One single door to all your LLMs
The AI gateway open-source LiteLLM, installed and maintained by DINAO. Call 100+ models in OpenAI format from a single point, hosted on our French servers — keys, costs, and access under your control.
What is LiteLLM?
LiteLLM est une passerelle IA (AI Gateway) open-source offrant un accès unifié à plus de 100 fournisseurs de modèles de langage — OpenAI, Anthropic, Gemini, Bedrock, Azure, Cohere, modèles locaux… — le tout via un format compatible OpenAI. Vous appelez n'importe quel LLM avec la même interface, et vous changez de modèle sans réécrire votre code.
Au-delà du routage, LiteLLM apporte une couche de gouvernance : clés virtuelles par équipe ou projet, suivi des dépenses (spend tracking), guardrails, load-balancing, logging et un dashboard d'administration. C'est l'outil pour centraliser et maîtriser tous les appels IA d'une organisation.
Côté technique, le proxy s'appuie sur Python, PostgreSQL en production et Redis en forte charge, avec des performances de l'ordre de quelques millisecondes de latence à un millier de requêtes par seconde. Déployé en conteneurs par DINAO et routé vers un modèle local, il devient une passerelle souveraine hébergée en France.
Host LiteLLM at DINAO
Resource tiers compatible with LiteLLM prerequisites (minimum 1 vCPU / 1 Go / 5 Go). Hosted in France, fully managed.
- 1 dedicated vCPU
- 2 Go RAM
- 20 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 2 dedicated vCPU
- 4 Go RAM
- 40 GB NVMe
- Daily backups
- Managed & monitored by DINAO
- 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
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…
Does LiteLLM host models or call them?
It's a gateway : it routes your calls to the provider of your choice, in OpenAI format. You can point it to a local model hosted by DINAO (sovereign) or to an external API.
Do my prompts stay in France?
The gateway, its keys, and its logs run on your instance in France. If you route it to a local model, your prompts do not leave; if routed to an external API, the content goes to the provider (see sovereignty note).
How to control costs and access?
LiteLLM allows creating virtual keys by team or project, tracking spending (spend tracking), setting quotas and guardrails — centralized governance of your AI calls.
Do I need to rewrite my code to change models?
No. All models are exposed in OpenAI format : you change provider or model via configuration, without modifying your application.
Can the gateway handle the load?
Yes. With PostgreSQL and Redis, LiteLLM handles high traffic (around 1000 req/s) with low latency. Higher tiers add load-balancing and high availability.
