Open Report › Self-hosted AI analytics
AI analytics that runs on your servers
Open Report is an open source reporting tool with an AI assistant, and you choose where both of them run.
Your model, not ours
- A local model through Ollama or LM Studio: nothing leaves your network.
- Any OpenAI-compatible API, Anthropic, or your own Claude subscription.
- One model for the whole instance, or a personal one per user.
You decide what the AI reads
- Schema only: the AI sees field names and labels, never a figure.
- Schema and cached data: it also reads the aggregates already cached for your reports.
- Query results: it reads up to 50 rows per answer, as the user, with their rights.
The AI proposes, you apply
The assistant has no write access: every chart, report, measure or join it suggests comes as a card you check and apply yourself.
Open source, and it stays that way
The code is published under the GNU AGPL-3.0: you can read what is sent to the model, run it at work, and keep the version you installed.
Running it
One Docker container on any machine, a Raspberry Pi included; or the hosted Cloud, in France, under EU law.
Questions
Can I use Open Report with Ollama?
Yes. Point the assistant to your Ollama or LM Studio server; any model with tool calling works, and the settings page tests it for you.
Does my data go to OpenAI or Anthropic?
Only if you choose one of them, and only what the sharing level allows. With a local model, nothing leaves your servers.
Is the AI assistant part of the free edition?
Yes. The self-hosted Community edition includes every feature, the AI assistant too, with no user limit.
What does the AI see under row-level security?
What the user sees: it reads through the same query route, with the same rights and the same row-level rules.
See also AI data analysis · AI report generator · vs Superset