# US 2026/0029997 A1 · Decentralized architecture using artificial intelligence driven autonomous self-healing of distributed software

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0029997 A1
- Document ID: US20260029997A1
- Status: Published application
- Kind: A1
- Application number: 18/784,509
- Filed: 2024-07-25
- Published: 2026-01-29
- Earliest claimed: 2024-07-25
- Assignees: American Express Travel Related Services Company, Inc.
- Applicants: American Express Travel Related Services Company, Inc.
- Inventors: Andras L. FERENCZI; Pedro Burglin PAES; Alaric M. EBY; Aniesh CHAWLA; Nithin Kumar Ullal KULAPURAM; Joseph GYAWALI
- Web page: https://patentgrep.com/patent/US20260029997A1

## CPC

- G06F 8/35: model driven
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Disclosed herein are system, method, and computer program product embodiments for autonomously repairing software by leveraging a large language model (LLM). A control system may detect a first error associated with an application executing in a region. The control system may then repair the first error associated with the application by: identifying a source of the first error within the application; generating a solution by inputting the source of the first error to an LLM; and implementing the solution via the LLM. The control system may then determine that the application is repaired by: executing the application; generating an output; and comparing the output to a predefined value. The control system may then deploy the application in the region.

## Claims

### Claim 1

A computer implemented method for autonomous software repair, the method comprising:
detecting a first error associated with an application executing in a region;
repairing the first error associated with the application comprising:
identifying a source of the first error within the application;
generating a solution by inputting the source of the first error to a large language model (LLM); and
implementing the solution via the LLM; and
determining that the application is repaired by:
executing the application;
generating, by the application, an output; and
comparing the output to a predefined value; and
deploying the application in the region in response to determining that the application is repaired.

### Claim 2 (depends on 1)

The computer implemented method of claim 1, wherein identifying the source of the first error comprises identifying an error message within a log file associated with the application.

### Claim 3 (depends on 1)

The computer implemented method of claim 1, wherein identifying the source of the first error comprises determining a telemetry value associated with the application is greater than a predefined threshold.

### Claim 4 (depends on 1)

The computer implemented method of claim 1, wherein the first error is associated with source code of the application and generating the solution further comprises generating new source code by the LLM, wherein the new source code is designed to repair the first error.

### Claim 5 (depends on 1)

The computer implemented method of claim 1, wherein the first error is associated with a configuration value of the application and implementing the solution further comprises updating the configuration value.

### Claim 6 (depends on 1)

The computer implemented method of claim 1, wherein generating the solution further comprises:
generating, by the LLM, a summary of the first error;
converting the summary to a summary vector;
calculating a similarity value between the summary vector and a stored error vector; and
outputting the solution linked with the stored error vector, wherein the stored error vector linked to the solution has a highest similarity value to the summary vector.

### Claim 7 (depends on 1)

The computer implemented method of claim 1, further comprising:
detecting a second error associated with a second instance of the application executing in a second region;
determining the first region has a higher priority than the second region; and
in response to the determination, deploying the application to the first region prior to the second region.

### Claim 8 (depends on 1)

The computer implemented method of claim 1, further comprising:
detecting a second error associated with the application; and
repairing the second error before the first error, based on a comparison of an effect of the first error and an effect of the second error on the application.

### Claim 9 (depends on 1)

The computer implemented method of claim 1, wherein the predefined value is at least one of: (i) an expected output defined by a function unit test, (ii) CPU usage, (iii) memory usage, or (iv) network usage.

### Claim 10

A system, comprising:
a memory; and
at least one processor coupled to the memory and configured to:
detect a first error associated with an application executing in a region;
repair the first error associated with the application comprising:
identifying a source of the first error within the application;
generating a solution by inputting the source of the first error to a large language model (LLM); and
implementing the solution via the LLM; and
determine that the application is repaired by:
executing the application;
generating an output; and
comparing the output to a predefined value; and
deploy the application in the region in response to determining that the application is repaired.

### Claim 11 (depends on 10)

The system of claim 10, wherein identifying the source of the first error comprises identifying an error message within a log file associated with the application.

### Claim 12 (depends on 10)

The system of claim 10, wherein identifying the source of the first error comprises determining a telemetry value associated with the application is greater than a predefined threshold.

### Claim 13 (depends on 10)

The system of claim 10, wherein the first error is associated with source code of the application and generating the solution further comprises generating new source code by the LLM, wherein the new source code is designed to repair the first error.

### Claim 14 (depends on 10)

The system of claim 10, wherein the first error is associated with a configuration value of the application and implementing the solution further comprises updating the configuration value.

### Claim 15 (depends on 10)

The system of claim 10, wherein generating the solution further comprises:
generating, by the LLM, a summary of the first error;
converting the summary to a summary vector;
calculating a similarity value between the summary vector and a stored error vector; and
outputting the solution linked with the stored error vector, wherein the stored error vector linked to the solution has a highest similarity value to the summary vector.

### Claim 16 (depends on 10)

The system of claim 10, further comprising:
detecting a second error associated with the application; and
repairing the second error before the first error, based on a comparison of an effect of the first error and an effect of the second error on the application.

### Claim 17

A non-transitory computer-readable device having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations comprising:
detecting a first error associated with an application executing in a region;
repairing the first error associated with the application comprising:
identifying a source of the first error within the application;
generating a solution by inputting the source of the first error to a large language model (LLM); and
implementing the solution via the LLM; and
determining that the application is repaired by:
executing the application;
generating an output; and
comparing the output to a predefined value; and
deploying the application in the region in response to determining that the application is repaired.

### Claim 18 (depends on 17)

The non-transitory computer-readable device of claim 17, wherein the first error is associated with source code of the application and generating the solution further comprises generating new source code by the LLM, wherein the new source code is designed to repair the first error.

### Claim 19 (depends on 17)

The non-transitory computer-readable device of claim 17, wherein identifying the source of the first error comprises determining a telemetry value associated with the application is greater than a predefined threshold.

### Claim 20 (depends on 17)

The non-transitory computer-readable device of claim 17, wherein generating the solution further comprises:
generating, by the LLM, a summary of the first error;
converting the summary to a summary vector;
calculating a similarity value between the summary vector and a stored error vector; and
outputting the solution linked with the stored error vector, wherein the stored error vector linked to the solution has a highest similarity value to the summary vector.
