Self-hosted nivq documentation

The governed, on-prem agentic AI data assistant for regulated enterprises — read-only by design, audit-first, and yours to run.

The handful of ideas — workspaces, agents, governance, audit, and licensing — that everything else in nivq builds on.

What you need before installing nivq — host, runtime, datastores, registry access, an LLM provider, and a licence.

Pull the nivq image and run it on a single host — either with Docker Compose (Postgres and Redis included) or as a single container pointed at datastores you already run.

Deploy nivq to a Kubernetes cluster with the official Helm chart — image pull secret, values, install, activation, ingress, and scaling.

Run nivq directly on a Linux host with a Java runtime — no container engine. Install the JAR, configure it, and run it as a systemd service.

Run the nivq web client — the browser UI users open. A static single-page app served by nginx (or any webserver), configured at runtime so one build runs anywhere — Docker, Kubernetes, or bare metal.

Every environment variable nivq reads — datastores, security, LLM providers, identity, payments, and observability.

How users sign in to nivq — social login, generic OIDC (Keycloak/Okta/Azure AD), per-workspace SSO, a local bootstrap admin, and mapping IdP groups to workspaces and roles.

Embed the nivq chat experience on any site with one tag — a framework-agnostic <nivq-chat> Web Component, a small token broker so your secret never reaches the browser, theming, and configuration.

How nivq licensing works — activate a fresh deployment, bind to a host, and understand the licence lifecycle.

Take nivq from a first install to a hardened production deployment — image pinning, persistence and backups, TLS, observability, and health probes.

How nivq is built — the request path, multi-tenant isolation, the read-only guard, audit, and the governed-agentic design.