GENIE.AI Documentation

Sovereign, multilingual RAG for the public sector — installation, operations, architecture, and contribution guides.

GENIE.AI — Sovereign, multilingual RAG for the public sector.

GENIE.AI is an open-source generative-AI framework that lets public-sector organizations run a retrieval-augmented-generation (RAG) chat assistant on their own infrastructure — no third-party model API required, no data leaves your perimeter, fully configurable, and shipped under a DPG-compliant license. The stack is built on OPEA (Open Platform for Enterprise AI) and supports 14+ languages out of the box.

Pick your path

GENIE.AI serves four audiences. Pick the one that matches your goal — each links to a 10-minute quickstart and a curated set of follow-on docs.

End user

Use the chat assistant, switch language, give feedback. No installation.

  1. Open the chat URL your admin gave you
  2. Sign in with the SSO button
  3. Ask a question about an indexed document
Deployer

Install GENIE.AI on a single host with Docker Compose — production-ready in 15 minutes.

  1. Clone the repo and cp env .env
  2. Set the required secrets (DB, Keycloak, HF token)
  3. docker compose up -d and verify the health endpoint
Developer

Extend the backend or the RAG pipeline — local dev loop in 10 minutes.

  1. Run npm install in components/gov-chat-backend
  2. Spin up dependencies with docker compose up -d
  3. Run npm run dev and hit localhost:3000/api/health
Contributor

Open your first merge request — fork, branch, test, request review.

  1. Fork and clone the repo
  2. Create a feature branch and write tests first
  3. Open the MR using the template — CI + review gate the merge

Sections

13 sections, ordered by the path most readers take. Each card lists the section’s mission in one line.

Get started

Be productive in 15 minutes — quickstarts for all four personas, concepts, and FAQ.

Core

High-level project overview, integration architecture, source tree, dev guide.

Deploy

Install GENIE.AI on Docker Compose, Swarm, or a remote GPU node.

Operate

Run a live deployment — backup, scaling, updates, troubleshooting.

Observe

Metrics, logs, traces, dashboards, and alerting via VictoriaMetrics + Grafana.

Configure

Realm setup, IdP integration, branding, dashboard, CORS / CSP, locale whitelist.

Architecture

C4 context/container view, auth flows, trust boundaries, OPEA contract.

RAG pipeline

Embedding, retrieval, reranking, generation, translation — every stage.

Knowledge base

Ingestion, taxonomy, document lifecycle, content quality guidance.

Reference

Single-source env-var table, HTTP API, OPEA extensions, glossary.

Frontend

Vue 3 SPA — auth flow, chat UX, sidebar, admin dashboard, theming.

Mobile

Flutter app — architecture, deployment, OIDC PKCE login, SSE chat.

Contribute

Repo layout, dev workflow, MR guide, i18n, security triage, style.

Search the docs

Use the search box at the top (or press /). It indexes titles, headings, and body text — search by:

  • an env-var name (e.g. RERANKER_TOP_N)
  • a service name (e.g. chatqna, dataprep, keycloak)
  • an error message (e.g. dataprep.status failed)
  • a command (e.g. docker stack deploy)

For the canonical reference of every supported env-var with default and scope, see Environment variables.

Recent updates

This site is rebuilt continuously. Major waves in the past 48 hours:

  • 2026-09-29: Wave 36 — fix broken cross-section refs + callout shortcode.
  • 2026-09-29: Wave 39 — 7 dispersion fixes (labelling + ingestion-log + api-contracts consolidations + locale-whitelist + mode fixes).
  • 2026-09-29: Wave 41 — grader P0+P1 fixes (alias mechanism + content accuracy).
  • 2026-09-29: Wave 43 — rebase on main + 12 P0+P1 fixes (VL/Collector always-on, VlUnavailableError, /api/logger removed, bge-large sweep, ChatQnA token rules).
  • 2026-09-29: Wave 44 — P2 sweep (legacy dirs removed, ArangoDB self-heal, aiohttp, weather i18n).
  • 2026-09-29: Wave 47 — full audit post-rebase (15 P0+P1+P2 issues).
  • 2026-09-29: Wave 49 — 15 grader consensus fixes (env-vars Keycloak/admin bootstrap, KEYCLOAK_URL callout, branding/dashboard inbound, deployment-flavors doc, 25 backend/api-contracts-backend refs rewritten + stub deleted).
  • 2026-09-29: Wave 51 — 5 final fixes (chat-ux.md link, development-guide.md e2e-tests URLs, configure/index.md deployment-flavors item, install-guide.md Step 11d, env-vars.md EMAIL* expansion).
  • 2026-09-29: Wave 53 — final fixes (env-vars.md DATAPREP_CONTEXTUAL_DOC_BUDGET row, locale-whitelist.md absolute paths, deployment-flavors.md cross-link).
  • 2026-09-29: Wave 56 — fresh audit P0 fixes.
  • 2026-09-29: Wave 58 — 12 grader fixes (8888 port, VICTORIAMETRICS_RETENTION, deployment-flavors paths, etc.).

For per-file changes, see each doc’s footer “last_reviewed” stamp.

Report a problem

Docs are maintained alongside the code in the same repository. If a page is wrong, missing, or out of date, open an issue or send a merge request following the contribution flow.

Latest review stamp on this site: 2026-09-29. Pages show their own last_reviewed in the footer.

Get started

Be productive in 15 minutes — quickstarts, concepts, and FAQ for all four personas.

Core & Project Overview

Top-level project context for GENIE.AI: scope, source-tree map, integration architecture, and contributor onboarding.

Deployment

End-to-end installation, Docker Compose / Swarm deployment, and GPU setup guides.

Operations

Day-to-day operations for a running GENIE.AI deployment — backup, restore, updates, scaling, troubleshooting, and operator-facing diagnostic surfaces.

Observability

OpenTelemetry-native metrics, logs, and traces for the GENIE.AI stack — collection, storage, dashboards, and alerting.

Configuration

Operator-facing configuration guides for Keycloak and external identity providers.

Architecture

System-wide architecture: C4 container view, authentication flows, the service-to-service auth matrix, and the OPEA microservices contract surface.

RAG Pipeline

Sovereign retrieval-augmented generation pipeline — embedding, hybrid retrieval, reranking, generation, and optional translation.

Knowledge Base

Managing the GENIE.AI knowledge base — ingesting documents, the processing pipeline, the label taxonomy, document lifecycle, and content guidance.

Reference

Canonical lookup — env-vars, HTTP API, OPEA protocol extensions, glossary.

Frontend (gov-chat Web)

Vue 3 + Vuex single-page app — UI component inventory, state-management patterns, theme system, auth flow, chat UX, sidebar, and the admin dashboard.

Mobile (genie_ai_mobile)

Flutter client docs: UI inventory, architecture, deployment, end-user auth, and chat pipeline.

Contribute

How to contribute to GENIE.AI — open an MR, follow the dev workflow, add a translation, triage a CVE.

Last modified September 30, 2026: fix(docs): lean fix Commit 1 — IA weight reorder + about.md link fix (72c35f1ed)