IP Library Granted Patent US 12670405
Granted Patent B2
US 12670405 · App. 18/976,068 · Granted Jun 30, 2026

Generating a model application comprising discrete model functions

Inventors: Richard Chan (Flushing, NY); Jennifer Gradone (San Francisco, CA); Theo Richardson (Toronto, CA); Hiten Shah (Redwood City, CA)
Assignee: Dropbox, Inc.
G06N3/0895
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12670405
App. No.
18/976,068
Granted
Jun 30, 2026
Kind
B2
Abstract

The present disclosure is directed toward systems, methods, and non-transitory computer readable media for generating a model application comprising one or more discrete model functions classified into a function category comprising a sense category, a reason category, or an act category. In some embodiments, the disclosed systems can combine a discrete model function with one or more additional model functions to generate a model application that defines data processing for a customized instance of a large language model. The disclosed systems can surface a model interaction interface comprising selectable application elements to instantiate respective applications of a large language model and modify the model interaction interface to surface a recommended source content item within a source selection window to utilize with the large language model.

Claims (94)

1 . A computer-implemented method comprising:

generating a customized user group context for a group of defined user accounts based on one or more source content items associated with the group of defined user accounts and monitoring user interaction data of the group of defined user accounts with the one or more source content items associated with the group of defined user accounts;

determining discrete model functions that are executable by a large language model and combinable with one another to form combined model functions;

classifying a discrete model function from the discrete model functions into a function category comprising one or more of a sense category, a reason category, or an act category;

combining the discrete model function with one or more additional model functions;

generating, in response to combining the discrete model function with the one or more additional model functions, a model application defining data processing for a customized instance of the large language model;

integrating the discrete model function with one or more autonomous agents based on one or more functions performed by the discrete model function;

determining one or more capabilities of the model application based on one or more functions performed by the discrete model function;

performing one or more tasks for the group of defined user accounts with the one or more autonomous agents integrated with the model application and the customized instance of the large language model according to the customized user group context based on the customized user group context corresponding to the one or more capabilities of the model application; and

based on detecting a change to the customized user group context, generating a modified model application by replacing the discrete model function with an additional discrete model function that shares the function category.

2 . The computer-implemented method of claim 1 , further comprising:

classifying the discrete model function into the sense category;

receiving, utilizing the discrete model function classified into the sense category, one or more user interactions with one or more content items within a content management system or third-party server linked to the content management system via one or more software connectors; and

ingesting, via the discrete model function classified into the sense category, the one or more content items based on the one or more user interactions with the one or more content items.

3 . The computer-implemented method of claim 1 , further comprising:

classifying the discrete model function into the reason category;

generating, utilizing the discrete model function classified into the reason category, one or more signals based on one or more content items or one or more user interactions with the one or more content items; and

generating, based on the one or more signals, one or more tasks to perform within the model application.

4 . The computer-implemented method of claim 1 , wherein the discrete model functions classified into the act category further comprises:

classifying the discrete model function into the act category; and

performing, utilizing the discrete model function classified into the act category, one or more tasks within the model application, based on one or more signals generated by the discrete model function classified into the reason category.

5 . The computer-implemented method of claim 1 , further comprising:

receiving, from a client device associated with a user account, one or more user interactions with the model application;

determining, based the one or more user interactions, a user context associated with the user account; and

performing one or more tasks according to the user context associated with the user account.

6 . The computer-implemented method of claim 1 , further comprising:

receiving, from a client device associated with a user account, one or more user interactions with the model application;

generating, for one or more user accounts, a historic log comprising the one or more user interactions, and one or more tasks performed by the model application; and

in response to receiving a request to access the one or more tasks performed by the model application from a client device associated with the one or more user accounts, providing the one or more tasks for display on the client device.

7 . The computer-implemented method of claim 1 , further comprising:

receiving, from a client device associated with a user account, one or more user interactions with one or more source content items or the model application;

determining based on the one or more user interactions, a target audience associated with an output of the model application;

determining an audience context based on the target audience, by accessing one or more source content items or historic outputs associated with the target audience; and

generating a targeted output specific to the target audience, based on the model application utilizing the audience context.

8 . A system comprising:

at least one processor; and

a non-transitory computer readable medium comprising instructions that, when executed by the at least one processor, cause the system to:

generate a customized user group context for a group of defined user accounts based on one or more source content items associated with the group of defined user accounts and monitoring user interaction data of the group of defined user accounts with the one or more source content items associated with the group of defined user accounts;

generate a default model application comprising default discrete model functions that are executable by a large language model, wherein the default discrete model functions correspond to at least one of a function category comprising a sense category, a reason category, or an act category;

determine discrete model functions that are executable by the large language model and combinable with the default discrete model functions;

classify a discrete model function from the discrete model functions into a function category comprising one or more of the sense category, the reason category, or the act category;

combine the default discrete model functions and the discrete model functions;

generate, in response to combining the default discrete model functions with the discrete model functions, a model application defining data processing for a customized instance of the large language model;

integrate the discrete model function with one or more autonomous agents based on one or more functions performed by the discrete model function;

determine one or more capabilities of the model application based on one or more functions performed by the discrete model function;

perform one or more tasks for the group of defined user accounts with the one or more autonomous agents integrated with the model application and the customized instance of the large language model according to the customized user group context based on the customized user group context corresponding to the one or more capabilities of the model application; and

based on detecting a change to the customized user group context, generate a modified model application by replacing the discrete model function with an additional discrete model function that shares the function category.

9 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

replace a default discrete model function with a discrete model function that corresponds to the function category of the default discrete model function.

10 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

monitor, from a client device associated with a user account, one or more user interactions with one or more content items within a content management system or on a third-party server linked to the content management system via one or more software connectors as dictated by the discrete model function classified into the sense category;

determine a user context for the user account based on the one or more user interactions; and

perform a task with the customized instance of the large language model according to the user context associated with the user account.

11 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

detect, from a set of client devices associated with a set of user accounts, one or more user interactions with one or more content items within a content management system or on a third-party server linked to the content management system via one or more software connectors according to a default discrete model function classified into the sense category; and

analyze at least the one or more content items or the one or more user interactions with the one or more content items according to a discrete model function classified in the reason category.

12 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

process one or more content items within a content management system or on a third-party server linked to the content management system via one or more software connectors according to a default discrete model function or a discrete model function classified into the sense category;

generate, one or more signals associated with a group of user accounts according to a default discrete model function or a discrete model function classified into the reason category; and

perform, one or more tasks based on the one or more signals associated with the group of user accounts according to at least a default discrete model function or a discrete model function classified into the act category.

13 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

perform, via at least a default discrete model function or a discrete model function classified into the act category one or more tasks according to the customized user group context associated with the group of defined user accounts.

14 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to:

perform one or more tasks based on receiving one or more user interactions with the model application from one or more client devices associated with a group of user accounts;

generate, for the group of user accounts, a historic group log comprising the one or more tasks performed by the model application; and

in response to receiving a request to access a task from the one or more tasks stored in the historic group log, provide the task for display on a client device associated with the group of user accounts.

15 . A non-transitory computer readable medium comprising instructions that, when executed by at least one processor, cause the at least one processor to:

generate a customized user group context for a group of defined user accounts based on one or more source content items associated with the group of defined user accounts and monitoring user interaction data of the group of defined user accounts with the one or more source content items associated with the group of defined user accounts;

determine discrete model functions that are executable by a large language model and combinable with one another to form combined model functions;

classify the discrete model functions into one or more function categories comprising one or more of a sense category, a reason category, or an act category;

combine a discrete model function from the discrete model functions with one or more additional discrete model functions classified into a different function category;

generate, in response to combining the discrete model function with the one or more additional discrete model functions classified into the different function category, a model application defining data processing for a customized instance of the large language model;

integrate the discrete model function with one or more autonomous agents based on one or more functions performed by the discrete model function;

determine one or more capabilities of the model application based on one or more functions performed by the discrete model function;

perform one or more tasks for the group of defined user accounts with the one or more autonomous agents integrated with the model application and the customized instance of the large language model according to the customized user group context based on the customized user group context corresponding to the one or more capabilities of the model application; and

based on detecting a change to the customized user group context, generate a modified model application by replacing the discrete model function with an additional discrete model function that shares a function category.

16 . The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to:

determine that the model application comprises a first discrete model function classified into the sense category, a second discrete model function classified into the reason category, and a third discrete model function classified into the act category.

17 . The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to:

receive, from a client device associated with a user account, one or more user interactions with one or more content items within a content management system or third-party server linked to the content management system via one or more software connectors;

determine a model application context based on the one or more user interactions; and

select, based on the model application context, a first discrete model function classified into the sense category, a second discrete model function classified into the reason category, and a third discrete model function classified into the act category.

18 . The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to:

receive, from a client device associated with a user account, one or more user interactions with the model application;

determine at least a tone, style, or voice associated with the user account based on the one or more user interactions; and

perform a task with the model application according to at least the tone, style, or voice associated with the user account.

19 . The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to:

detect, one or more user interactions with one or more content items within a content management system or third-party server linked to the content management system via one or more software connectors according to a first discrete model function classified into the sense category;

generate one or more signals to automate one or more tasks associated with the one or more content items based on the one or more user interactions according to a second discrete model function classified into the reason category; and

perform a task based on the one or more signals associated with the one or more content items, according to a third discrete model function classified into the act category.

20 . The non-transitory computer readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the at least one processor to:

perform one or more tasks based on receiving from a client device associated with a user account one or more user interactions with the model application;

generate, for the user account, a historic log comprising the one or more tasks performed by the model application; and

in response to receiving from the client device associated with the user account, a request to access a task from the one or more tasks stored in the historic log, provide the task for display on the client device associated with the user account.