accounting. The platform comprises a background support type module, an application service type module, an accounting knowledge base related module, a plurality of MCP servers, an intelligent body main control module and a client module, wherein the invention uses a large language model as an intelligent core based on an MCP protocol, builds an accounting knowledge base by integrating knowledge loaded by an approval process and knowledge in an accounting rule base in an accounting system, and assists accounting staff to realize interpretation of policies, enterprise accounting rules and related management requirements and match with the accounting rules through a natural language interactive interface, thereby assisting the accounting staff to build and adjust the accounting rules, improving the working efficiency and guaranteeing the consistency and the integrity of the accounting rules.

Inventors

FENG LIUJUN
PENG JINSONG
LI BO
ZHU KE
BAO RUI

Assignees

北京南天软件有限公司
云南南天电子信息产业股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251212

Claims (9)

1. The platform for realizing the management and control of the accounting intelligent rules is characterized by comprising a background support type module, an application service type module, an accounting knowledge base related module, a plurality of MCP servers, an agent main control module and a client module; The background support type module adopts a large language model and is used for providing large language model service; the application service class module comprises an accounting system, wherein the accounting system stores an accounting rule base and provides an accounting rule query service and an accounting rule change application service; the accounting knowledge base related module comprises a knowledge warehouse-in management system and an accounting knowledge base; the accounting knowledge base is a vector database and is used for storing vectorized knowledge text blocks and metadata thereof; The knowledge storage management system is used for receiving a knowledge storage request from the client and/or the MCP server, analyzing, dividing and vectorizing the knowledge document, and storing the knowledge document into an accounting knowledge base; The plurality of MCP servers includes: The accounting rule MCP server is used for providing an interface with the accounting system and converting the JSON format and the transaction data format of the accounting system; the knowledge base MCP server is used for providing a natural language query and metadata condition query interface for the accounting knowledge base; the knowledge warehouse-in MCP server is used for receiving the knowledge of the accounting rules and the knowledge generated in the interaction process of the intelligent agent main control module and the large language model, and submitting the knowledge to a knowledge warehouse-in management system; The intelligent assistant intelligent agent is used for receiving the session request from the client module, scheduling the large language model for reasoning, and calling the corresponding MCP server according to the reasoning result to complete the task, including calling the knowledge base MCP server according to the user inquiry to obtain the related information, sending the information to the large language model to generate an answer after prompting the engineering organization, and calling the accounting rule MCP server or the functions and parameters of the knowledge warehouse MCP server according to the user instruction organization; the client module comprises an accounting intelligent assistant client and a knowledge warehouse-in management client which are respectively used for providing an interactive interface between a user and the main control module of the intelligent body and a management and approval interface of the knowledge warehouse-in.
2. The platform for implementing intelligent rule management of accounting according to claim 1, wherein the knowledge base management system is configured to call a large language model, automatically generate classified words or phrases based on the content of the management class knowledge document, store document content text blocks and the generated classified words and phrases as independent sets in an accounting knowledge base, and establish correspondence by associating document identifications in metadata of the classified words and phrases.
3. The platform for realizing intelligent rule management and control of accounting according to claim 1, wherein the accounting rule knowledge received by the knowledge-base MCP server is JSON format data after serializing relation data in the accounting rule base, and fields are added in the JSON data to describe the corresponding relation between the accounting rule and related management class knowledge classification words and phrases, and metadata of the accounting rule includes one or more of entry type codes, version numbers, enabling marks, enabling dates and disabling dates.
4. The platform for implementing intelligent rule management and control of accounting according to claim 1, wherein the knowledge warehouse management system includes a data loading module for analyzing a document, text-dividing a plain text obtained by analysis, and vectorizing by using an embedded model.
5. The platform for implementing intelligent rule management for accounting according to claim 1, wherein the interface functions provided by the accounting rules MCP server include invoking an accounting rule query transaction function of the accounting system and the accounting rule add, modify, and disable application transaction functions.
6. The method for managing the rules of the intelligent assistant for accounting by adopting the platform for realizing the management and control of the intelligent rules for accounting according to any one of claims 1 to 5 is characterized by comprising the following steps: S1, a user initiates a session containing natural language instructions through an accounting intelligent assistant client; s2, the intelligent agent main control module receives the session, organizes callable MCP server function information and user instructions into prompt words, and submits the prompt words to the large language model; S3, returning a JSON format instruction containing functions to be called and parameters by the large language model; S4, the main control module of the intelligent agent calls a corresponding MCP server to execute operation according to the JSON format instruction; s5, the agent main control module acquires a return result of the MCP server, and returns the result to the client module directly according to the task type, or sends the result to the large language model to generate a final answer to return to the client after prompting engineering organization.
7. The method according to claim 6, wherein the MCP server called in the step S4 is a knowledge base MCP server, and the operations performed include querying the highest version of the accounting rules existing in the accounting knowledge base according to the user instruction, further querying the associated management class knowledge based on the associated classification words and phrases in the accounting rules, and returning the query result to the agent host module.
8. The method according to claim 6, wherein the MCP servers called in step S4 are an accounting rules MCP server and a knowledge-warehousing MCP server, and the operations performed include a process of loading the updated accounting rules into an accounting knowledge base, specifically: s41, calling an accounting rule MCP server, and acquiring a currently-started accounting rule set A from an accounting system; s42, calling a knowledge base MCP server, and acquiring a highest version accounting rule set B corresponding to each entry type code from an accounting knowledge base; s43, comparing the set A with the set B, and identifying a missing or lower version accounting rule in the accounting knowledge base to form a set C to be loaded; s44, calling a knowledge base MCP server to acquire all management class knowledge classification words and phrases; s45, submitting the accounting rules, the classification words and the short sentences in the set C to be loaded to a large language model, and requesting to add associated classification words and short sentence fields for the accounting rules; s46, receiving the reinforced accounting rules returned by the large language model, and displaying the reinforced accounting rules to a user through a client for confirmation or adjustment; s47, calling a knowledge-base MCP server according to a user confirmation instruction, submitting the final version of calculation rules to be loaded to a knowledge-base management system, and starting a base approval process.
9. The method according to claim 6, wherein the MCP server called in step S4 is an accounting rules MCP server, and the operations performed include a process of generating and submitting an accounting rules application through a dialogue, specifically: s41, according to a user instruction, calling a knowledge base MCP server to inquire related policy difference information and existing accounting rule information; s42, submitting the queried policy difference information, the existing accounting rule information and the user instruction to a large language model, and requesting the large language model to generate an newly added or modified accounting rule draft; S43, returning the generated accounting rule draft to the client module for verification and adjustment by a user; s44, according to the application instruction submitted by the user, the new application function or the new application function modification corresponding to the accounting rule MCP server is called, and the accounting rule draft confirmed by the user is submitted to the accounting system to form the application items to be processed.

Description

Platform for realizing management and control of intelligent rules of accounting Technical Field The invention belongs to the technical field of accounting, and particularly relates to a platform for realizing management and control of intelligent rules of accounting. Background At present, an application system of a financial institution generally performs an accounting process automatically by linking a business function module and an accounting module (or a module providing an accounting function such as an accounting center). In order to realize automatic accounting processing, the module needs to be configured with accounting rules, and the automatic accounting processing is realized through rule data driving. The accounting rule data carries out association matching on the specific business and the corresponding accounting method information. The accounting rules are quite large in scale. When the policy, the enterprise accounting rules and the related management requirements are changed, the accounting rules need to be correspondingly adjusted according to the requirements. Currently, policy, criteria, and management requirements are read, and accounting rule adjustment work is handled by accounting personnel entirely manually. Because the quantity of the accounting rules is huge, the logic relationship is complex, and the query checking and the adjustment of the accounting rules are very huge in workload and easy to cause errors. Therefore, how to overcome the defects of the prior art is a problem to be solved in the technical field of the current accounting technology. Disclosure of Invention The invention aims to solve the defects of the prior art and provides a platform for realizing management and control of intelligent rules of accounting. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: the platform for realizing the management and control of the accounting intelligent rules comprises a background support type module, an application service type module, an accounting knowledge base related module, a plurality of MCP servers, an agent main control module and a client module; The background support type module adopts a large language model and is used for providing large language model service; the application service class module comprises an accounting system, wherein the accounting system stores an accounting rule base and provides an accounting rule query service and an accounting rule change application service; the accounting knowledge base related module comprises a knowledge warehouse-in management system and an accounting knowledge base; the accounting knowledge base is a vector database and is used for storing vectorized knowledge text blocks and metadata thereof; The knowledge storage management system is used for receiving a knowledge storage request from the client and/or the MCP server, analyzing, dividing and vectorizing the knowledge document, and storing the knowledge document into an accounting knowledge base; The plurality of MCP servers includes: The accounting rule MCP server is used for providing an interface with the accounting system and converting the JSON format and the transaction data format of the accounting system; the knowledge base MCP server is used for providing a natural language query and metadata condition query interface for the accounting knowledge base; the knowledge warehouse-in MCP server is used for receiving the knowledge of the accounting rules and the knowledge generated in the interaction process of the intelligent agent main control module and the large language model, and submitting the knowledge to a knowledge warehouse-in management system; The intelligent assistant intelligent agent is used for receiving the session request from the client module, scheduling the large language model for reasoning, and calling the corresponding MCP server according to the reasoning result to complete the task, including calling the knowledge base MCP server according to the user inquiry to obtain the related information, sending the information to the large language model to generate an answer after prompting the engineering organization, and calling the accounting rule MCP server or the functions and parameters of the knowledge warehouse MCP server according to the user instruction organization; the client module comprises an accounting intelligent assistant client and a knowledge warehouse-in management client which are respectively used for providing an interactive interface between a user and the main control module of the intelligent body and a management and approval interface of the knowledge warehouse-in. Further, the knowledge warehouse-in management system is configured to call a large language model, automatically generate classified words or phrases based on the content of the management knowledge document, store document content text blocks and the generated classified words and phrases as independe