Run it locally

Get the factory running on your own machine, to try it or to work on it.

Before you start

• Python 3, and a clone of the repository.
• Run every command from the repository root. The default setup needs no account, no warehouse and no API key.

Steps

1. Create a virtual environment and install the dependencies:

python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

2. Load some data. showcase loads sample source systems and several data products already in flight, which is the better choice for exploring the web app; demo walks the worked example from empty to a live release.

.venv/bin/python -m dpf showcase
# or: .venv/bin/python -m dpf demo

Both write to var/, created on first use and ignored by git. python -m dpf reset deletes it, to start over.

3. Run the web interface:

.venv/bin/python -m dpf serve

It listens on http://127.0.0.1:8200. Set PORT to use a different one.

4. Optional — let the agents draft. Without ANTHROPIC_API_KEY, the three agent steps (requirements, the data model, the data contract) show their Run the agent button disabled. Every step except the first section of the data contract v0.1 for a system can still be done by hand — see Known limitations. To turn them on, put the key in a .env file at the repository root:

echo 'ANTHROPIC_API_KEY=sk-ant-…' >> .env

.env is read once per process and never overrides a key already exported in your shell. It is in .gitignore and never gets committed.

5. Run the tests:

.venv/bin/python -m pytest -q

The suite is offline: no network call, no API key, no real warehouse. Tests marked llm, which call the Anthropic API, are deselected by default; tests marked databricks skip themselves unless a real workspace is configured.

Check that it worked

showcase prints the products it built and how far each one walked; serve prints the URL it is listening on. Open it, and the sidebar lists the data products you loaded. pytest -q ends with every test passing.

If it does not work

• ModuleNotFoundError on startup. The virtual environment is not active, or the install step did not finish. Re-run .venv/bin/pip install -r requirements.txt and use .venv/bin/python, not whatever python resolves to on your PATH.
• Port already in use. Something else is on 8200: PORT=8300 .venv/bin/python -m dpf serve.
• Run the agent stays disabled after adding the key. Restart serve — .env is read once per process, not on every request.

Related

• Deploy — putting this on a server instead of a laptop.
• Your first data product — walking the showcase data step by step.
• Command line — every command, generated from the parser.