Granted Patent B1

US 12,481,411 · App. 19/196,582 · Granted Nov 25, 2025

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

Assignee: Dropbox, Inc.

G06F3/04817G06F40/166

Abstract

Claims (97)

1 . A computer-implemented method comprising:

generating, for display overlaid on a first webpage within a web browser, a floating widget comprising a large language model (LLM) function button corresponding to first content of the first webpage overlaid by the floating widget;

updating, in response to detecting a navigational input within the web browser navigating from the first webpage to a second webpage comprising second content different from the first content, the floating widget to overlay the second content of the second webpage;

modifying, in response to detecting the floating widget overlaying the second content, the floating widget to include a different LLM function button corresponding to the second content of the second webpage;

generating, based on detecting an interaction with the LLM function button, a prompt for an LLM associated with the LLM function button, the prompt comprising an instruction for the LLM to complete a function of the LLM function button; and

providing the instruction and the first content of the first webpage to the LLM to cause the LLM to complete a function according to the prompt that includes the instruction and the first content.

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

generating the floating widget to display a first subset of a plurality of LLM function buttons; and

modifying the floating widget to display a second subset of the plurality of LLM function buttons in response to detecting the navigational input.

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

responsive to receiving an indication of a selection of a summarize function button from among the plurality of LLM function buttons:

extracting current content of a current webpage overlaid by the floating widget at selection of the summarize function button;

generating a summary prompt comprising the current content and instructions to summarize the current content;

generating, based on providing the summary prompt to an LLM, a summary of the current content; and

providing the summary within a side panel of the web browser.

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

responsive to receiving an indication of a selection of a chat function button from among the plurality of LLM function buttons:

instantiating a chat interface in a side panel of the web browser;

extracting current content of a current webpage overlaid by the floating widget at selection of the chat function button;

generating, in response to a query provided via the chat interface, a chat prompt comprising the current content, the query, and an instruction to generate a response to the query from the current content; and

generating the response to the query for display in the chat interface by providing the chat prompt to an LLM.

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

responsive to receiving an indication of a selection of a tone function button from among the plurality of LLM function buttons:

extracting current content of a current webpage overlaid by the floating widget at selection of the tone function button;

generating a tone prompt comprising the current content and an instruction to generate tone-specific composed text according to a tone associated with the current content of the current webpage;

generating, based on providing the tone prompt to an LLM, the tone-specific composed text; and

providing the tone-specific composed text for display within a side panel of the web browser.

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

responsive to receiving an indication of a selection of a template function button from among the plurality of LLM function buttons:

extracting current content of a current webpage overlaid by the floating widget at selection of the template function button;

generating a template prompt comprising the current content and instructions to generate a templated content item according to the current content within the current webpage;

generating, based on providing the template prompt to an LLM, the templated content item; and

providing the templated content item within a side panel of the web browser.

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

responsive to receiving an indication of a selection of an autonomous workflow function button from among the plurality of LLM function buttons:

extracting current content of a current webpage overlaid by the floating widget at selection of the autonomous workflow function button;

generating an autonomous workflow prompt comprising the current content and instructions to execute an autonomous workflow based on the current content; and

executing the autonomous workflow by providing the autonomous workflow prompt to an LLM for modifying one or more fields within the current webpage.

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:

generate, for display overlaid on a first webpage within a web browser, a floating widget comprising a first subset of a plurality of large language model (LLM) function buttons corresponding to first content of the first webpage overlaid by the floating widget;

update, in response to detecting a navigational input within the web browser navigating from the first webpage to a second webpage comprising second content different from the first content, the floating widget to overlay the second content of the second webpage;

modify, in response to detecting the floating widget overlaying the second content, the floating widget to include a second subset of the plurality of LLM function buttons corresponding to the second content of the second webpage;

generate, based on detecting an interaction with an LLM function button of the first subset of LLM function buttons, a prompt for an LLM associated with the LLM function button, the prompt comprising an instruction for the LLM to complete a function of the LLM function button; and

provide the instruction and the first content of the first webpage to the LLM to cause the LLM to complete a function according to the prompt that includes the instruction and the first content.

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

determine a context set relevant to the second content of the second webpage, the context set comprising a plurality of webpages open in respective tabs of the web browser; and

modify the floating widget to include the second subset of the plurality of LLM function buttons based on the context set.

10 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to, responsive to receiving an indication of a selection of a summarize function button from among the first subset of the plurality of LLM function buttons:

extract current content of a current webpage overlaid by the floating widget at selection of the summarize function button;

generate a summary prompt comprising the current content and instructions to summarize the current content;

generate, based on providing the summary prompt to an LLM, a summary of the current content; and

provide the summary within a side panel of the web browser.

11 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to, responsive to receiving an indication of a selection of a chat function button from among the first subset of the plurality of LLM function buttons:

instantiate a chat interface in a side panel of the web browser;

extract current content of a current webpage overlaid by the floating widget at selection of the chat function button;

generate, in response to a query provided via the chat interface, a chat prompt comprising the current content, the query, and an instruction to generate a response to the query from the current content; and

generate the response to the query for display in the chat interface by providing the chat prompt to an LLM.

12 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to, responsive to receiving an indication of a selection of a tone function button from among the plurality of LLM function buttons:

extract current content of a current webpage overlaid by the floating widget at selection of the tone function button;

generate a tone prompt comprising the current content and an instruction to generate tone-specific composed text according to a tone indication within the current content of the current webpage;

generate, based on providing the tone prompt to an LLM, the tone-specific composed text; and

provide the tone-specific composed text within a side panel of the web browser.

13 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to, responsive to receiving an indication of a selection of a template function button from among the plurality of LLM function buttons:

extract current content of a current webpage overlaid by the floating widget at selection of the template function button;

generate a template prompt comprising the current content and instructions to generate a templated content item according to the current content within the current webpage;

generate, based on providing the template prompt to an LLM, the templated content item; and

provide the templated content item within a side panel of the web browser.

14 . The system of claim 8 , further comprising instructions that, when executed by the at least one processor, cause the system to embed, based on current content of a current webpage, at least one LLM function button of the first subset of the plurality of LLM function buttons in the current webpage.

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

generate, for display overlaid on a first webpage within a web browser, a floating widget comprising a large language model (LLM) function button corresponding to first content of the first webpage overlaid by the floating widget;

update, in response to detecting a navigational input within the web browser navigating from the first webpage to a second webpage comprising second content different from the first content, the floating widget to overlay the second content of the second webpage;

modify, in response to detecting the floating widget overlaying the second content, the floating widget to include a different LLM function button corresponding to the second content of the second webpage;

generate, based on detecting an interaction with the LLM function button, a prompt for an LLM associated with the LLM function button, the prompt comprising an instruction for the LLM to complete a function of the LLM function button; and

provide the instruction and the first content of the first webpage to the LLM to cause the LLM to complete the function according to the prompt that includes the instruction and the first content.

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:

generate, based on detecting the interaction with the LLM function button, a side panel adjacent to the first webpage within the web browser; and

provide, for display in the side panel, a description of the LLM function button.

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:

generate the floating widget to display a first subset of a plurality of LLM function buttons; and

modify the floating widget to display a second subset of the plurality of LLM function buttons in response to detecting the navigational input.

18 . The non-transitory computer-readable medium of claim 17 , further comprising instructions that, when executed by the at least one processor, cause the computing device to, based on determining that the LLM function button is a summarize function button:

extract current content of a current webpage overlaid by the floating widget at selection of the summarize function button;

generate a summary prompt comprising the current content and instructions to summarize the current content;

generate, based on providing the summary prompt to an LLM, a summary of the current content; and

provide the summary within a side panel of the web browser.

19 . The non-transitory computer-readable medium of claim 17 , further comprising instructions that, when executed by the at least one processor, cause the computing device to, based on determining that the LLM function button is a chat function button:

instantiate a chat interface in a side panel of the web browser;

extract current content of a current webpage overlaid by the floating widget at selection of the chat function button;

generate, in response to a query provided via the chat interface, a chat prompt comprising the current content, the query, and an instruction to generate a response to the query from the current content; and

generate the response to the query for display in the chat interface by providing the chat prompt to an LLM.

20 . The non-transitory computer-readable medium of claim 17 , further comprising instructions that, when executed by the at least one processor, cause the computing device to, based on determining that the LLM function button is a template function button:

extract current content of a current webpage overlaid by the floating widget at selection of the template function button;

generate a template prompt comprising the current content and instructions to generate a templated content item according to the current content within the current webpage;

generate, based on providing the template prompt to an LLM, the templated content item; and

provide the templated content item within a side panel of the web browser.

Assignments (2)

Continuity (1)

References Cited (36)