Self-hosted RAG

Search & chat your data.
Self-hosted, on your keys.

Vectorize your documents and data, then add search and chat to any site with one line of HTML, or build your own on the API. Bring your own model keys, or run a fully local model so nothing leaves your server. One command, on your own machine.

Join the beta Read the docs

Private beta — join for early access.

Planned pricing: $299 first year, then $99/yr.

Not ready to buy? See it live below, or ask a question.

1: Docker container
5: Model providers + local
0: Token markup

install.sh

$ curl -fsSL https://get.datalumo.app/install.sh | bash

// drop-in widgets

Search & chat, in one line

One script tag, one element, scroll through both right here. Switch layouts, or peek at the exact code. Prefer to build your own UI? Every widget runs on the same HTTP API and headless client.

datalumo ·

The chat bubble is live — look in the bottom-right corner ↘

Search

Drop a search box anywhere — or a ⌘K command palette. Instant, typo-tolerant results grounded in your own content.

Inline box or ⌘K modal
Semantic + keyword, ranked together
One line of HTML, your own keys

Read the widget docs →

Chat

A grounded chat assistant your visitors can talk to. Embed it inline in the page, or as a floating bubble in the corner.

Inline panel or floating bubble
Cited answers, grounded in your data
Streams responses in real time

Read the widget docs →

// what you get

Everything to ship private AI

Ingest your content, then serve it through whatever fits, a full HTTP API and CLI for builders, ready-to-go widgets for everyone else.

Hybrid search

Semantic + keyword retrieval, ranked together with citations.

Grounded chat

RAG chat that streams answers cited back to your sources.

HTTP API

Build your own search and chat on a clean, documented API.

CLI & SDKs

Ingest and query from the terminal, PHP, WordPress or Laravel.

Drop-in widgets

Search box, command palette, chat bubble, one line of HTML.

Your infrastructure

One container on your server. Your keys, your data, no markup.

Run it fully offline

Point it at a local model (Ollama, vLLM) for chat and embeddings. Nothing leaves your server, not even the question.

Ingest anything

Upload files, sync an S3 bucket or folder, or push from your app. PDFs, Word, Markdown, HTML, and text.

Many sites, one instance

Separate indexes and apps, each with its own content, keys, and widget. Power several sites or sections from one box.

// analytics

See what your AI is doing

Every query, answer and token is yours to inspect, plus visitor feedback and conversation insights. Track usage per app, watch latency and grounding quality, and find the questions your content doesn't answer yet.

Token usage per app and per key
Thumbs-up/down feedback on every answer
Conversation insights and top questions
Usage quotas and rate limits you control

datalumo · analytics

Queries

24.8k

Tokens

1.2M

Helpful

94%

Avg latency

180ms

Queries · last 7 days

+12%

Runs anywhere Docker does

A laptop, a box in the closet, or a cheap cloud VM. Datalumo itself needs no GPU; it handles indexing, search and serving, and your model provider does the inference. (Running a model locally on the same box is the exception, that needs its own hardware.)

Usage	RAM	CPU	Storage
Light	2 GB	1	20 GB
Moderate	4 GB	2	40–80 GB
Large	8 GB+	4+	80 GB+

No GPU required

SQLite for small sets, Postgres + pgvector when you scale.

Your favorite provider

OpenAI, Anthropic, Mistral (EU), Gemini, xAI, or a local model on your own hardware.

// faq

Frequently asked questions

What does a license include?

A valid license gives you the Datalumo Docker image and its updates, set up via the installer. Stop renewing and you keep the current version forever, you just stop getting updates and support. One instance per license unless it's a multi-instance license; reselling is not allowed.

Can you help me set it up?

Happy to. There's no managed or hosted version yet, so Datalumo runs on your own server, but if you'd like a hand, we offer paid setup: point us at a fresh DigitalOcean (or similar) box and we'll install and configure it for you. For anything more involved, reach out and we'll work out a custom quote.

What support is included?

Support is intentionally light-touch. The documentation is the first stop, and with a valid license we ship updates and follow up on genuine Datalumo bugs. We can offer best-effort guidance, but support does not include hands-on server administration, we won't log into your infrastructure or fix issues unrelated to Datalumo. Need hands-on help? That's the paid setup arrangement above.

Is AI usage included?

No. You bring your own provider keys (OpenAI, Anthropic, Gemini, Mistral, xAI) and pay them directly, with no markup from us. Or run a local model and pay nothing per query. Datalumo uses both embedding and LLM models; Anthropic and xAI are LLM only.

Is it open source?

Our tooling, the PHP SDK, WordPress plugin and Laravel package, is open source on GitHub. Datalumo itself is not.

Is it GDPR / data-residency friendly?

You control where data lives and which provider receives your queries for embedding and chat. Or run a fully local model, and nothing leaves your server at all, not even the question. Datalumo only "phones home" to check your license, which can be disabled, and sends nothing beyond what that check needs.

Does it scale?

Yes. Use the SQLite path for small setups, or the Postgres path with external Postgres and Redis for higher loads.

How do updates work?

With a valid license you're entitled to updates. The installer can enable automatic updates, or you can update manually.

What is the refund policy?

Full refund if requested within 14 days of purchase.

How do I ingest my data?

Through the API, the CLI, S3-compatible storage or file uploads, plus helpers like the WordPress plugin, Laravel package and PHP SDK.

Private AI, on your own server.

One container, your keys, no token markup. Set it up in an afternoon.