Skip to main content

Quick Start

Get your first AI Gateway route running and callable in under 5 minutes.

Prerequisites​

  • Docker (with Compose support)
  • An API key for at least one supported provider (e.g. OpenAI)

1. Start the Gateway​

Clone the repository and start the stack using pre-built images:

git clone https://github.com/radicalbit/radicalbit-ai-gateway
cd radicalbit-ai-gateway

Create a secrets.yaml file in the root directory with your provider API key:

secrets.yaml
OPENAI_API_KEY: sk-your-key-here

Then start the gateway:

GATEWAY_TAG=latest docker compose up -d

This starts the full stack: gateway, database, cache, and observability services. Wait a few seconds for the services to be healthy, then open http://localhost:9000 — you should see the Gateway UI.

2. Create a Project​

In the UI, create a new project and give it a short name, e.g. quickstart.

The project starts in DRAFT state. You will load your configuration into it next.

3. Write Your Configuration​

In the project editor, paste the following configuration:

chat_models:
  - model_id: gpt-4o-mini
    model: openai/gpt-4o-mini
    credentials:
      api_key: !secret OPENAI_API_KEY

routes:
  my-assistant:
    chat_models:
      - gpt-4o-mini

This defines one model and one route named my-assistant.

4. Approve and Serve​

Click Load to load the configuration into the project, then Approve, then Serve.

Your route is now live. You will see it listed in the Routes section of the UI.

5. Create a Group and API Key​

Go to Groups and create a group (e.g. my-team). Then go to API Keys, create a key, and associate it with the my-assistant route inside the quickstart project.

warning

The key is shown only once — copy it and store it securely.

6. Make Your First Call​

curl http://localhost:9000/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "quickstart/my-assistant",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

You should get a response from the model. That's it — your gateway is running.

Next Steps​