Run and publish

Choose how to run the application separately from how to publish its documentation. GitHub Pages serves static documentation; the Python application needs its own process and persistent data home.

Option	What it runs	Start here
Source + uv or pip	UI, CLI and MCP from the checkout	Setup and deployment
Python wheel	Packaged launchers and application assets	Wheel installation
Docker	Linux container with a persistent /data volume	Container setup
Standalone bundle	Host-native directory distribution	Build and release
GitHub Pages	Searchable public documentation only	Publish the documentation

Source version 0.1.0 is package metadata, not a claim of a published release. Build artifacts must be verified outside the checkout; native executables need target-platform testing. The build guide records dated evidence and remaining gates.

Before operating a data home

Use an explicit --home when switching between launchers. One process owns each DuckDB home. The UI binds to loopback by default and is intended for one trusted local user. Externally bound MCP HTTP needs authentication, workspace scope and a TLS reverse proxy; it does not add multi-user authentication to the Streamlit application.

Supply credentials at runtime. Keep originals, databases, exports and backups out of source control and static hosting. General cleanup requires a successful full-home backup; restore archives into a new destination. See data management and security before cleanup, upgrades or recovery.