Skip to main content
This guide runs Fleet RLM locally with the shipped SQLite configuration. You don’t need PostgreSQL to try Fleet.

​
Prerequisites

  • Python 3.11, 3.12, or 3.13
  • uv
  • Node.js 22.19 or later and pnpm, for the terminal client
  • A Daytona account, API key, and organization ID
  • A key for the model provider named in config/fleet.toml. The shipped configuration uses the Databricks Unity AI Gateway.
1

Install

Clone the repository, install the Python environment, and install the terminal client.
2

Add credentials

Copy the example environment file.
Open .env and set these values:The example already sets FLEET_DATABASE_URL to a local SQLite file. Keep .env out of Git. To use a different provider, see Configure models.
3

Initialize the database

Fleet doesn’t migrate the database on startup. Apply the Alembic migrations once.
4

Start Fleet

fleet cli starts the backend, waits for it to become ready, and opens the terminal client. Try a prompt such as “Use Python to calculate the first 20 Fibonacci numbers and explain the result.” Type /help to see every command.
5

Resume the Session later

Fleet prints the Session ID. Pass it after -- to return to the conversation.
If the first live Turn fails, run uv run fleet doctor daytona. It creates a disposable Sandbox, checks your credentials, database, and Volume mount, and prints an action for each failure. See Troubleshooting.

​
Run the backend without the terminal

Start only the FastAPI backend when you want to call the API from your own client.
The backend listens on http://127.0.0.1:8000 and has no caller authentication. See the HTTP API reference.

​
Next steps

Last modified on September 4, 2026