AI-Powered QA Intelligence Suite AI-powered QA Intelligence Suite using CrewAI, Jira Integration, Defect Intelligence, Root Cause Analysis, Production Risk Prediction, and Regression Recommendation.

AI-Powered Defect Analysis and Assignment Assistant

Bug Triage Agent is an intelligent multi-agent system built using CrewAI that automates the initial defect triage process by analyzing bug reports, identifying probable root causes, assessing severity and priority, and recommending the most suitable assignee or team for resolution.

The agent leverages Large Language Models (LLMs), project knowledge bases, historical defect data, application documentation, and test execution results to provide consistent and data-driven triage decisions.

Key Capabilities

1. Automated Bug Analysis Reads bug descriptions, logs, screenshots, stack traces, and test results. Extracts critical information from defect reports. Identifies affected modules and functionalities.
2. Severity and Priority Recommendation Evaluates business impact and technical impact. Suggests Severity (Critical, Major, Minor, etc.). Recommends Priority (P1, P2, P3, etc.). Reduces subjective decision-making during triage meetings.
3. Root Cause Prediction Analyzes defect patterns and historical issues. Predicts likely causes such as: UI defect Backend issue API failure Configuration issue Environment issue Data-related issue
4. Intelligent Assignment Recommendation Identifies the most relevant developer or team. Uses module ownership and historical defect resolution data. Reduces reassignment cycles.
5. Duplicate Defect Detection Compares new defects against historical defects. Flags potential duplicates. Helps reduce defect backlog noise.
6. Risk and Impact Assessment Identifies affected business areas. Predicts regression risks. Suggests impacted test suites.
7. Knowledge Base Integration Connects with: Jira Confluence Test Management Tools Historical Defect Repositories Requirement Documents
8. Triage Summary Generation

Automatically generates:

Bug Summary

Module: Login Severity: High Priority: P1 Probable Cause: Token Validation Failure Recommended Team: Backend Authentication Team Duplicate Risk: Low Suggested Action: Immediate Investigation Proposed Multi-Agent Architecture Agent Responsibility Defect Analyzer Agent Analyze bug details Log Analysis Agent Review logs and stack traces Root Cause Agent Predict probable cause Severity Assessment Agent Recommend severity Assignment Agent Suggest assignee/team Duplicate Detection Agent Find similar historical defects Report Generator Agent Create triage summary Business Benefits Reduces manual triage effort by 50–70%. Improves consistency in defect classification. Faster assignment to the right team. Reduces triage meeting duration. Enables data-driven defect management. Accelerates defect resolution lifecycle.

ShinuAI Crew — QA Bug Triage Agent

Lightweight agent for triaging QA bug reports and assisting automated bug classification.

Files

• main.py — entry point: fetches the bug report, builds the crew, runs the triage, and generates the Production Risk Intelligence report
• agents/llm.py — Groq LLM setup (GroqLLM, groq_llm)
• agents/qa_agents.py — agent definitions (Bug Triage Analyst, Root Cause Investigator, Test Recommender, Production Risk Analyst)
• agents/qa_tasks.py — task definitions for the crew (build_tasks, build_release_risk_task)
• utils/jira_client.py — JIRA ticket fetching helpers and sample bug report fallback
• utils/historical_defects.py — loads and summarizes historical defect data from resources/historical_defects.csv
• utils/risk_intelligence.py — Production Risk Intelligence engine: computes the Release Risk Score, Top Risk Areas, Potential Production Failures, Recommended Regression Areas, and Release Readiness recommendation from resources/historical_defects.csv, resources/release_scope.txt, and resources/project_context.md
• resources/historical_defects.csv — historical defect/bug records (Jira export format) used for risk analysis
• resources/project_context.md — module risk profile and production risk rules used to enrich the risk report
• resources/release_scope.txt — list of modules included in the current release
• output/ — generated production_risk_report.json and production_risk_report.txt (created on each run)
• requirement.txt — Python dependencies

Requirements

• Python 3.8+
• Install dependencies:

pip install -r requirement.txt

Environment

Create a .env file from .env_Sample and set the required variables.

copy .env_Sample .env
# then edit .env and fill values

Common variables:

• OPENAI_API_KEY — API key for language model access
• GITHUB_TOKEN — (optional) token for GitHub API operations

Usage

Run the main script:

python main.py

Production Risk Intelligence

In addition to the per-bug triage workflow, every run also produces an executive-friendly Production Risk Intelligence report for the upcoming release. This is computed deterministically with pandas (no ML models, no vector DB) from:

• resources/historical_defects.csv — historical defects, severities, and modules
• resources/release_scope.txt — modules included in the current release
• resources/project_context.md — module risk profile and known failure patterns

The report includes:

1. Release Risk Score (1.0–10.0) with a transparent, weighted breakdown
2. Top Risk Areas — risk level, defect count, historical trend, and reason per module
3. Potential Production Failures — likely failure scenarios with impact and confidence level
4. Recommended Regression Areas — prioritized regression plan per module
5. Release Readiness Recommendation — GO / CONDITIONAL GO / NO GO with supporting reasons

A short executive summary is added by the Production Risk Analyst agent on top of these computed figures (it does not change the score or readiness call). On each run, the full report is printed to the console and saved to:

• output/production_risk_report.json — structured output
• output/production_risk_report.txt — human-readable report

Contributing

Feel free to open issues or PRs in the repository.

License

Use as you like. Add a license file if needed.