Granted Patent B1

US 12,737,180 · App. 19/248,315 · Granted Sep 15, 2026

Inventors: Girish C. Sharma (Pennington, NJ); Biswaranjan Panigrahi (Bangalore, IN); Omaran Bazna (Dearborn, MI); Vivek Sharma (Ghaziabad, IN)

Assignee: Morgan Stanley Services Group Inc.

G06F8/71G06F8/10G06F16/9024G06F16/953

Abstract

Claims (65)

1 . A method of performing context-sensitive application support for software development using artificial intelligence, comprising:

receiving a sequential input of browser events, including a user query input into the browser involving a first network endpoint;

identifying an application type based on the sequential input;

obtaining a configuration prompt using the application type based on configuration data specifying a graph in a graph database and a schema for the graph database,

each node in the graph represents a network endpoint of a plurality of network endpoints, and each edge in the graph represents a relationship between two network endpoints;

combining the sequential input and the configuration prompt, producing a sequential input prompt;

sending the sequential input prompt to a large language model (LLM) to produce an intermediate data object as a flow embedding vector representing a subgraph connected to a node corresponding to the first network endpoint;

receiving a task configuration file describing a file format conforming to a task record data structure format based on the configuration data;

mapping the intermediate data object onto the task configuration file, producing a task congruency determination indicating how a second network endpoint is affected by the first network endpoint; and

producing an output stream based on the task congruency determination within a browser interface,

wherein the method is performed by one or more processors.

2 . The method of claim 1 , wherein:

the intermediate data object includes a graph query, the graph query directed to querying information related to a first node in the graph corresponding to the first network endpoint,

the output stream includes metadata associated with the first node.

3 . The method of claim 2 , wherein:

the output stream further includes a list of nodes which share an edge with the first node, each edge between two nodes in the list of nodes representing a respective network traffic event of a sequence of network traffic events.

4 . The method of claim 2 , further comprising:

receiving feedback from an administrator device; and

updating a network endpoint sequence, the network endpoint sequence describing a plurality of node and edge relationships in the graph database, based on the feedback, producing a revised network endpoint sequence.

5 . The method of claim 2 , wherein:

the sequential input including a list of network traffic events involving one or more network endpoints, and

the method further comprising updating the graph database based on the list of network traffic events.

6 . The method of claim 5 , further comprising:

inserting a certain node into the graph, the certain node representing a backend network endpoint not directly connected with the browser.

7 . The method of claim 1 , further comprising:

receiving browser events related to one or more web applications;

accessing log data of one or more native applications;

building the graph based on the browser events and the log data,

each node in the graph representing a website, webpage, web application, native application, or device,

each edge in the graph representing a relationship between two nodes,

each node and each edge having a weight indicating a frequency.

8 . A non-transitory, computer-readable storage medium storing one or more sequences of instructions which when executed cause one or more processor to perform:

receiving a sequential input of browser events, including a user query input into a browser involving a first network endpoint;

obtaining a configuration prompt based on configuration data specifying a graph in a graph database and a schema for the graph database,

each node in the graph represents a network endpoint of a plurality of network endpoints, and each edge in the graph represents a relationship between two network endpoints;

combining the sequential input and the configuration prompt, producing a sequential input prompt;

sending the sequential input prompt to a large language model (LLM) to produce an intermediate data object as a flow embedding vector representing a subgraph connected to a node corresponding to the first network endpoint;

receiving a task configuration file describing a file format conforming to a task record data structure format based on the configuration data;

mapping the intermediate data object onto the task configuration file, producing a task congruency determination indicating how a second network endpoint is affected by the first network endpoint; and

producing an output stream based on the task congruency determination within a browser interface.

9 . The non-transitory, computer-readable storage medium of claim 8 , the one or more sequences of instructions when executed causing the one or more processor to further perform:

the user query including a description of the first network endpoint; and

wherein mapping the intermediate data object onto the task configuration file includes:

combining the intermediate data object and the user query,

sending the combined intermediate data object and the user query to the LLM to produce a second intermediate data object based on the intermediate data object, and

mapping the second intermediate data object onto the task configuration file, producing the task congruency determination.

10 . The non-transitory, computer-readable storage medium of claim 8 , the one or more sequences of instructions when executed causing the one or more processor to further perform:

the user query including a description of a relationship with the first network endpoint; and

wherein mapping the intermediate data object onto the task configuration file includes:

combining the intermediate data object and the user query,

sending the combined intermediate data object and the user query to the LLM to produce a second intermediate data object based on the intermediate data object, and

mapping the second intermediate data object onto the task configuration file, producing the task congruency determination.

11 . A computer system for performing context-sensitive application support for software development using artificial intelligence, comprising:

a memory;

one or more processors coupled to the memory and configured to perform:

receiving a sequential input of browser events, including a user query input into a browser involving a first network endpoint;

obtaining a configuration prompt based on configuration data specifying a graph in a graph database and a schema for the graph database,

each node in the graph represents a network endpoint of a plurality of network endpoints, and each edge in the graph represents a relationship between two network endpoints;

combining the sequential input and the configuration prompt, producing a sequential input prompt;

sending the sequential input prompt to a large language model (LLM) to produce an intermediate data object as a flow embedding vector representing a subgraph connected to a node corresponding to the first network endpoint;

receiving a task configuration file describing a file format conforming to a task record data structure format based on the configuration data;

mapping the intermediate data object onto the task configuration file, producing a task congruency determination indicating how a second network endpoint is affected by the first network endpoint; and

producing an output stream based on the task congruency determination within a browser interface.

12 . The non-transitory, computer-readable storage medium of claim 8 ,

wherein the sequential input and the configuration prompt are further combined with the user query, producing the sequential input prompt.

Assignments (1)

References Cited (28)