Deployment Architecture

Guide

Deploy RADAR self-hosted beside your existing agent stack. Evidence stays in your infrastructure — VPC, on-prem, or air-gapped.

System Model#

ArchitectureSidecar deployment

Your Infrastructure

AgentsLangChain \u2022 CrewAI

LLM GatewayOpenAI \u2022 Anthropic

SIEMSplunk \u2022 Sentinel

↓

RADAR

CollectorTraces \u2022 Events

PII Detection50+ patterns

Control MapEU AI Act \u2022 GDPR

ExportJSON \u2022 PDF \u2022 SIEM

↓

Regulator-Ready Evidence

Trace RecordsSession timeline

Control MappingFramework-aligned

Export BundleAudit-ready

Self-hosted \u2022 Zero egress \u2022 Your data stays inside

Sidecar

RADAR runs beside your LLM gateway and tool infrastructure. It observes without intercepting the execution path.

Your Infrastructure

All evidence — traces, findings, exports — stays inside your VPC, on-prem environment, or air-gapped network.

No Execution Ownership

RADAR does not run your agents, orchestrate workflows, or replace your existing stack. It monitors and records.

Docker Compose#

Recommended for evaluation and mid-scale production deployments. RADAR runs as a single service with local volume storage.

docker-compose.yml

version: '3.8'
services:
  radar:
    image: akioudai/radar:latest
    ports:
      - "8080:8080"
    volumes:
      - radar-data:/data
    environment:
      - RADAR_LICENSE_KEY=${RADAR_LICENSE_KEY}
      - RADAR_RETENTION_DAYS=30
      - RADAR_STORAGE_PATH=/data
    restart: unless-stopped

volumes:
  radar-data:

Volumes

• radar-data: evidence storage
• /data: trace, finding, export persistence

Environment

• RADAR_LICENSE_KEY: your evaluation license
• RADAR_RETENTION_DAYS: evidence retention period

Kubernetes#

For enterprise deployments requiring auto-scaling, rolling updates, and service mesh integration.

deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: radar
  labels:
    app: radar
spec:
  replicas: 2
  selector:
    matchLabels:
      app: radar
  template:
    metadata:
      labels:
        app: radar
    spec:
      containers:
      - name: radar
        image: akioudai/radar:latest
        ports:
        - containerPort: 8080
        env:
        - name: RADAR_LICENSE_KEY
          valueFrom:
            secretKeyRef:
              name: radar-secrets
              key: license-key
        - name: RADAR_RETENTION_DAYS
          value: "90"
        volumeMounts:
        - name: evidence-storage
          mountPath: /data
        resources:
          requests:
            memory: "2Gi"
            cpu: "500m"
          limits:
            memory: "4Gi"
            cpu: "2"
        livenessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 30
          periodSeconds: 15
      volumes:
      - name: evidence-storage
        persistentVolumeClaim:
          claimName: radar-evidence

Kubernetes considerations

Use persistent volumes for evidence retention. For air-gapped environments, ensure the image is available in your private registry. RADAR does not require external network access after deployment.

Air-Gapped Deployment#

RADAR is designed for zero-trust and air-gapped environments. No external API calls, no telemetry, no cloud dependencies.

No Egress Required

RADAR does not phone home. All evidence processing, retention, and export is local.

Offline License Validation

License keys are validated at startup with optional offline activation for air-gapped networks.

Local Storage Only

All traces, findings, and exports reside on local or networked storage inside your security perimeter.

Evidence Retention#

Configure how long RADAR retains traces, findings, and export history. Retention is enforced at the storage layer and can be customized per environment.

bash

# Set retention to 90 days
docker exec radar radar retention set --days 90

# Check current retention and usage
docker exec radar radar retention status
# Retention: 90 days
# Current usage: 12.4 GB / 50 GB
# Oldest record: 2026-02-03

Next Steps#

Self-Hosted Evaluation

Walk through the evaluation process step by step

Evidence Model

Understand trace records, findings, and control mapping