Docglow replaces dbt Core’s built-in docs site with a single-page app you can generate and host anywhere. It reads your compiled dbt artifacts and spits out a static site—no dbt Cloud needed, no backend required. You get unlimited models and viewers, a drag-and-drop lineage explorer that shows upstream/downstream dependencies, and column-level tracing with transformation labels (direct, derived, aggregated). A built-in health score gauges coverage of descriptions, tests, and docs completeness. Installing and running Docglow takes two commands: • pip install docglow • docglow generate --project-dir /path/to/dbt --output-dir ./site Then serve locally with docglow serve --dir ./site. You can also bundle everything into one self-contained HTML file via --static. Optional features include full-text search, dark/light/auto themes, AI chat powered by your Anthropic key, and an MCP server for editor integration. Health checks run with docglow health, and you can gate CI pipelines using --fail-under. Configuration lives in an optional docglow.yml you scaffold with docglow init. You can tweak layer definitions, display settings, theme, and AI options there. The AI chat panel streams queries against embedded project metadata without shipping your key, and the MCP server lets local AI editors query model details over stdio. For large projects, skip column lineage or slim output to cut payload by up to 60%. Prewritten GitHub Pages, S3, GitLab CI workflows and a pre-commit hook simplify continuous deployment.