Your data and privacy

Datalumo runs on your server, under your model keys. This page explains exactly what is stored where, and what leaves your server, so you can answer your own security questions with confidence.

Where your content lives

Your documents and their embeddings are stored only on your server, in your database (SQLite file or Postgres). Datalumo does not copy them anywhere, and we cannot see them. There is no Datalumo cloud holding your content.

What leaves your server

One thing leaves, and only when someone searches or chats: the question and the matching passages Datalumo retrieved to answer it. These go to the AI provider you chose, on your key, so the model can write the answer.

You control this completely:

• You pick the provider (OpenAI, Anthropic, Mistral, Gemini, or xAI), or a local model.
• You use your own account and key, so the provider's data terms apply to you directly.
• If a provider offers a no-training or zero-retention tier, you can use it.

Datalumo is never in that path. The request goes straight from your server to the provider.

What we can and can't see

• We can't see your documents, your queries, your answers, or your users.
• We do see a license check. Your instance contacts the Datalumo portal to validate your license. That call carries your license and activation identifiers, not your content.

Going fully private

If you cannot send anything to a third-party provider, run a local model. Datalumo's custom provider talks to any OpenAI-compatible endpoint, so you can point it at a model running on your own hardware (Ollama, vLLM, LM Studio) for both chat and embeddings. Then nothing leaves your server at all, not even the query. See Configuration.

In short