Chat with your documents, privately and in France
A document AI assistant powered by PrivateGPT, installed and maintained by DINAO. Your files are ingested, vectorized, and queried entirely on our French servers — nothing goes to OpenAI.
What is PrivateGPT?
PrivateGPT is an open source AI application that allows you to query your local documents in natural language using locally run language models. Everything happens on your machine or network : your documents remain private and are never transmitted to an external service.
The core of the tool is a complete RAG (Retrieval Augmented Generation) pipeline : PrivateGPT loads your files (PDF, TXT, DOCX…), splits them into chunks, calculates open-source embeddings to capture their meaning, and stores these vectors in a local vector database. It can then answer your questions based on your own content, citing sources.
The application is built around an OpenAI-compatible API (FastAPI), which facilitates switching between local and cloud models. Highly modular (LlamaIndex, Pydantic), it allows customization of the LLM, embedding model, and vector store. Its main asset is the guarantee of confidentiality : your proprietary data stays within your infrastructure.
Host PrivateGPT at DINAO
Resource tiers compatible with PrivateGPT prerequisites (minimum 4 vCPU / 8 Go / 20 Go). Hosted in France, fully managed.
- 4 dedicated vCPU
- 8 Go RAM
- 80 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 8 dedicated vCPU
- 16 Go RAM
- 160 GB NVMe
- Daily backups
- Managed & monitored by DINAO
2 hidden tier(s) (insufficient resources for this app) : Découverte, Standard
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.
Ollama
Managed Ollama by DINAO: run Llama, Mistral, and Qwen on GPUs in France, GDPR compliant. P…
Text Generation Web UI
Text Generation Web UI managed by DINAO: your local LLM interface hosted in France, GDPR-c…
Chatpad AI
Chatpad AI managed by DINAO: elegant AI chat interface hosted in France, GDPR compliant. A…
Flowise
Flowise managed by DINAO: build AI agents and RAG chatbots with no-code, hosted in France …
LocalAI
Managed LocalAI by DINAO: replace the OpenAI API with a private instance in France, GDPR-c…
Open WebUI
Managed Open WebUI by DINAO: your private ChatGPT hosted in France, GDPR compliant, connec…
You might be wondering…
Do my documents go to OpenAI?
No. In local mode, ingestion, vectorization, and the model run entirely on your DINAO instance in France. Your files never leave the server and are not used to train any third-party models.
What file formats can I ingest?
PDF, Word (DOCX), plain text, and other common formats. PrivateGPT splits documents, calculates embeddings, and stores them in a local vector database to answer your questions.
Are the answers reliable?
PrivateGPT relies on the actually ingested content (RAG) and can cite source passages. Quality depends on the chosen model and your documents: a GPU and a larger model improve relevance.
Can I connect my own tools?
Yes. PrivateGPT exposes an OpenAI-compatible API (FastAPI), allowing your internal applications to query it as they would with the cloud, while remaining sovereign.
Is a GPU required?
Not mandatory to start with small models on CPU. For large corpora or more powerful models, a GPU (optional or dedicated) significantly speeds up ingestion and responses.
