Granted Patent B2

US 12,487,819 · App. 18/379,872 · Granted Dec 2, 2025

Inventors: Neelakantan Sundaresan (Bellevue, WA); Alexey Svyatkovskiy (Bellevue, WA); Michele Tufano (Bellevue, WA); Shuo Zhang (Issaquah, WA)

Assignee: Microsoft Technology Licensing, LLC

G06F8/77G06F16/24578G06N3/0455G06N3/0475

Abstract

Claims (61)

1 . A system comprising:

a processor; and

a memory that stores at least one program that is configured to be executed by the processor, the at least one program includes instructions to perform actions that:

obtain a plurality of code changes from a pull request of a repository, wherein a code change represents a change to a file, class or method of the repository;

prioritize the plurality of code changes based on a score that represents an impact of a code change on the file, class or method in the repository;

rank the plurality of code changes based on top-k highest scores;

generate a first prompt to a large language model comprising top-k code changes;

obtain from the large language model, given the first prompt, a summary of the top-k code changes; and

display the summary in a pull request description for the pull request in the repository.

2 . The system of claim 1 , wherein the at least one program includes instructions to perform actions that:

perform code review of a top code change;

receive input on outcome of the code review of the top code change; and

upon successful outcome of the code review of the top code change, merge the code change back to the repository.

3 . The system of claim 1 , wherein the at least one program includes instructions to perform actions that:

obtain from the large language model, given the first prompt, a classification of the top-k code changes.

4 . The system of claim 1 , wherein the at least one program includes instructions to perform actions that:

generate an open issue index comprising a plurality of open issues associated with the repository, wherein the open issue index is accessed through an embedding of an issue.

5 . The system of claim 4 , wherein the at least one program includes instructions to perform actions that:

search the open issue index for an issue related to the summary of the top-k code changes using an embedding of the summary of the top-k code changes.

6 . The system of claim 5 , wherein the at least one program includes instructions to perform actions that:

extract from the retrieved issue at least one code reviewer to review the pull request.

7 . The system of claim 6 , wherein the at least one program includes instructions to perform actions that:

output the summary of the top-k code changes, the classification of the top-k code changes, a related open issue, and the at least one code reviewer into the pull request description of the pull request.

8 . A computer-implemented method, comprising:

retrieving a plurality of code changes from a pull request of a repository;

scoring each code change based on an impact of a code change on files, classes and methods of the repository;

ranking the plurality of code changes based on a respective score;

selecting top-k code changes having highest scores;

generating a prompt to a large language model comprising the top-k code changes having highest scores; and

obtaining from the large language model, given the prompt, a summary of the pull request, wherein the summary of the pull request comprises a description of the top-k code changes.

9 . The computer-implemented method of claim 8 , comprising:

reviewing a top code change;

receiving input on outcome of a review of the top code change; and

upon successful outcome of the review of the top code change, merging the code change back to the repository.

10 . The computer-implemented method of claim 8 , comprising:

obtaining from the large language model, given the prompt, a classification of the top-k code changes.

11 . The computer-implemented method of claim 8 , comprising:

obtaining at least one open issue from the repository associated with the top-k code changes based on the at least one open issue matching the summary of the pull request.

12 . The computer-implemented method of claim 11 , comprising:

extracting from the at least one open issue at least one code reviewer to review the pull request.

13 . The computer-implemented method of claim 12 , comprising:

displaying in a pull request description field of the pull request, the summary of the top-k code changes, a classification of the top-k code changes, the at least one open issue, and the at least one code reviewer.

14 . The computer-implemented method of claim 8 , wherein the large language model is a conversational pre-trained generative neural transformer models with attention.

15 . A hardware storage device having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

generate a dependency graph for a repository, wherein the repository comprises source code files having methods and classes, wherein the dependency graph comprises a plurality of nodes representing the source code files, methods and classes within the repository, wherein an edge between two nodes represents a dependency between the two nodes;

obtaining a pull request comprising a plurality of code changes;

mapping each of the plurality of code changes to a respective node in the dependency graph;

generating a score for each of the plurality of code changes based on a size of connected components of the respective mapped node in the dependency graph and a page rank of the respective mapped node;

selecting top-k code diff hunks based on highest scores;

generate a prompt to a large language model, wherein the prompt comprises top-k code changes;

obtain from the large language model a natural language summarization of the top-k code changes of the pull request; and

output the natural language summarization of the top-k code changes in the pull request.

16 . The hardware device of claim 15 having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

obtain from the large language model a classification of the top-k code changes in the pull request.

17 . The hardware device of claim 16 having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

search for a related open issue of the repository based on a similarity of an embedding of the related open issue to an embedding of the natural language summarization of the pull request.

18 . The hardware device of claim 17 having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

extract at least one code reviewer from the related open issue of the repository.

19 . The hardware device of claim 18 having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

output the classification, related open issue and the at least one code reviewer in the pull request.

20 . The hardware device of claim 15 , wherein the large language model is a conversational pre-trained generative neural transformer models with attention.

Assignments (1)

Continuity (2)

References Cited (22)