AGENTS

Jira Defect Analysis Agent

Jira-URL-entry multi-modal defect analysis combining logs, video, screenshots and RAG knowledge.

Core Pain Points

1. Jira information and on-site evidence are scattered: Defect URLs, logs, videos, screenshots, and problem descriptions live in different locations; humans repeatedly open and consolidate them.
2. Defect reproduction and correlation analysis are time-consuming: Jira text alone is insufficient to reconstruct the on-site phenomenon; investigation efficiency depends on experience.
3. Conclusions lack a unified format: Fault classification, root cause, evidence, and fix suggestions are often written freely by different people — hard to compare later or consolidate.

Solution & Technical Implementation

1. Defect context aggregation: Uses the defect URL as entry point, jointly reading logs, video or screenshots, and problem descriptions.
2. Multimodal + RAG analysis: Uses visual analysis to understand videos/screenshots, combined with RAG to retrieve historical defects and diagnostic knowledge.
3. Structured report: Outputs fault classification, root-cause location, evidence chain, and fix suggestions — forming a diagnostic report directly usable in defect reviews.

Highlights

• Unified Jira entry: Launches analysis directly from the defect URL.
• Log + visual fusion: Analyzes both time information and visual phenomena at the same time.
• Knowledge-enhanced localization: Uses historical defect experience to improve similarity-based issue recognition.
• Standardized reports: Fixed output fields facilitate collaboration across R&D, testing, and project management.

Business Value

• Shortens the time from defect submission to initial classification, localization, and ownership confirmation.
• Structured evidence and historical knowledge retrieval reduce repeat investigation work.
• The same analysis framework can be used across different Jira projects, vehicle models, and versions — suitable for building a unified defect-analysis service.

---