Skip to Content
DocumentationSelf-HostedDeployment

Self-Hosted Deployment

QuantumVerifi can be deployed in your own cloud environment or air-gapped infrastructure. Self-hosted deployment gives you full control over data residency, network access, and compliance requirements.

Architecture Overview

A self-hosted Scout deployment consists of these components:

ComponentPurposeReplicas
API ServerREST API, authentication, job coordination2+
WorkerRuns analysis workflows, LLM orchestration2+ (auto-scaled)
Web FrontendNext.js dashboard and analysis UI2+
PostgreSQLAnalysis results, billing, compliance data1 (managed recommended)
RedisJob queues, caching, real-time events1
Object StorageTest artifacts, evidence chains, training dataS3-compatible
LLM GatewayMulti-provider routing with failover1

Optional components:

  • Temporal — Durable workflow orchestration (recommended for production)
  • Qdrant — Vector database for RAG-enhanced generation
  • V-JEPA Server — GPU-accelerated visual AI (requires NVIDIA GPU)

Deployment Options

Scout ships as a set of container images with Kubernetes manifests. Supports any Kubernetes distribution:

  • AKS (Azure Kubernetes Service) — production-tested
  • EKS (AWS Elastic Kubernetes Service)
  • GKE (Google Kubernetes Engine)
  • Self-managed Kubernetes clusters

Manifests use Kustomize for environment-specific configuration. ArgoCD is supported for GitOps deployments.

Docker Compose

For evaluation or small-team deployments, a Docker Compose configuration is available that runs all components on a single host.

Prerequisites

  • Kubernetes 1.28+ (or Docker Compose for evaluation)
  • Container registry access (images provided via Azure Container Registry)
  • PostgreSQL 15+ (managed or self-hosted)
  • Redis 7+ (managed or self-hosted)
  • S3-compatible object storage (Azure Blob, AWS S3, MinIO)
  • LLM API access (Azure OpenAI, Anthropic, or self-hosted via Ollama)

Configuration

All configuration is managed through environment variables and Kubernetes ConfigMaps/Secrets.

Required Secrets

SecretDescription
DATABASE_URLPostgreSQL connection string
REDIS_URLRedis connection string
LLM_API_KEYAPI key for your LLM provider
AUTH_SECRET_KEYAuthentication provider secret
ENCRYPTION_KEYToken encryption key

Sandbox Execution

Tests run in isolated sandbox containers. Self-hosted deployments support:

ProviderDescription
Kubernetes JobsOn-demand pods per test run (recommended)
Docker-in-DockerContainer-based isolation
LocalHost execution (testing only)

Each sandbox gets a fresh filesystem, pre-installed language runtimes, and network isolation.

Air-Gapped Mode

For environments with no internet access, Scout supports fully offline operation:

  • Offline LLM — Use Ollama with locally-hosted models (Qwen, CodeLlama)
  • Offline training — QLoRA fine-tuning runs on local GPU infrastructure
  • No telemetry — All analytics and observability stay within your network
  • Container mirroring — Images served from your internal registry

Health Monitoring

EndpointPathPurpose
API Server/healthLiveness and readiness checks
Web Frontend/Frontend availability
LLM Gateway/healthLLM provider connectivity

Getting Started

  1. Contact us for access to container images and deployment manifests
  2. Review the Architecture Overview for detailed component diagrams
  3. Configure your environment variables and secrets
  4. Deploy using Kustomize or Docker Compose
  5. Verify health endpoints and run your first analysis

For deployment assistance, reach out to [email protected].