AWS HealthOmics AI Troubleshooting Assistant

AI-assisted troubleshooting for AWS HealthOmics genomic workflows with custom knowledge base support

Overview

The HealthOmics AI Troubleshooter is an intelligent troubleshooting system for bioinformatics engineers working with AWS HealthOmics genomic workflows. It integrates with Kiro IDE to provide natural language troubleshooting, AI-powered root cause analysis, and actionable recommendations in seconds.

Key Capabilities

• 🤖 AI-Powered Analysis: Specialized bioinformatics agent built on AWS Bedrock AgentCore
• 🔍 Root Cause Detection: Automatically correlates data across HealthOmics, CloudWatch, CloudTrail, and X-Ray
• 💬 Natural Language Interface: Ask questions like "Why did my workflow fail?" and get instant answers
• 🧬 Genomics Expertise: Understands WGS, WES, RNA-Seq workflows and tools like GATK, BWA-MEM2, Samtools
• ⚡ Fast Results: Get troubleshooting insights in 5-30 seconds, not hours
• 🚀 Turnkey Deployment: One-command infrastructure setup with automated IAM policies
• 📚 Custom Knowledge: Ingest your organization's documentation and historical troubleshooting data
• 🔄 Multi-Workflow Support: Works with Nextflow, WDL, and CWL workflows

Quick Start

Installation

1. Install the Power in Kiro IDE from the Powers marketplace
2. Install required dependency Powers when prompted
3. Follow the Setup Wizard to deploy infrastructure

First Query

"Why did workflow run omics-abc123 fail?"

The agent will analyze the failure, identify root causes, and provide specific recommendations.

Architecture

Kiro IDE
  ↓
AgentCore Bioinformatics Agent (AWS Bedrock)
  ↓
HealthOmics + Observability Powers
  ↓
AWS Services (HealthOmics, CloudWatch, CloudTrail, X-Ray, S3)

Development

Prerequisites

• Node.js >= 18.0.0
• TypeScript 5.6+
• AWS Account with HealthOmics access

Setup

npm install
npm run build
npm test

Project Structure

.
├── src/
│   ├── agent/              # AgentCore agent implementation
│   ├── powers/             # HealthOmics and Observability Powers
│   ├── orchestration/      # Query orchestrator and analyzers
│   ├── infrastructure/     # CDK stack definitions
│   ├── setup/              # Setup wizard
│   ├── knowledge/          # Knowledge base management
│   └── types/              # TypeScript type definitions
├── tests/                  # Unit and property-based tests
├── docs/                   # Documentation
├── examples/               # Example workflows and scenarios
└── cdk/                    # CDK app entry point

Documentation

• Installation Guide
• User Guide
• Architecture Overview
• API Reference
• Contributing Guide

How It Works

1. Query: User asks a natural language question about a workflow failure
2. Retrieve: Agent fetches data from HealthOmics, CloudWatch, CloudTrail, X-Ray
3. Analyze: Root cause analyzer correlates data and identifies failure reasons
4. Recommend: Recommendation engine provides specific fixes with parameter values
5. Act: User applies recommendations and re-runs workflow

Example Queries

• "What's the status of my latest workflow run?"
• "Why did workflow run omics-abc123 fail?"
• "Show me resource utilization for run omics-abc123"
• "What IAM permissions are missing?"
• "Which tasks failed in my last run?"
• "Show me the error logs for task xyz"

Custom Knowledge Base

Enhance the agent with your organization's knowledge:

• SharePoint document libraries
• Confluence spaces
• Internal wikis and runbooks
• Historical troubleshooting logs
• Best practices and SOPs

The agent will prioritize your organization's knowledge when providing recommendations.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

This project is licensed under the Apache License 2.0 - see LICENSE for details.

Security

See SECURITY.md for reporting security issues.

Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

Related Projects

• AWS HealthOmics
• AWS Bedrock AgentCore
• Kiro IDE
• AWS Enhanced Monitoring Solution