Self-hosting
Lalabase on your own server.
Lalabase is open source under AGPL-3.0. You run it on your own server for free, with Docker Compose and every feature of our cloud.
// What you need
A server, Docker and a domain.
- At least 2 vCPU, 4 GB RAM and 20 GB SSD. For a larger team 4 vCPU, 8 GB RAM and 40 GB.
- Docker 24 or later with Docker Compose v2.
- A domain pointing at the server. The bundled Caddy fetches the TLS certificate itself.
- For the AI features, a key of your own or a model of your own. Without one the AI stays off and everything else runs.
Five commands to a running instance
git clone https://github.com/lalabase/lalabase.git && cd lalabase
git checkout "$(git tag --list '[0-9][0-9][0-9][0-9].[0-9]*.[0-9]*' --sort=-v:refname | head -1)"
cp .env.docker.example .env
docker compose up -d --build
docker compose exec web bin/rails lalabase:setupThe guide explains every step, the values in .env and running the instance afterwards. Read the installation guide
// Cloud or your own server
The same code, two ways to run it.
| Question | Your own server | Our cloud |
|---|---|---|
| Cost | €0 for the software. You pay for the server and the operation. | A monthly price from the price list, cancel monthly. |
| Features | All of them. New features arrive with the next release. | All of them, always up to date. |
| Operation, updates, backups | You. The guide covers updates and backups. | Us. Backed up every night, encrypted, in a second location. |
| Where the data lives | On your server, wherever it stands. | In a data centre in Germany, with a DPA. |
| AI | Your own key: OpenAI, Anthropic, EU providers such as Mistral or IONOS, or a model of your own through an OpenAI-compatible interface. | OpenAI, Gemini or Mistral, off by default. |
| Help | Bugs and ideas as an issue on GitHub. We help with the installation on request. | By email and phone, on weekdays. |
// Honestly
What you should know
Self-hosting means Lalabase is yours, and so is running it.
- Servers, updates and backups are yours. We do not see your instance and cannot step in when it fails.
- Without a key or a model of your own, the AI features stay off. Everything else runs.
- New features appear in our cloud first and on your server with the next release.
- We do not accept pull requests. Bugs and ideas are welcome as issues.
- If you change Lalabase and offer it to others over a network, the AGPL requires that they get the source code of your version.
// Frequently asked
Common questions about self-hosting
- What does Lalabase cost if I host it myself?
- The software costs nothing, it is free on GitHub under AGPL-3.0. You pay for the server, the time it takes to run it and, if you use the AI, your AI provider.
- Does the self-hosted version lack features?
- No. It is the same code as our cloud. New features arrive with the next release.
- May I run Lalabase for my clients?
- Yes, commercially too. If you change the code, the AGPL requires that the users of your instance get the source code of your version.
- Can I use a local language model?
- Yes. Through the OpenAI-compatible interface you connect a model running in your own network. Without a model and without a key the AI stays off.
- Do you help with the installation?
- Yes, on request. Write to us and we will go through the installation together.
// Get started
The guide has everything else.
Installation, configuration, AI keys, backups and updates, in English and German.