Granted Patent B1

US 12,541,400 · App. 19/196,588 · Granted Feb 3, 2026

Inventors: Hiten Shah (Redwood City, CA); Richard Chan (Flushing, NY); Theo Richardson (Toronto, CA)

Assignee: Dropbox, Inc.

G06F9/5027G06F16/3329G06F16/345

Abstract

Claims (74)

1 . A computer-implemented method comprising:

determining a repeated access pattern and a repeated action pattern by a user account within a webpage of a web browser displayed on a client device;

receiving, from the webpage, an indication of selecting a large language model (LLM) function button for executing a series of functions related to content within the webpage;

generating a recording request that indicates an autonomously executable function of the series of functions that corresponds to the repeated access pattern and the repeated action pattern;

in response to receiving a confirmation of the recording request, recording a series of user account interactions relating to the webpage;

extracting a workflow comprising a plurality of steps from the series of user account interactions;

determining the autonomously executable function based on the plurality of steps;

in response to the indication of selecting the LLM function button, providing, within a function execution panel, a preliminary description of an automated execution of the autonomously executable function of the series of functions to be performed by an LLM;

requesting, via the function execution panel, a confirmation of the autonomously executable function of the series of functions to be executed;

based on receiving the confirmation of the autonomously executable function, executing the autonomously executable function within the series of functions by generating a function prompt comprising the content within the webpage and further comprising LLM instructions to execute the autonomously executable function based on the content within the webpage; and

providing, for display within the web browser on the client device, a function execution panel presenting a function ticker depicting a description of how the automated execution of the autonomously executable function was performed by an LLM using the function prompt.

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

determining the autonomously executable function within the series of functions according to the content within the webpage.

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

generating the preliminary description of the automated execution of the autonomously executable function by using the function ticker to generate a natural language description of the autonomously executable function.

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

receiving, via a chat interface of the function execution panel, a modification to the preliminary description of the automated execution of the autonomously executable function; and

executing the autonomously executable function according to the modification based on receiving an approval of a modified preliminary description of the automated execution of the autonomously executable function via the function execution panel.

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

displaying the LLM function button in a floating widget within the webpage.

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

instantiating the function execution panel based on determining an indication of an interaction with the LLM function button within the floating widget.

7 . The computer-implemented method of claim 1 , wherein executing the autonomously executable function comprises performing one or more of: modifying a field of the webpage, mimicking a navigational input, or extracting data from a content item.

8 . A system comprising:

at least one processor; and

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

determine a repeated access pattern and a repeated action pattern by a user account within a webpage of a web browser displayed on a client device;

receive, from the webpage, an indication of selecting a large language model (LLM) function button for executing a series of functions related to content within the webpage;

generate a recording request that indicates an autonomously executable function of the series of functions that corresponds to the repeated access pattern and the repeated action pattern;

in response to receiving a confirmation of the recording request, record a series of user account interactions relating to the webpage;

extract a workflow comprising a plurality of steps from the series of suer account interactions;

determine the autonomously executable function based on the plurality of steps;

in response to the indication of selecting the LLM function button, provide, within a function execution panel, a preliminary description of an automated execution of autonomously executable function of the series of functions to be performed by an LLM;

request, via the function execution panel, a confirmation of the autonomously executable function of the series of functions to be executed;

based on receiving the confirmation of the autonomously executable function, execute the autonomously executable function within the series of functions by generating a function prompt comprising the content within the webpage and further comprising LLM instructions to execute the autonomously executable function based on the content within the webpage; and

provide, for display within the web browser on the client device, a function execution panel presenting a function ticker depicting a description of how the automated execution of the autonomously executable function was performed by an LLM using the function prompt.

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

perform a first function comprising navigating to a field of the webpage;

perform a second function to extract data that is relevant to the field of the webpage from a content item hosted within a content management system; and

perform a third function comprising modifying the field of the webpage by inputting the data.

10 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to perform the autonomously executable function by executing a call to an application processing interface (API).

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

generate the preliminary description of the automated execution of the autonomously executable function by using the function ticker to generate a natural language description of the autonomously executable function.

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

receive, via a chat interface of the function execution panel, a modification to the preliminary description of the automated execution of the autonomously executable function; and

execute the autonomously executable function according to the modification based on receiving an approval of a modified preliminary description of the automated execution of the autonomously executable function via the function execution panel.

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

display the LLM function button in a floating widget within the webpage.

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

instantiate the function execution panel based on determining an indication of an interaction with the LLM function button within the floating widget.

15 . A non-transitory computer-readable medium storing instructions thereon that, when executed by at least one processor, cause a computing device to:

determine a repeated access pattern and a repeated action pattern by a user account within a webpage of a web browser displayed on a client device;

receive, from the client device, an indication of selecting a large language model (LLM) function button for executing a series of functions related to content within the webpage;

generate a recording request that indicates an autonomously executable series of functions that corresponds to the repeated access pattern and the repeated action pattern;

in response to receiving a confirmation of the recording request, record a series of user account interactions relating to the webpage;

extract a workflow comprising a plurality of steps from the series of user account interactions;

determine the autonomously executable series of functions based on the plurality of steps;

in response to the indication of selecting the LLM function button, provide, within a function execution panel, a preliminary description of an automated execution of the autonomously executable the series of functions to be performed by an LLM;

request, via the function execution panel, a confirmation of the autonomously executable series of functions to be executed;

based on receiving the confirmation of the autonomously executable series of functions, execute the autonomously executable series of functions by generating a function prompt comprising the content within the webpage and further comprising LLM instructions to execute the autonomously executable series of functions based on the content within the webpage; and

provide, for display within the web browser on the client device, a function execution panel presenting a function ticker depicting a description of how the automated execution of the autonomously executable series of functions was performed by an LLM using the function prompt.

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

receive a modification to the description of how the automated execution of the autonomously executable series of functions was performed by the LLM;

generate a modified function prompt according to the modification to the description;

execute the autonomously executable series of functions according to the modified function prompt; and

provide, for display within the web browser on the client device, the function execution panel presenting the function ticker depicting a modified description of how the automated execution of the autonomously executable series of functions was performed according to the modified function prompt.

17 . The non-transitory computer-readable medium of claim 15 , further comprising instructions that, when executed by the at least one processor, cause the computing device to perform the autonomously executable series of functions by executing a subroutine defined by computer code embedded in the LLM function button.

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

generate the preliminary description of the automated execution of the autonomously executable series of functions by using the function ticker to generate a natural language description of the series of functions.

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

receive, via a chat interface of the function execution panel, a modification to the preliminary description of the automated execution of the autonomously executable series of functions; and

execute the series of functions according to the modification based on receiving an approval of a modified preliminary description of the automated execution of the autonomously executable series of functions via the function execution panel.

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

display the LLM function button in a floating widget within the webpage.

Assignments (2)

Continuity (1)

References Cited (35)