About GENIE.AI
GENIE.AI is an open-source generative-AI framework for the public sector: a sovereign, Digital Public Good (DPG)-compliant Retrieval-Augmented Generation (RAG) platform with multilingual support. It answers questions strictly from a knowledge base you control, running every model on infrastructure you own. It integrates with OPEA (Open Platform for Enterprise AI) for AI/ML services.
Why it matters
Public-sector institutions have constraints that generic AI services cannot meet. GENIE.AI is built around them:
- Sovereignty. Data, models, and the answering pipeline all run on infrastructure the operator controls. No query, document, or user interaction leaves the deployment to a third-party model API.
- Grounded answers, not guesses. The LLM is constrained to answer only from retrieved knowledge-base content and to abstain when the answer is not there — critical where a confident but wrong answer is worse than none.
- No vendor lock-in. Standards-based throughout: OIDC identity, OpenTelemetry observability, OpenAI-compatible model APIs, W3C trace context. Models, identity providers, and storage can be swapped without rewriting the platform.
- Digital Public Good. GENIE.AI is developed as a DPG — open and reusable across governments and public agencies.
Security & data protection
GENIE.AI’s security model assumes a deployment on infrastructure the operator trusts:
- Self-hosted models — embedding, reranking, generation, and translation run on your own GPU/CPU via OPEA / vLLM / TEI. There is no call to an external model provider by default.
- Identity & access — Keycloak (OIDC) handles authentication and role-based access; the API gateway (Kong) enforces routing and rate limits. See Configuration → Keycloak.
- Document safety — every uploaded document is virus-scanned (ClamAV) before it is stored or processed. See Knowledge base → Ingestion.
- PII-safe telemetry — observability spans are filtered to strip sensitive attributes (tokens, passwords, user PII) before export. See Observability → Tracing.
- Hardened backend — security headers (
helmet), rate limiting, and CORS/CSP policies are applied at the backend and gateway. - Data residency — because the whole stack is self-hosted, data stays within the deployment’s jurisdiction.
Who it is for
GENIE.AI serves a range of public-sector and institutional contexts — government agencies, public services, NGOs, and international organisations that need a trustworthy, self-hosted question-answering capability over their own content.
The people who use this documentation fall into four roles:
| Role | What you’ll do | Start here |
|---|---|---|
| Evaluator / decision-maker | Assess fit, sovereignty, compliance. | This page + Project overview |
| Operator / admin | Deploy, configure, run, observe. | Deployment + Observability |
| Knowledge / content manager | Curate the knowledge base. | Knowledge base |
| Developer / integrator | Extend, integrate, customise. | Architecture + RAG pipeline |
What you can do with it
- Stand up a multilingual, grounded Q&A service over your own document base.
- Organise content with a service-category taxonomy for precise, on-topic retrieval.
- Observe the full RAG pipeline (retrieval, reranking, generation) with distributed tracing and pre-built dashboards.
- Deploy on a single host (Compose) or a cluster (Swarm), with or without local GPUs.
For the full feature surface and the architecture behind it, read the Project overview and the Architecture section.
Getting started
New to GENIE.AI? The fastest path:
- Read the Project overview for the 10-minute orientation.
- Follow the Installation & Configuration Guide to stand up a deployment.
- Load documents via the Knowledge base workflow and start asking questions.