Package managers
Install the closed-source static daemon binary through apt, AUR, Chocolatey, or Homebrew. Supervise with systemd, launchd, or your platform init.
Deployment
Ship memory like infrastructure.
LibraVDB Memory runs as an operator-managed daemon. Package-manager installs, Docker images, Compose stacks, and a production Helm chart are available now. Every path runs the same cognitive kernel — the difference is how you operate it.
Docker
Production Alpine image pulls pre-built binaries from GitHub Releases — no build step. Supports linux/amd64 and linux/arm64. Exposes gRPC on :50051 with a built-in HEALTHCHECK. GGUF embedding backend by default for 3–5× inference throughput over ONNX.
Docker Compose
Single-command start with named volumes for data and model persistence. Pre-configured with GGUF backend, health checks, and restart policy. Mount ONNX models by switching the backend env var.
Kubernetes Helm
Full Helm chart at v1.9.3 — ConfigMap, PVC, Service, Deployment, and optional Secret for mTLS. Separate data and models PersistentVolumeClaims. Resource limits, probes, security context, and affinity rules all configurable.
mTLS transport
Enable gRPC over TLS with a single Helm value. Provide base64-encoded cert, key, and CA — the chart provisions a Kubernetes TLS Secret and configures encrypted transport between your agents and the memory kernel.
Hosted service
Join the waitlist for the managed LibraVDB Memory service. We operate the daemon. Hosted privacy, billing, and data terms are handled through managed signup — separate from the self-managed daemon path.
docker compose -f deploy/docker-compose.yml up -dhelm install libravdbd ./deploy/helm/libravdbd \ --set persistence.dataSize=50Gi \ --set resources.limits.memory=4Gihelm install libravdbd ./deploy/helm/libravdbd \ --set tls.enabled=true \ --set tls.cert=$(base64 -i server.crt) \ --set tls.key=$(base64 -i server.key)Infrastructure as code
Terraform — deploy to any cloud.
Terraform modules wrap the Helm chart for AWS EKS, GCP GKE, and Azure AKS with provider-native storage classes and IAM integration. A standalone EC2 module runs Docker directly — no Kubernetes required. Same cognitive kernel, same gRPC surface, same GGUF inference path — the only difference between terraform apply on EKS and docker compose up is where the daemon process lives.
AWS EKS GCP GKE Azure AKS Standalone VM
AWS EKS
Connects to an existing EKS cluster via the Helm provider. Configures gp3 storage with the EBS CSI driver for low-latency persistent volumes. IAM roles for service accounts map daemon permissions without node-level credential sharing. Cluster name, region, and instance type exposed as Terraform variables.
cd deploy/terraform/eks && terraform apply -var cluster_name=my-clusterGCP GKE
Targets an existing GKE cluster with standard-rwo StorageClass backed by persistent SSD. Workload Identity binds a GCP service account to the daemon pod for Cloud KMS and artifact registry access. Project ID, cluster name, and zone are configurable variables.
cd deploy/terraform/gke && terraform apply -var project=my-project -var cluster_name=my-clusterAzure AKS
Deploys to an existing AKS cluster using azurefile StorageClass with ReadWriteOnce access mode. Supports Azure AD pod identity for Key Vault and container registry authentication. Resource group, cluster name, and location passed as Terraform variables.
cd deploy/terraform/aks && terraform apply -var resource_group=my-rg -var cluster_name=my-clusterStandalone VM
EC2 instance running Docker on Ubuntu 22.04 LTS — no Kubernetes, no Helm, no cluster overhead. Provisions a security group opening gRPC :50051 to a configurable CIDR range. User data script installs Docker and starts the Compose stack on boot. Ideal for single-machine evaluation, edge deployment, or CI test environments.
cd deploy/terraform/vm && terraform applyapt AUR Chocolatey Homebrew Docker Helm v1.9.3 Terraform EKS GKE AKS mTLS linux/amd64 linux/arm64