Method and system for selecting and executing function in multi-task environment based on large model

Abstract

The invention discloses a function selection and execution method and system in a multi-task environment based on a large model, belongs to the technical field of artificial intelligence, and aims to solve the technical problem of how to understand natural language input of a user and automatically select and execute corresponding functions to complete complex tasks in the multi-task environment. Comprising the following steps: for each function in a user API function library, defining attribute information of the function, and constructing a feature vector library based on the function and a feature vector corresponding to the function; searching a function matched with the user question based on the vector similarity between the feature vector of the user question and the feature vector of the function; based on the user problem and the function matched with the user problem, analyzing the demand and expectation of the user through a large model, formulating a task execution strategy based on the demand and expectation of the user, selecting the function to be executed based on the task execution strategy, and constructing a function execution plan; and for the function execution plan, executing each function in sequence, and monitoring and adjusting the currently executed function.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for selecting and executing functions in a multi-task environment based on a large model. Background Art

[0002] The application of natural language processing (NLP) and artificial intelligence (AI) technologies in various fields is becoming more and more extensive. When dealing with complex and changing task environments, how to improve the intelligence level of the system, reduce manual intervention, and enhance user experience has become a key point in the development of technology.

[0003] In the process of using traditional API function libraries, users often need to have a certain technical background to effectively call and configure functions to meet specific business needs. In this case, users face problems such as complex operations, high learning costs, and low efficiency. In addition, with the continuous expansion and update of API function libraries, how to efficiently manage and utilize these function libraries has become a severe challenge for developers.

[0004] In a multi-task environment, how to understand user natural language input, automatically select and execute corresponding functions to complete complex tasks is a technical problem that needs to be solved. Summary of the invention

[0005] The technical task of the present invention is to address the above shortcomings and provide a function selection and execution method and system in a multi-task environment based on a large model to solve the technical problem of how to understand user natural language input, automatically select and execute corresponding functions to complete complex tasks in a multi-task environment.

[0006] In a first aspect, the present invention provides a method for selecting and executing functions in a multi-task environment based on a large model, comprising the following steps:

[0007] Function library modeling: For each function in the user API function library, define the function's attribute information, extract the function's feature vector, and build a feature vector library based on the function and its corresponding feature vector. The function's attribute information includes description, input field, output field, type, and function.

[0008] User question analysis: For user questions, extract the feature vector of the user question, and find the function that matches the user question based on the vector similarity between the feature vector of the user question and the feature vector of the function;

[0009] Function execution plan construction: Based on user questions and matching functions, the user's needs and expectations are analyzed through a large model, and task execution strategies are formulated based on the user's needs and expectations. Based on the task execution strategies, the functions to be executed are selected and a function execution plan is constructed.

[0010] Task execution: For the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the successful task execution result is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. Among them, when monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the big model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

[0011] Preferably, the feature vector of the function is extracted by a text embedding model, and the feature vector of the user question is extracted by a text embedding model.

[0012] Preferably, monitoring and adjusting the currently executed function includes the following steps:

[0013] During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function;

[0014] After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function;

[0015] If the current function is executed successfully, determine whether there is a function to be executed after the current function. If yes, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user.

[0016] If the current function fails to execute, the currently executed function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the big model. If the currently executed function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function execution is determined to have failed and the execution result of the failed execution task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

[0017] Preferably, the return value or status code returned by the function is used as the execution result, and based on the execution log of the current function and the returned execution result, the current function is subjected to log analysis, error type identification, and task execution strategy adjustment through a large model, to determine whether to adjust the currently executed function and / or the function execution plan, wherein adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

[0018] In a second aspect, the present invention provides a system for selecting and executing functions in a multi-task environment based on a large model, which is used to implement the selection and execution of functions in a multi-task environment through a method for selecting and executing functions in a multi-task environment based on a large model as described in any one of the first aspects, wherein the system includes a function library modeling module, a user problem analysis module, a function execution planning construction module, and a task execution module;

[0019] The function library modeling module is used to perform the following: for each function in the user API function library, define the attribute information of the function, extract the feature vector of the function, and build a feature vector library based on the function and the feature vector corresponding thereto, wherein the attribute information of the function includes description, input field, output field, type and function;

[0020] The user question analysis module is used to perform the following: for a user question, extract a feature vector of the user question, and based on the vector similarity between the feature vector of the user question and the feature vector of the function, find a function matching the user question;

[0021] The function execution plan building module is used to perform the following: based on user questions and matching functions, analyze user needs and expectations through a large model, formulate a task execution strategy based on user needs and expectations, select the function to be executed based on the task execution strategy and build a function execution plan;

[0022] The task execution module is used to perform the following: for the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the successful task execution result is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. Among them, when monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the large model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

[0023] Preferably, the function library modeling module is used to extract the feature vector of the function through the text embedding model, and the user question analysis module is used to extract the feature vector of the user question through the text embedding model.

[0024] Preferably, the task execution module is used to perform the following monitoring and adjustment on the currently executed function:

[0025] During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function;

[0026] After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function;

[0027] If the current function is executed successfully, determine whether there is a function to be executed after the current function. If yes, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user.

[0028] If the current function fails to execute, the currently executed function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the big model. If the currently executed function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function execution is determined to have failed and the execution result of the failed execution task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

[0029] Preferably, the task execution module is used to use the return value or status code returned by the function as the execution result, and based on the execution log of the current function and the returned execution result, perform log analysis on the current function through a large model, identify the error type, and adjust the task execution strategy, determine whether to adjust the currently executed function and / or adjust the function execution plan, wherein adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

[0030] The method and system for selecting and executing functions in a multi-task environment based on a large model of the present invention have the following advantages:

[0031] 1. Improved function usage efficiency: For each function in the user API function library, define the function's attribute information, extract the function's feature vector, and build a feature vector library based on the function and its corresponding feature vector, so as to facilitate understanding of the function and intelligent selection of the function;

[0032] 2. Optimized user experience: Analyze user questions through a large model to obtain user needs and expectations, formulate task execution strategies based on user needs and expectations, select functions to be executed based on the task execution strategies and build function execution plans, thereby realizing the automatic and intelligent generation of function execution plans. Users do not need to have an in-depth understanding of the function, and can obtain function execution plans by asking questions, which meets the needs of users, especially non-professionals, and reduces the complexity of operations;

[0033] 3. Reduce manual intervention: Automated problem-solving processes reduce the need for manual operations and reduce the frequency and degree of manual intervention;

[0034] 4. Reduced labor costs: Reduced reliance on professionals, thereby reducing labor costs;

[0035] 5. Improved problem-solving accuracy: Driven by large models, the system can more accurately understand and execute user intentions, thus improving the accuracy of problem solving;

[0036] 6. Enhanced flexibility: The large model can work in a multi-task environment, adapt to different task types and requirements, and has high flexibility;

[0037] 7. Improved automation level: Automated execution steps and error handling mechanisms improve the automation level of the entire large model system;

[0038] 8. Improved task completion rate: Through detailed planning and execution steps, the consistency and correctness of the tasks are ensured, thereby improving the task completion rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] The present invention is further described below in conjunction with the accompanying drawings.

[0041] Figure 1 This is a flowchart of a method for selecting and executing functions in a multi-task environment based on a large model according to Example 1. DETAILED DESCRIPTION

[0042] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments may be combined with each other.

[0043] The embodiments of the present invention provide a method and system for selecting and executing functions in a multi-task environment based on a large model, which are used to solve the technical problem of how to understand user natural language input, automatically select and execute corresponding functions to complete complex tasks in a multi-task environment.

[0044] Embodiment 1:

[0045] The present invention discloses a method for selecting and executing functions in a multi-task environment based on a large model, comprising four steps: function library construction, user problem analysis, function execution plan construction and task execution.

[0046] Step S100: Function library modeling: For each function in the user API function library, define the function's attribute information, extract the function's feature vector, and build a feature vector library based on the function and its corresponding feature vector, wherein the function's attribute information includes description, input field, output field, type, and function.

[0047] In this embodiment, the feature vector of the function is extracted through a text embedding model.

[0048] Step S200: User question analysis: for the user question, extract the feature vector of the user question, and search for a function matching the user question based on the vector similarity between the feature vector of the user question and the feature vector of the function.

[0049] In this embodiment, for the user's questions or problems, the feature vector of the user's question is extracted through the text embedding model, the cosine similarity between the user's question feature vector and the function feature vector is calculated, and the function matching the user's question is found based on the similarity between the feature vectors.

[0050] Step S300: Constructing a function execution plan: Based on user questions and matching functions, analyze user needs and expectations through a large model, formulate a task execution strategy based on user needs and expectations, select the function to be executed based on the task execution strategy and construct a function execution plan.

[0051] In this embodiment, a large model of prompt word input is constructed based on user questions and matching functions. The large model analyzes the user's needs and expectations based on the input prompt words to evaluate the user's intentions. If the current output user intention does not meet the user's expectations, the user will continue to be questioned and the user's intentions will be evaluated again based on the user's questions and problems until the user's needs and expectations are met.

[0052] Step S400 task execution: for the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the task execution result of successful execution is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. When monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the large model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

[0053] In this embodiment, monitoring and adjusting the currently executed function includes the following steps:

[0054] (1) During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function;

[0055] (2) After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function;

[0056] (3) If the current function is executed successfully, determine whether there is a function to be executed after the current function. If so, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user;

[0057] (4) If the current function fails to execute, the current function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the large model. If the current function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function is determined to have failed and the execution result of the failed task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

[0058] As a specific implementation, the return value or status code returned by the function is used as the execution result. Based on the execution log of the current function and the returned execution result, the current function is subjected to log analysis, error type identification, and task execution strategy adjustment through a large model, to determine whether to adjust the currently executed function and / or the function execution plan. Adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

[0059] During the task execution process, after all planning steps have been executed, the big model will perform a final check to confirm that all user requirements have been met and there are no remaining execution steps. If everything is normal, a task completion notification will be sent to the user and the task execution results will be provided.

[0060] Embodiment 2:

[0061] The present invention discloses a function selection and execution system in a multi-task environment based on a large model, comprising a function library modeling module, a user problem analysis module, a function execution planning construction module and a task execution module.

[0062] The function library modeling module is used to perform the following: for each function in the user API function library, define the function's attribute information and extract the function's feature vector, and build a feature vector library based on the function and its corresponding feature vector, where the function's attribute information includes description, input field, output field, type and function.

[0063] Among them, the function library modeling module in this embodiment is used to extract the feature vector of the function through the text embedding model.

[0064] The user question analysis module is used to perform the following: for a user question, extract a feature vector of the user question, and based on the vector similarity between the feature vector of the user question and the feature vector of the function, find a function matching the user question.

[0065] In this embodiment, for the user's questions or problems, the user question analysis module is used to perform the following: extract the feature vector of the user question through the text embedding model, calculate the cosine similarity between the user question feature vector and the function feature vector, and find the function that matches the user question based on the similarity between the feature vectors.

[0066] The function execution planning construction module is used to perform the following: based on user questions and matching functions, analyze user needs and expectations through a large model, formulate a task execution strategy based on user needs and expectations, select the function to be executed based on the task execution strategy and build a function execution plan.

[0067] In this embodiment, the function execution planning construction module is used to construct a prompt word input large model based on user questions and matching functions. The large model analyzes the user's needs and expectations based on the input prompt words to evaluate the user's intentions. If the current output user intention does not meet the user's expectations, the user will continue to be questioned and the user's intentions will be evaluated again based on the user's questions and problems until the user's needs and expectations are met.

[0068] The task execution module is used to perform the following: for the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the successful task execution result is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. Among them, when monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the large model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

[0069] In this embodiment, the task execution module is used to perform the following monitoring and adjustment on the currently executed function:

[0070] (1) During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function;

[0071] (2) After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function;

[0072] (3) If the current function is executed successfully, determine whether there is a function to be executed after the current function. If so, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user;

[0073] (4) If the current function fails to execute, the current function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the large model. If the current function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function is determined to have failed and the execution result of the failed task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

[0074] As a specific implementation, the return value or status code returned by the function is used as the execution result. Based on the execution log of the current function and the returned execution result, the current function is subjected to log analysis, error type identification, and task execution strategy adjustment through a large model, to determine whether to adjust the currently executed function and / or the function execution plan. Adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

[0075] During the task execution process, after all planning steps have been executed, the big model will perform a final check to confirm that all user requirements have been met and there are no remaining execution steps. If everything is normal, a task completion notification will be sent to the user and the task execution results will be provided.

[0076] The above is a detailed introduction to the function selection and execution method and system in a multi-tasking environment based on a large model provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for selecting and executing functions in a multi-task environment based on a large model, characterized in that: The steps include: Function library modeling: For each function in the user API function library, define the function's attribute information, extract the function's feature vector, and build a feature vector library based on the function and its corresponding feature vector. The function's attribute information includes description, input field, output field, type, and function. User question analysis: For user questions, extract the feature vector of the user question, and find the function that matches the user question based on the vector similarity between the feature vector of the user question and the feature vector of the function; Function execution plan construction: Based on user questions and matching functions, the user's needs and expectations are analyzed through a large model, and task execution strategies are formulated based on the user's needs and expectations. The functions to be executed are selected based on the task execution strategies and a function execution plan is constructed. Task execution: For the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the successful task execution result is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. Among them, when monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the big model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

2. The method for selecting and executing functions in a multi-task environment based on a large model according to claim 1, characterized in that: The feature vector of the function is extracted through the text embedding model, and the feature vector of the user question is extracted through the text embedding model.

3. The method for selecting and executing functions in a multi-task environment based on a large model according to claim 1, characterized in that: Monitor and adjust the currently executed function, including the following steps: During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function; After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function; If the current function is executed successfully, determine whether there is a function to be executed after the current function. If yes, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user. If the current function fails to execute, the currently executed function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the big model. If the currently executed function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function execution is determined to have failed and the execution result of the failed execution task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

4. The method for selecting and executing functions in a multi-task environment based on a large model according to claim 1, characterized in that: Taking the return value or status code returned by the function as the execution result, based on the execution log of the current function and the returned execution result, log analysis of the current function is performed through a large model, error type identification, and task execution strategy adjustment, it is determined whether to adjust the currently executed function and / or the function execution plan, wherein adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

5. A function selection and execution system in a multi-task environment based on a large model, characterized in that: Used to realize the selection and execution of functions in a multi-task environment by a method for selecting and executing functions in a multi-task environment based on a large model as described in any one of claims 1 to 4, the system comprising a function library modeling module, a user problem analysis module, a function execution planning construction module and a task execution module; The function library modeling module is used to perform the following: for each function in the user API function library, define the attribute information of the function, extract the feature vector of the function, and build a feature vector library based on the function and the feature vector corresponding thereto, wherein the attribute information of the function includes description, input field, output field, type and function; The user question analysis module is used to perform the following: for a user question, extract a feature vector of the user question, and based on the vector similarity between the feature vector of the user question and the feature vector of the function, find a function matching the user question; The function execution plan building module is used to perform the following: based on user questions and matching functions, analyze user needs and expectations through a large model, formulate a task execution strategy based on user needs and expectations, select the function to be executed based on the task execution strategy and build a function execution plan; The task execution module is used to perform the following: for the function execution plan, execute each function in turn, and monitor and adjust the currently executed function. If all functions in the function execution plan are executed successfully, the successful task execution result is returned to the user. If there is a function execution failure in the function execution plan, the execution result of the failed task is returned to the user. Among them, when monitoring and adjusting the currently executed function, if the currently executed function returns the execution result of success, the next function is executed. If the currently executed function returns the execution result of failure, based on the execution log of the currently executed function and the returned execution result, the currently executed function and the function execution plan are adjusted and judged through the large model. Based on the adjustment judgment result, it is selected to re-execute the adjusted function or re-execute the adjusted function execution plan. If the adjusted function is re-executed, the adjusted function is monitored and judged. If the adjusted function execution plan is re-executed, it is determined that the current function execution has failed and the execution result of the failed task is returned to the user.

6. The function selection and execution system in a multi-task environment based on a large model according to claim 5, characterized in that: The function library modeling module is used to extract the feature vector of the function through the text embedding model, and the user question analysis module is used to extract the feature vector of the user question through the text embedding model.

7. The function selection and execution system in a multi-task environment based on a large model according to claim 5, characterized in that: The task execution module is used to monitor and adjust the currently executed function as follows: During the execution of the current function, the execution process of the current function is recorded to form the execution log of the current function; After the current function is executed, determine whether the current function is executed successfully based on the execution result returned by the current function; If the current function is executed successfully, determine whether there is a function to be executed after the current function. If yes, execute the next function. If not, determine that the task is executed successfully and return the successful task execution result to the user. If the current function fails to execute, the currently executed function and the function execution plan are adjusted and judged based on the execution log of the current function and the returned execution result through the big model. If the currently executed function is adjusted, the adjusted function is re-executed and the adjusted function is monitored and judged. If the function execution plan is adjusted, the current function execution is determined to have failed and the execution result of the failed execution task is returned to the user, and the task execution operation is performed based on the adjusted function execution plan.

8. The function selection and execution system in a multi-task environment based on a large model according to claim 5, characterized in that: The task execution module is used to use the return value or status code returned by the function as the execution result, based on the execution log of the current function and the returned execution result, log analysis of the current function through a large model, error type identification and task execution strategy adjustment, to determine whether to adjust the currently executed function and / or the function execution plan, wherein adjusting the currently executed function includes redefining the attribute information of the function, and adjusting the function execution plan includes selecting the function to be executed and the execution order of the functions to form a new function execution plan.

Images