Skip to content

QUICKSTART_RAG_ENTERPRISE.md

Latest commit

 

History

History
156 lines (119 loc) · 3.57 KB

QUICKSTART_RAG_ENTERPRISE.md

File metadata and controls

156 lines (119 loc) · 3.57 KB

Quick Start: Enterprise RAG with US Customs Tariffs

Goal: Get the NVIDIA RAG Blueprint deployed and tariff PDFs ingested in ~30 minutes.

Prerequisites (5 minutes)

  1. EKS Cluster Running

    kubectl cluster-info

    If not, deploy it first:

    cd infrastructure/terraform
./install.sh

  2. NGC API Key

    export NGC_API_KEY="nvapi-..."

  3. kubectl Configured

    aws eks update-kubeconfig --region us-west-2 --name aiq-udf-eks

Step 1: Deploy RAG Blueprint (15 minutes)

cd infrastructure/helm
./deploy-rag-blueprint.sh

What's deploying:

  • ✅ Milvus vector database
  • ✅ RAG ingest server (PDF processing)
  • ✅ RAG query server (search & retrieval)

Wait for it to finish, then verify:

./verify-rag-deployment.sh

Expected output:

✅ Milvus - 1/1 pods running
✅ RAG Query Server - 2/2 pods running
✅ RAG Ingest Server - 1/1 pods running

Step 2: Ingest Tariff PDFs (20 minutes)

cd ../../scripts
./setup_tariff_rag_enterprise.sh

What's happening:

  1. Port-forwards to RAG ingest service
  2. Creates us_tariffs collection
  3. Uploads 99 tariff PDF chapters
  4. Runs test queries

Expected output:

✅ Success: 99
📦 Total:   99

Step 3: Test in UI (2 minutes)

  1. Get frontend URL:

    kubectl get svc -n aiq-agent aiq-agent-frontend -o jsonpath='{.status.loadBalancer.ingress[0].hostname}'

  2. Open in browser

  3. Enter these test queries:

    • "What is the tariff for replacement batteries for a Raritan remote management card?"
    • "What's the tariff of Reese's Pieces?"
    • "Tariff of a replacement Roomba vacuum motherboard, used"

  4. Set collection name: us_tariffs

Troubleshooting

Pods not starting?

kubectl get pods -n rag-blueprint
kubectl describe pod <pod-name> -n rag-blueprint

Ingestion failing?

# Check port-forward
curl http://localhost:8082/health

# Check logs
kubectl logs -n rag-blueprint -l app=rag-ingest-server -f

No query results?

# Verify collection was created
kubectl logs -n rag-blueprint -l app=rag-ingest-server | grep "us_tariffs"

# Re-run ingestion
cd scripts
./setup_tariff_rag_enterprise.sh

Architecture At-A-Glance

User Query
    ↓
AI-Q Agent Backend
    ↓
RAG Query Server (8081) ← Milvus Vector Store
    ↓                         ↑
Embedding NIM (8000)          |
                              |
                    RAG Ingest Server (8082)
                              ↑
                        Tariff PDFs (99)

What You Get

✅ Enterprise Vector Store: Milvus (production-grade)
✅ Hybrid Search: Vector + keyword (BM25) for tariff codes
✅ GPU-Accelerated: PDF processing with NVIDIA NIMs
✅ Scalable: Auto-scales with Karpenter
✅ Citation Support: Returns source documents with answers

Next Steps

  • Add more document collections (regulations, trade agreements)
  • Integrate RAG into UDR dynamic strategies
  • Scale query server for production traffic
  • Set up monitoring and alerts

Full Documentation

For detailed configuration, troubleshooting, and operations:

Deployed in ~30 minutes! Now you have enterprise-grade RAG powered by NVIDIA blueprints. 🚀