CN-122022422-A - Flow execution method, device, equipment and medium based on large language model

CN122022422ACN 122022422 ACN122022422 ACN 122022422ACN-122022422-A

Abstract

The invention provides a flow execution method, a device, equipment and a medium based on a large language model, wherein the flow execution method based on the large language model comprises the steps of obtaining a plurality of target flow nodes configured by a user; the target process node comprises at least one preconfigured process node constructed based on a large language model, a process instance is created based on the target process node, user input information is obtained, and the process instance is executed based on the user input information. The invention can carry out intelligent remodelling on the business process by using the large language model for the process treatment. The large language model enables the process to be more automatic, more accurate and more intelligent through deep fusion of cognitive abilities.

Inventors

ZHAO YUNTAO
HUANG CHUAN
REN SIYU
LUO ZIHAN
HUANG DAN

Assignees

吉旗（成都）科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251202

Claims (10)

1. A large language model-based flow execution method, comprising: obtaining a plurality of target flow nodes configured by a user, wherein the target flow nodes comprise at least one preconfigured flow node constructed based on a large language model; creating a flow instance based on the target flow node; user input information is acquired, and the process instance is executed based on the user input information.
2. The large language model based flow execution method of claim 1, wherein the pre-configuring the flow node comprises: A start node for determining a data type of user input information; Model nodes for interacting with the large language model; an intention judgment node for determining intention information based on user input information; The knowledge base node is used for inquiring the knowledge base based on the input of the knowledge base node; The problem classification node is used for requesting the large language model to carry out classification selection according to the input information of the user and the output of the last flow node, and routing to the downstream flow node according to the classification selection result; And the tool node is used for extracting parameters of the tool by using the large language model according to the input information of the user or the output of the last flow node, and calling the tool to return an extraction result.
3. The large language model-based flow execution method of claim 2, wherein the model node is specifically configured to perform the invocation of the target large language model by combining the output of the previous flow node according to the target large language model selected by the user and the configured prompt template; the knowledge base node is specifically configured to perform keyword search, similarity search or vector search in the knowledge base according to user input information or output of a previous flow node.
4. The large language model based flow execution method of claim 2, wherein the pre-configuring the flow node further comprises: a document processing node for converting user input information into text data; the document generation node is used for converting the output of the last process node into a document in a target format for output; and the chart generating node is used for generating echart codes through a large language model according to the user input information and the output of the last flow node and generating chart data based on the echart codes.
5. The large language model based flow execution method of claim 2, wherein the pre-configuring the flow node further comprises: the direct replying node is used for outputting an execution result to a user based on the configured text information or the output of the last node; And the MCP node is used for determining a target MCP service according to the input information of a user or the output of the last flow node, extracting parameters and executing tools through a model selection tool, and returning the result.
6. The large language model based flow execution method according to any one of claims 1 to 5, wherein after the obtaining user input information and executing the flow instance based on the user input information, the method further comprises: and storing the session information in the process of executing the flow instance.
7. A large language model-based flow execution device, comprising: The system comprises an acquisition module, a configuration module and a configuration module, wherein the acquisition module is configured to acquire a plurality of target flow nodes configured by a user, and the target flow nodes comprise at least one preconfigured flow node constructed based on a large language model; A creating module configured to create a flow instance based on the target flow node; and the execution module is configured to acquire user input information and execute the flow instance based on the user input information.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements the large language model based flow execution method of any one of claims 1 to 6 when the computer program is executed.
9. A non-transitory computer-readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the large language model-based flow execution method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program which, when executed by a processor, implements a large language model based flow execution method according to any one of claims 1 to 6.

Description

Flow execution method, device, equipment and medium based on large language model Technical Field The present invention relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a medium for executing a flow based on a large language model. Background With the rapid development of information technology, new challenges are brought to the construction of an automation flow. Most of the traditional automatic flow node generation methods depend on manual design, so that not only is a designer required to have knowledge reserves across fields, but also a great deal of time is consumed to comb the association relation among data, and the situations of data volume surge and modal diversification are difficult to adapt, so that the method cannot be completely and accurately executed. The process flow of the enterprise task is strict and standard, and the process flow of the enterprise task is standard and complex, so that the existing constructed automatic flow cannot completely and accurately process the flow. Disclosure of Invention The invention provides a flow execution method, device, equipment and medium based on a large language model, which are used for solving the defects in the prior art. The invention provides a flow execution method based on a large language model, which comprises the following steps: obtaining a plurality of target flow nodes configured by a user, wherein the target flow nodes comprise at least one preconfigured flow node constructed based on a large language model; creating a flow instance based on the target flow node; user input information is acquired, and the process instance is executed based on the user input information. According to the flow execution method based on the large language model provided by the invention, the pre-configuration flow node comprises the following steps: A start node for determining a data type of user input information; Model nodes for interacting with the large language model; an intention judgment node for determining intention information based on user input information; The knowledge base node is used for inquiring the knowledge base based on the input of the knowledge base node; The problem classification node is used for requesting the large language model to carry out classification selection according to the input information of the user and the output of the last flow node, and routing to the downstream flow node according to the classification selection result; And the tool node is used for extracting parameters of the tool by using the large language model according to the input information of the user or the output of the last flow node, and calling the tool to return an extraction result. According to the flow execution method based on the large language model, the model node is specifically used for calling the target large language model according to the target large language model selected by the user and the configured prompt template and combining the output of the last flow node; the knowledge base node is specifically configured to perform keyword search, similarity search or vector search in the knowledge base according to user input information or output of a previous flow node. According to the flow execution method based on the large language model provided by the invention, the pre-configuration flow node further comprises: a document processing node for converting user input information into text data; the document generation node is used for converting the output of the last process node into a document in a target format for output; and the chart generating node is used for generating echart codes through a large language model according to the user input information and the output of the last flow node and generating chart data based on the echart codes. According to the flow execution method based on the large language model provided by the invention, the pre-configuration flow node further comprises: the direct replying node is used for outputting an execution result to a user based on the configured text information or the output of the last node; And the MCP node is used for determining a target MCP service according to the input information of a user or the output of the last flow node, extracting parameters and executing tools through a model selection tool, and returning the result. According to the flow execution method based on the large language model provided by the invention, after the user input information is acquired and the flow instance is executed based on the user input information, the method further comprises the following steps: and storing the session information in the process of executing the flow instance. The invention also provides a flow executing device based on the large language model, which comprises: The system comprises an acquisition module, a configuration module and a configuration module, wherein the acquisition module is configured t