USER INTERFACE FOR IMPROVED LARGE LANGUAGE MODEL INTEGRATION
An online system improves the development and deployment of LLM-based applications providing offering an application workflow user interface to developers of these applications. The application workflow UI is a user interface that includes different sections for the efficient design of a workflow and logic for an application operating on the online system. The application workflow UI allows the user to generate UI elements for stages within the overall application workflow and connect those stages through a simple to use interface. The application workflow UI improves the development process of LLM-based applications by allowing the user to set parameters for stages that involve prompting an LLM through the same interface. Specifically, the application workflow UI includes a parameters section that allows the user to specify parameters for prompting the LLM.
1 . A non-transitory computer-readable medium storing instructions that, when executed by a computer system, cause the computer system to perform operations comprising:
receiving a plurality of prompt stages through an application workflow user interface presented on a client device, wherein each prompt stage is a stage in a workflow for an application executing on an online system and wherein each prompt stage represents a prompt to be transmitted to a large language model;
presenting a prompt stage element for each of the plurality of prompt stages in a workflow section of the application workflow user interface, wherein each prompt stage element is a user interface element representing the corresponding prompt stage element in the workflow section of the application workflow user interface;
receiving prompt text for each prompt stage element through a parameters section of the application workflow user interface, wherein the prompt text for a prompt stage element comprises text for the prompt for the corresponding prompt stage;
translating the plurality of prompt stage elements into computer-executable instructions to execute the workflow of the application on the online system; and
executing the computer-executable instructions on the online system.
2 . The non-transitory computer-readable medium of claim 1 , wherein the application is a chatbot application.
3 . The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise:
receiving a plurality of query stages through the application workflow user interface, wherein each query stage is a stage in the workflow for the application and wherein each query stage represents a query to a database to collect information stored in the database; and
presenting a query stage element for each of the plurality of query stages in the workflow section of the application workflow user interface.
4 . The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise:
receiving a plurality of query stages through the application workflow user interface, wherein each query stage is a stage in the workflow for the application and wherein each query stage represents a query to a database to collect information stored in the database; and
presenting a query stage element for each of the plurality of query stages in the workflow section of the application workflow user interface.
5 . The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise:
receiving, through the application workflow user interface, an indication of a connection between a first prompt stage and a second prompt stage of the plurality of prompt stages, wherein the connection indicates that an output of the first prompt stage is an input to the second prompt stage; and
presenting a connection element corresponding to the connection in the workflow section of the application workflow user interface, wherein the connection element is a user interface element representing the connection between the first prompt stage and the second prompt stage.
6 . The non-transitory computer-readable medium of claim 1 , wherein the large language model is a machine-learning model executed on the online system or on a third-party system.
7 . The non-transitory computer-readable medium of claim 1 , the operations further comprising:
receiving input instructions and output instructions for each prompt stage element of the plurality of prompt stage elements through the parameters section of the application workflow user interface.
8 . The non-transitory computer-readable medium of claim 7 , wherein the output instructions for a prompt stage element comprise computer-readable instructions for generating a data structure storing information from a response from the large language model, wherein the data structure is generated for use in a subsequent stage in the workflow.
9 . The non-transitory computer-readable medium of claim 8 , wherein the data structure is a JSON file.
10 . The non-transitory computer-readable medium of claim 7 , wherein the input instructions for a prompt stage element comprise computer-readable instructions for extracting information from a data structure for inclusion in a prompt corresponding to the prompt stage element.
11 . The non-transitory computer-readable medium of claim 1 , wherein the operations further comprise:
receiving, through the parameters section, a selection of which large language model to use for each prompt stage of the plurality of prompt stages.
12 . The non-transitory computer-readable medium of claim 1 , wherein executing the computer-executable instructions on the online system comprises:
receiving an input from the user through a testing section of the application workflow user interface;
executing the computer-executable instructions for the application on the input; and
providing an output of the executed computer-executable instructions to the user through the testing section.
13 . The non-transitory computer-readable medium of claim 12 , wherein executing the computer-executable instructions comprises:
continually identifying a current stage of the application workflow as the workflow is executed through the computer-executable instructions; and
continually modifying the workflow section to identify the prompt stage element corresponding to the current stage of the workflow.
14 . A method comprising:
receiving a plurality of prompt stages through an application workflow user interface presented on a client device, wherein each prompt stage is a stage in a workflow for an application executing on an online system and wherein each prompt stage represents a prompt to be transmitted to a large language model;
presenting a prompt stage element for each of the plurality of prompt stages in a workflow section of the application workflow user interface, wherein each prompt stage element is a user interface element representing the corresponding prompt stage element in the workflow section of the application workflow user interface;
receiving prompt text for each prompt stage element through a parameters section of the application workflow user interface, wherein the prompt text for a prompt stage element comprises text for the prompt for the corresponding prompt stage;
translating the plurality of prompt stage elements into computer-executable instructions to execute the workflow of the application on the online system; and
executing the computer-executable instructions on the online system.
15 . The method of claim 14 , wherein the application is a chatbot application.
16 . The method of claim 14 , further comprising:
receiving a plurality of query stages through the application workflow user interface, wherein each query stage is a stage in the workflow for the application and wherein each query stage represents a query to a database to collect information stored in the database; and
presenting a query stage element for each of the plurality of query stages in the workflow section of the application workflow user interface.
17 . The method of claim 14 , further comprising:
receiving a plurality of query stages through the application workflow user interface, wherein each query stage is a stage in the workflow for the application and wherein each query stage represents a query to a database to collect information stored in the database; and
presenting a query stage element for each of the plurality of query stages in the workflow section of the application workflow user interface.
18 . The method of claim 14 , further comprising:
receiving, through the application workflow user interface, an indication of a connection between a first prompt stage and a second prompt stage of the plurality of prompt stages, wherein the connection indicates that an output of the first prompt stage is an input to the second prompt stage; and
presenting a connection element corresponding to the connection in the workflow section of the application workflow user interface, wherein the connection element is a user interface element representing the connection between the first prompt stage and the second prompt stage.
19 . The method of claim 14 , wherein the large language model is a machine-learning model executed on the online system or on a third-party system.
20 . The method of claim 14 , further comprising:
receiving input instructions and output instructions for each prompt stage element of the plurality of prompt stage elements through the parameters section of the application workflow user interface.