Service plug-in calling method based on large language model

The invention discloses a large language model-based service plug-in calling method, which comprises the following steps of: (a) setting middleware in a large language model and a user side, so that the middleware firstly receives user information, performs semantic retrieval and acquires context information, forms a semantic vector and inputs the semantic vector to the large language model; the method comprises the following steps: (a) obtaining a big language model, (b) disassembling a semantic vector by using the big language model so as to disassemble a question and an entity parameter input by a user, (c) enabling middleware to find out a service needing to be called according to the disassembled question, inputting the entity parameter into the service, assembling information returned by the service into a complete answer through the big language model, and sending the complete answer to the middleware. And returning to the user side. By setting the middleware for connecting the large language model wit