How mckinsey/vizro works
Project type and purpose
• This is a monorepo containing three interrelated Python packages:
• vizro-core – the low-code dashboard framework (Pydantic models + Plotly/Dash integration).
• vizro-ai – an extension that layers LLM-powered chart/dashboard generation on top of vizro-core.
• vizro-mcp – a minimal server component that can host or orchestrate Vizro dashboards (e.g., multi-component platform).
• It includes docs (MkDocs/Sphinx), examples, CI/CD (GitHub Actions), Docker devcontainer, and release tooling.
File structure
• Top-level folders:
• .devcontainer – Dockerfile + setup for VS Code dev containers.
• .github – Issue templates, GitHub Actions workflows.
• tools – release scripts, helper utilities (e.g., PyCafé link checker).
• vizro-core, vizro-ai, vizro-mcp – each follows the same layout:
– src/<package> with Python source
– tests (unit, integration, e2e)
– docs (MkDocs or ReadTheDocs config)
– examples & changelog directories
– packaging config (pyproject.toml, hatch.toml)
• Patterns:
• Monorepo with sibling packages
• Pydantic model definitions → code generation → Dash + Plotly runtime
• CI pipelines per package + monorepo‐wide workflows
• Devcontainer to unify development environment
README insights
• The README emphasizes: low-code configuration via Pydantic/JSON/YAML, built‐in best practices, extensibility via Python/JS/CSS, production deployment with Plotly/Dash, and an LLM “AI” layer for natural-language generation.
• It cites core tech: Python 3.9–3.13, Plotly, Dash, Pydantic, Hatch (build), MkDocs/ReadTheDocs.
Be detailed but modular
• Five major areas: Developer tools, Core library, AI extension, Server, CI/CD & publishing.