Predicate-guided preparation for LLM integrations in spreadsheet environments
Technology is disclosed herein for the integration of spreadsheet environments and LLM services. In an implementation, an application receives a natural language input from a user associated with a spreadsheet. The application generates a prompt based on the user input and at least a portion of the spreadsheet. The prompt includes a statement of the problem, a request for an LLM service to identify preparatory steps before generating a solution to the problem, and a request to include the preparatory steps in output that includes the solution to problem. The application service receives a reply to the prompt from the LLM service that includes the output. The application service implements the preparatory steps with respect to data in the spreadsheet and implements the solution to the problem with respect to the data in the spreadsheet.
1 . A computing apparatus comprising:
one or more computer-readable storage media;
one or more processors operatively coupled with the one or more computer-readable storage media; and
an application comprising program instructions stored on the one or more computer-readable storage media that, when executed by the one or more processors, direct the computing apparatus to at least:
receive, via a user interface of the application, natural language input associated with a user of a spreadsheet;
generate a prompt based on the natural language input and a portion of the spreadsheet, wherein the prompt includes a statement of a problem, a request for a large language model (LLM) service to identify one or more preparatory steps before generating a solution to the problem, and a second request to include the one or more preparatory steps in output that includes the solution to the problem;
receive a reply to the prompt that includes the output, wherein the output comprises a spreadsheet formula and the one or more preparatory steps, wherein the one or more preparatory steps include evaluating spreadsheet data for data that violates a predicate, and modifying the data using a cleaning function associated with the predicate;
display, in the user interface, a graphical input device to accept the spreadsheet formula;
receive, via the user interface, a selection of the graphical input device to accept the spreadsheet formula;
implement the one or more preparatory steps with respect to the spreadsheet data, including by evaluating the spreadsheet data for the data that violates the predicate, and executing the cleaning function to modify the data to satisfy the predicate; and
recalculate the spreadsheet data according to the spreadsheet formula.
2 . The computing apparatus of claim 1 , wherein the predicate evaluates a format of the spreadsheet data identified by the LLM service based on the portion of the spreadsheet.
3 . The computing apparatus of claim 2 , wherein the predicate comprises a Boolean operator.
4 . The computing apparatus of claim 1 , wherein the one or more preparatory steps include generating assumptions about the data and testing the assumptions.
5 . The computing apparatus of claim 4 , wherein the one or more preparatory steps include generating predicates for testing the data based on the assumptions.
6 . The computing apparatus of claim 5 , wherein the one or more preparatory steps further include generating a cleaning function comprising operations performed on the data, wherein the operations are associated with the predicates.
7 . The computing apparatus of claim 1 , wherein the program instructions further direct the computing apparatus to display, in the user interface of the application, the solution to the problem.
8 . The computing apparatus of claim 7 , wherein the program instructions further direct the computing apparatus to receive user input indicative of a command to implement the solution to the problem with respect to the data in the spreadsheet.
9 . A method comprising:
receiving, via a user interface of an application, natural language input associated with a user of a spreadsheet;
generating a prompt based on the natural language input and a portion of the spreadsheet, wherein the prompt includes a statement of a problem, a request for a large language model (LLM) service to identify one or more preparatory steps before generating a solution to the problem, and a second request to include the one or more preparatory steps in output that includes the solution to the problem;
receiving a reply to the prompt that includes the output, wherein the output comprises a spreadsheet formula and the one or more preparatory steps, wherein the one or more preparatory steps include evaluating spreadsheet data for data that violates a predicate, and modifying the data using a cleaning function associated with the predicate;
displaying, in the user interface, a graphical input device to accept the spreadsheet formula;
receiving, via the user interface, a selection of the graphical input device to accept the spreadsheet formula;
implementing the one or more preparatory steps with respect to the spreadsheet data in the spreadsheet, including by evaluating the spreadsheet data for the data that violates the predicate, and executing the cleaning function to modify the data to satisfy the predicate; and
recalculating the spreadsheet data according to the spreadsheet formula.
10 . The method of claim 9 , wherein the predicate evaluates a format of the spreadsheet data identified by the LLM service based on the portion of the spreadsheet data.
11 . The method of claim 10 , wherein the predicate comprises a Boolean operator.
12 . The method of claim 9 , wherein the one or more preparatory steps include generating assumptions about the data and testing the assumptions.
13 . The method of claim 12 , wherein the one or more preparatory steps include generating predicates for testing the data based on the assumptions.
14 . The method of claim 13 , wherein the one or more preparatory steps further include generating a cleaning function comprising operations performed on the data, wherein the operations are associated with the predicates.
15 . One or more computer readable storage media having program instructions stored thereon that, when executed by one or more processors operatively coupled with the one or more computer-readable storage media, direct a computing device to:
receive, via a user interface of an application, natural language input associated with a user of a spreadsheet;
generate a prompt based on the natural language input and a portion of the spreadsheet, wherein the prompt includes a statement of a problem, a request for a large language model (LLM) service to identify one or more preparatory steps before generating a solution to the problem, and a second request to include the one or more preparatory steps in output that includes the solution to the problem;
receive a reply to the prompt that includes the output, wherein the output comprises a spreadsheet formula and the one or more preparatory steps, wherein the one or more preparatory steps include evaluating spreadsheet data for data that violates a predicate, and modifying the data using a cleaning function associated with the predicate;
display, in the user interface, a graphical input device to accept the spreadsheet formula;
receive, via the user interface, a selection of the graphical input device to accept the spreadsheet formula;
implement the one or more preparatory steps with respect to the spreadsheet data, including by evaluating the spreadsheet data for the data that violates the predicate, and executing the cleaning function to modify the data to satisfy the predicate; and
recalculate the spreadsheet data according to the spreadsheet formula.
16 . The one or more computer readable storage media of claim 15 , wherein the predicate evaluates a format of the spreadsheet data identified by the LLM service based on the portion of the spreadsheet data.
17 . The one or more computer readable storage media of claim 16 , wherein the predicate comprises a Boolean operator.
18 . The one or more computer readable storage media of claim 15 , wherein the one or more preparatory steps include generating assumptions about the data and testing the assumptions.
19 . The one or more computer readable storage media of claim 18 , wherein the one or more preparatory steps include generating predicates for testing the data based on the assumptions.
20 . The one or more computer readable storage media of claim 19 , wherein the one or more preparatory steps further include generating a cleaning function comprising operations performed on the data, wherein the operations are associated with the predicates.