Deployment

Deploy Marivo where the agent runs

Marivo is a Python library, not a hosted service. Install it in the environment where the agent runs, then check the project’s datasource and semantic declarations into version control. Provide datasource secrets as environment variables referenced by *_env fields; do not commit credentials into the project.

For development from this repository, create an environment, install the editable package with the needed extras, and verify the import:

python3 -m venv .venv
.venv/bin/pip install -e ".[dev,duckdb,trino]"
.venv/bin/python -c "import marivo; print(marivo.__name__)"

After install, run marivo doctor from the selected environment to verify the Marivo version, Python executable, loaded package path, and project state. If the environment is not activated, invoke the marivo executable belonging to that environment. The CLI is a bootstrap check only. After entry, use Python for focused live contracts:

import marivo

marivo.help("datasource.<target>")
marivo.help("semantic.<target>")
marivo.help("analysis.<target>")

For a project using a published release, install the required backend extra as described in Installation, then run marivo init in the project directory. Keep the committed semantic layer alongside the agent so every run resolves the same declarations.