Skip to content

Start with Docker

Docker Compose starts DeltaLLM and the services it needs. This is the easiest way to try DeltaLLM on your computer.

Before you start

You need:

  • Docker with Docker Compose v2 or later
  • Git
  • an OpenAI API key for the sample model

1. Download DeltaLLM

git clone https://github.com/deltawi/deltallm.git
cd deltallm

2. Copy the starter configuration

cp config.example.yaml config.yaml

The starter configuration includes an OpenAI model named gpt-4o-mini. To add that model to the database the first time DeltaLLM starts, make sure config.yaml contains:

general_settings:
  model_deployment_source: db_only
  model_deployment_bootstrap_from_config: true

After the first successful start, you can change model_deployment_bootstrap_from_config back to false.

3. Add your secrets

Create a .env file in the repository root with these values:

DELTALLM_MASTER_KEY=replace-with-a-generated-master-key
DELTALLM_SALT_KEY=replace-with-a-generated-salt-key
OPENAI_API_KEY=replace-with-your-openai-key
PLATFORM_BOOTSTRAP_ADMIN_EMAIL=admin@example.com
PLATFORM_BOOTSTRAP_ADMIN_PASSWORD=replace-with-a-strong-password

Generate the master key and salt instead of writing them yourself:

python3 -c 'import secrets; print("DELTALLM_MASTER_KEY=sk-" + secrets.token_hex(20) + "A1")'
python3 -c 'import secrets; print("DELTALLM_SALT_KEY=" + secrets.token_hex(32))'

Copy the generated values into .env. Keep this file private and do not commit it to source control.

The admin email and password create the first account you can use to sign in to the Admin UI.

4. Start DeltaLLM

docker compose --profile single up -d --build

The first start may take a few minutes while Docker downloads and builds the images. When it finishes, DeltaLLM is available at http://localhost:4002.

Docker also starts PostgreSQL and Redis. You do not need to install or configure them separately.

5. Check that it works

Check the service:

curl http://localhost:4002/health/liveliness

You should receive:

{
  "status": "ok"
}

Now continue to Check or add your first model.

If DeltaLLM does not start

View the container logs:

docker compose --profile single logs deltallm

Check that:

  • the values in .env are not placeholders
  • the master key is at least 32 characters and contains letters and numbers
  • the OpenAI API key is valid
  • port 4002 is not already in use

For multi-instance testing, operational limits, and production guidance, see Docker deployment.