Query your documents with AI, privately
🔒 VPS onlyTurn your internal documents into an AI-queryable knowledge base with LightRAG and its knowledge graph, installed and maintained by DINAO. Your content stays on our French servers.
What is LightRAG?
LightRAG is a lightweight RAG (Retrieval Augmented Generation) framework that combines a knowledge graph and vector embeddings to better understand and query a corpus of documents. Where classic RAG simply retrieves close passages, LightRAG builds a graph linking entities and concepts, enabling more comprehensive answers to cross-document questions.
It offers five query modes (local, global, hybrid, naive, and mixed), incremental updates to the database without reconstruction, and role-based configuration (extraction, query, keywords, vision model/VLM). For models, it connects to a local open-weight LLM (via Ollama) or any OpenAI-compatible API, with optional embedding and reranking choices. Multimodal document processing relies on MinerU and Docling.
LightRAG supports numerous storage backends (PostgreSQL, MongoDB, Neo4j, Memgraph, Milvus, Qdrant…) and exposes a REST API as well as a web interface. Self-hosted with DINAO and connected to a local model, it becomes a truly sovereign document assistant: your content never leaves France.
Host LightRAG at DINAO
No managed container for LightRAG: this application is installed on a VPS hosted in France (2 vCPU / 2 Go / 10 Go minimum).
Application available on a VPS
This application requires privileged access to the host machine, which is incompatible with managed shared hosting. It is installed on a VPS (virtual private server) that you administer: DINAO delivers it with Docker and the management panel of your choice, and you install and maintain LightRAG yourself. The VPS is not managed by DINAO unless you add the managed-services option when ordering.
Technical reason: bind-mount d'un chemin hôte (volumes nommés uniquement en managé).
Discover the Docker VPS →Requirements for LightRAG: 2 vCPU, 2 Go, 10 Go. Our Docker VPS at €11 excl. VAT per month covers them.
This application uses AI
The container hosts the application, not the AI engine (which requires dedicated GPUs) : inference is handled externally, with your own provider key. Prioritize a sovereign engine — Mistral AI, NumSpot, Scaleway, or OVHcloud AI Endpoints (France, GDPR) ; an international provider (OpenAI, Anthropic…) only if a specific capability requires it. Inference subscriptions are not included in hosting.
Technical details
You may also like…
Other applications managed by DINAO in the same category, hosted in France and GDPR compliant.
You might be wondering…
What is graph-based RAG?
RAG (Retrieval Augmented Generation) provides LLMs with excerpts from your documents so they can answer based on evidence. LightRAG adds a knowledge graph linking entities and concepts, yielding more comprehensive answers to cross-cutting questions.
Can I use my own model?
Yes. You connect a local open-weight LLM served by Ollama, or an OpenAI-compatible API. You also choose the embedding model and, optionally, a reranker.
Do my documents remain confidential?
In local mode, yes: documents, graph, and queries stay on your instance in France; nothing is transmitted to a third party. See the sovereignty note below for the external API case.
How do I add new documents?
LightRAG supports incremental updates: you add documents continuously without rebuilding the entire database, and multimodal content (PDFs, images) is handled via MinerU/Docling.
Where is the data hosted?
On DINAO infrastructure in France, in one of the available data centers. Your data does not leave the territory in local mode.
