Observability for your AI applications, hosted in France
The black box of your LLM applications powered by Langfuse, installed and maintained by DINAO. Traces, costs and evaluations remain on our French servers — your prompts never leave for a third party.
What is Langfuse?
Langfuse is an open-source LLM engineering platform that enables teams to trace, debug, analyze, evaluate and scale their generative AI applications. It centralizes observability, prompt management and evaluation of LLM applications in one place, both in development and production.
Tracing covers every step of a call: model requests, retrieval, embeddings, agent actions, with multi-turn sessions, agent graphs and per-user tracking. Prompt management provides versioning, deployment and caching, while evaluations combine LLM-as-a-judge, code evaluators, user feedback and benchmark datasets to measure quality and detect regressions.
On the integration side, Langfuse is OpenTelemetry compatible, offers Python/JS/TS SDKs, a comprehensive REST API and over twenty native integrations (LangChain, LlamaIndex, LiteLLM, Haystack, OpenAI…). Multi-tenant (organizations, projects, roles), it relies on a distributed architecture — PostgreSQL, ClickHouse, Redis/Valkey and S3 object storage — deployable via Docker or Kubernetes.
Host Langfuse at DINAO
Resource tiers compatible with Langfuse prerequisites (minimum 4 vCPU / 8 Go / 30 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 might be wondering…
Is Langfuse an AI that answers my questions?
No. Langfuse does not run any models: it is an observability platform that traces, measures and evaluates the AI applications you develop. It monitors your LLMs, it does not replace them.
Do my prompts and traces go to a third party?
No. The entire Langfuse stack (PostgreSQL, ClickHouse, Redis, object storage) runs on your instance hosted in France. Your traces, prompts and evaluations stay with you.
Which frameworks can I connect it to?
Langfuse is OpenTelemetry compatible and offers Python/JS SDKs as well as over 20 native integrations: LangChain, LlamaIndex, LiteLLM, Haystack, Vercel AI SDK, OpenAI, and many more.
Why go through a managed hosting provider?
Self-hosting Langfuse requires four components (PostgreSQL, ClickHouse, Redis/Valkey, S3 storage). DINAO installs, updates, backs up and monitors everything: you focus on your AI applications.
Can I export my data or change plans?
Yes. You can upgrade or downgrade at any time, and your data remains exportable via the API. Langfuse is open-source (MIT license) — no proprietary lock-in.
