AI Root Cause Analysis Using Knowledge Graphs for Build Logs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing software root cause analysis methods are inefficient and prone to errors due to the extensive size of logs exceeding language model prompts, format changes causing failures, and hallucinations leading to incorrect problem investigation.

Innovation Solution

A multi-model approach using a Small Graphical Language Model (SGLM) and a Large Language Model (LLM) to analyze build logs, generating a knowledge graph, extracting signatures, and constructing prompts for the LLM to accurately predict root causes, while the SGLM checks for hallucinations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If build logs are fed directly to a large language model for analysis, then the analysis can be performed, but the extensive size of the logs exceeds the prompt size limit

Engineering Contradiction:

Improveautomated root cause analysis capabilityVSAvoidlog data truncation

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The patent segments the build logs into multiple smaller chunks that can fit within the prompt size limit. The knowledge graph is divided into multiple sub-graphs, each containing a subset of nodes and edges, which are then processed separately by the language model. This segmentation allows the model to analyze large volumes of log data without exceeding prompt size constraints.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent extracts only the most relevant information from the build logs by generating a knowledge graph that filters and prioritizes critical entities and relationships. The extraction process identifies and isolates key diagnostic information, separating it from redundant log data, thereby enabling effective analysis within limited prompt sizes.

Inventive Principle:

Principle #2Taking out (Extraction)

2Adaptability or versatility

If the prompt structure or log structure changes, then the system adapts to new formats, but the automated analysis fails due to sensitivity to structural variations

Engineering Contradiction:

Improvehandling of log format changesVSAvoidanalysis consistency

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The knowledge graph serves as an intermediary representation that translates diverse log formats into a unified structural framework. This intermediate layer abstracts away format variations, allowing the system to handle different log structures consistently. The knowledge graph acts as a mediator between the variable input formats and the fixed analysis methodology, ensuring reliable processing across format changes.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Extent of automation

If a language model is used to analyze build logs, then root cause analysis can be automated, but the model may hallucinate false root causes leading to wasted investigation time

Engineering Contradiction:

Improveroot cause analysis automationVSAvoidaccuracy of root cause identification

Core Design Contradiction:

Extent of automationVSReliability

Solution Approach 1:

The patent implements feedback mechanisms where the knowledge graph continuously validates the language model's analysis outputs. The structured representation of log data provides factual feedback that corrects hallucinations, ensuring that only accurate root causes are identified. This feedback loop between the knowledge graph and language model maintains high reliability in automated analysis.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The knowledge graph is generated in advance as a comprehensive structured representation of the build logs before the language model performs analysis. This preliminary structuring establishes a factual foundation that guides the analysis process, preventing hallucinations by ensuring the model operates on verified information rather than unstructured data.

Inventive Principle:

Principle #10Preliminary action

4Measurement precision

If manual review of build logs is performed, then accurate root cause identification is possible, but the process is time-consuming and impractical for extensive logs

Engineering Contradiction:

Improveroot cause identification accuracyVSAvoidmanual analysis time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces the mechanical process of manual log review with an automated system combining knowledge graph generation and language model analysis. This substitution transforms the time-consuming manual inspection into an automated computational process that achieves comparable or superior accuracy while dramatically reducing analysis time.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12474921B2Multi-modal artificial intelligence root cause analysis

Publication Date: 2025.11.18 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

A data processing system implements obtaining build logs that include information associated with a software build problem; analyzing the logs to generate a knowledge graph identifying the relationship between various entities in the logs; extracting a signature of a candidate root cause of the build problem from the knowledge graph representing a subset of nodes and edges of the knowledge graph; providing the signature of the candidate root cause to a graphical language model to obtain a prediction of a category of root cause failure selected from among a plurality of root cause failures; constructing a prompt for a language model to generate a root cause failure analysis that describes the root cause of the build problem, the prompt including the category of root cause; receiving the root cause failure analysis from the language model; and performing one or more actions in response to receiving the root cause failure analysis.