Quick start

The fastest way to a running installation is the Docker Compose stack. It starts every service, generates the secrets on first run, and configures the Keycloak realm by itself.

On this page

Prerequisites
Starting up
Accounts in the local stack
What sits behind the single address
What works right away — and what does not
Managing and shutting down

Prerequisites

Docker Desktop, running — the whale icon in the menu bar must be steady.
openssl, curl and jq for the helper scripts. On macOS everything but jq is preinstalled (brew install jq).

Starting up

1
Change into the compose directory and start
On first run up.sh creates .env with random secrets, builds the console and agent images, starts the stack, and waits for the realm setup to finish.
```
cd deploy/compose
./up.sh
```
2
Wait out the first run
It pulls images and builds two applications, which takes a few minutes. These two commands show the progress.
```
docker compose ps
docker compose logs -f keycloak-init
```
3
Open the console
Everything sits behind one address. Login goes through Keycloak and returns to the console afterwards.
```
http://localhost:8080/
```

localhost, not 127.0.0.1

Only http://localhost counts as a secure context. Over 127.0.0.1, window.crypto.subtle is missing and the PKCE login cannot complete.

Accounts in the local stack

User	Password	Role
flaky-admin	flaky-admin-pw	Platform administrator — may set policies, users and the AI backend
flaky-tester	flaky-tester-pw	Ordinary user — useful for checking what a policy actually does

For the local stack only

The realm setup creates these two accounts so a fresh installation is usable at once. Anything beyond that follows the Kubernetes page.

What sits behind the single address

URL	Serves
http://localhost:8080/	Console
http://localhost:8080/api/agent	Agent backend
http://localhost:8080/auth	Keycloak (login)

What works right away — and what does not

Login, the SQL worksheet, the data browser, import, governance and search work immediately. Search needs no key: the embedding model ships with the platform. The chat, however, needs a language model, and none is bundled — without a backend it replies that none is configured. Setting one up is covered under Configuring the AI backend.

Managing and shutting down

Running the stack

docker compose ps              # status
docker compose logs -f agent   # follow one service's log
docker compose down            # stop, keep the data
docker compose down -v         # stop and wipe all data

If a docker command answers “Cannot connect to the Docker daemon”, Docker is off — that is not a fault of the platform.

Applies to: This page describes lavalake 0.16.0. Product and documentation live in separate repositories — when in doubt, what the console shows is authoritative.

Something missing?

If these pages do not answer your question, we will walk through the platform against your own case in half an hour.

Book a demo

Quick start

Prerequisites#

Starting up#

Accounts in the local stack#

What sits behind the single address#

What works right away — and what does not#