Trace and evaluate your AI applications with confidence
A LLM observability platform powered by Phoenix (Arize), installed and maintained by DINAO. Your traces, prompts, and evaluations remain on our French servers — never with a third-party provider.
What is Phoenix?
Phoenix, developed by Arize AI, is an open-source observability and evaluation platform for artificial intelligence applications. It allows you to trace every call to a large language model with its full context — prompts, responses, latencies, costs, agent steps — to understand, debug, and ensure the reliability of your LLM pipelines and agents.
Built on the OpenTelemetry standard (OpenInference), Phoenix can be instrumented in just a few lines and natively supports major providers (OpenAI, Anthropic, Google, AWS Bedrock…) and frameworks (LangChain, LlamaIndex, DSPy). It includes built-in LLM evaluators, dataset management, experimentation, and embedding analysis to measure response quality and track hallucinations.
Designed for self-hosting (SQLite or PostgreSQL, Docker/Kubernetes deployment), Phoenix is the ideal tool to maintain control over often sensitive traces. Managed by DINAO, it offers you complete AI observability without ever exposing your application data to a third-party SaaS.
Host Phoenix at DINAO
Resource tiers compatible with Phoenix 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 might be wondering…
What is Phoenix used for?
To understand what is happening in your AI applications : every call to an LLM is traced with its prompts, response, latency, and cost. You can then evaluate quality, debug a drifting agent, and detect regressions.
My traces contain sensitive data: are they protected?
Yes. Self-hosted at DINAO, Phoenix keeps your traces, prompts, and responses on your instance in France. Nothing is sent to a third-party observability SaaS, unlike cloud offerings.
Where is the data hosted?
On DINAO infrastructure in France, in one of the available data centers. Your traces and evaluations do not leave the territory.
Do I need technical skills?
Instrumenting your code (a few lines via OpenTelemetry) requires a technical team. However, DINAO fully manages the installation, database, security, and updates for Phoenix.
Does Phoenix work with any model?
Yes. Phoenix is agnostic : it instruments both OpenAI, Anthropic, Google, or Bedrock and your local open-weight models, and integrates with LangChain, LlamaIndex, DSPy, and other frameworks.
