Skip to content

QUICKSTART.md

Latest commit

 

History

History
184 lines (123 loc) · 4.03 KB

QUICKSTART.md

File metadata and controls

184 lines (123 loc) · 4.03 KB

🚀 Quick Start Guide

AI-Q + UDR Research Assistant - Get Running in 30 Minutes

This guide gets you up and running as quickly as possible. For detailed documentation, see DEPLOYMENT.md.

Prerequisites Checklist

  • AWS Account configured (aws configure)
  • Tools installed: terraform, kubectl, helm, docker
  • NVIDIA NGC API Key (Get it here)
  • Tavily API Key (optional, get here)

3-Step Deployment

Step 1: Setup Environment (2 minutes)

cd ~/repos/AIML/Research_as_a_Code

# Set your API keys
export TF_VAR_ngc_api_key="nvapi-YOUR_KEY_HERE"
export NGC_API_KEY="nvapi-YOUR_KEY_HERE"
export TAVILY_API_KEY="tvly-YOUR_KEY_HERE"  # Optional
export AWS_DEFAULT_REGION="us-west-2"

# Verify
echo $NGC_API_KEY
aws sts get-caller-identity

Step 2: Deploy Infrastructure (~20 minutes)

cd infrastructure/terraform
./install.sh

# Configure kubectl (command will be shown at end of install)
aws eks update-kubeconfig --region us-west-2 --name ai-q-udf-hackathon

Step 3: Deploy Application (~40 minutes)

cd ../kubernetes

# Deploy NIMs (30 minutes)
./deploy-nims.sh

# Deploy Agent (10 minutes)
./deploy-agent.sh

Access Your Application

# Get the URL
kubectl get svc aiq-agent-frontend -n aiq-agent

# Look for EXTERNAL-IP column (e.g., xxx.us-west-2.elb.amazonaws.com)
# Open in browser: http://<EXTERNAL-IP>

Test It Works

  1. Open the application URL in your browser
  2. Enter query: "What is Amazon EKS?"
  3. Watch the "Agentic Flow" panel show real-time execution
  4. See the generated report in the right panel

Advanced test:

Enter: "Generate a report on NIMs on EKS with cost-benefit analysis"

Watch as the agent:

  • Recognizes complexity → Selects UDR strategy
  • Compiles multi-step research plan
  • Executes dynamic code
  • Synthesizes comprehensive report

Troubleshooting One-Liners

# Check if everything is running
kubectl get nodes && kubectl get pods -n nim && kubectl get pods -n aiq-agent

# Watch NIMs start
kubectl get pods -n nim --watch

# View agent logs
kubectl logs -n aiq-agent -l component=backend -f

# Restart a stuck NIM
kubectl delete pod <pod-name> -n nim

Cleanup

# Delete agent only
kubectl delete namespace aiq-agent

# Delete NIMs
kubectl delete namespace nim

# Delete infrastructure (stops all costs)
cd infrastructure/terraform
terraform destroy

Estimated Timeline

Task Time
Prerequisites setup 5 min
Terraform (EKS + Karpenter) 20 min
NIMs deployment 30 min
Agent deployment 10 min
Total ~65 min

Costs

With Spot instances: ~$4-5/hour
With On-Demand: ~$15-20/hour

Remember to terraform destroy when done!

Next Steps

Key Features to Demo

  1. Simple RAG: Ask "What is Kubernetes?"
  2. UDR Strategy: Ask "Compare EKS vs GKE with cost analysis"
  3. Real-Time Flow: Watch the left panel stream agent decisions
  4. Report Quality: Download and review the generated markdown

Common Issues

Issue Solution
"Permission denied" on scripts chmod +x infrastructure/**/*.sh
NIMs stuck in Pending Wait 10min for Karpenter to provision GPU nodes
ImagePullBackOff Check NGC_API_KEY is correct
Can't access frontend Wait 5min for LoadBalancer DNS to propagate

Support

Happy Hacking! 🎉

Built for AWS & NVIDIA Agentic AI Unleashed Hackathon 2025