Wizard v1.0.2

Wizard is a local-first autonomous data analysis agent: point it at a CSV or a database, and it plans, writes, and runs Python in a sandbox to answer your question, showing its reasoning and code as it goes.

Get started

Download the zip for your platform below, then:

./cli/wizard init
./cli/wizard start

wizard init checks prerequisites and sets up the backend/frontend. See the README for the full quick start, including running from source or via Docker.

What's new in 1.0.2

• Gemini is now a first-class provider. wizard init --provider gemini --gemini-key AI... (or pick it from the Models page) — Gemini joins Anthropic, OpenAI, and any OpenAI-compatible gateway as a cloud backend.
• wizard init now sets up local, hybrid, or fully-cloud installs in one run, not just local Ollama:

  wizard init --provider anthropic --anthropic-key sk-ant-...
  wizard init --data-mode hybrid --anthropic-key sk-ant-...

  It writes straight into backend/.env, skips irrelevant local-model setup on a pure-cloud install, and pulls in the extra Python packages a cloud provider needs automatically.
• wizard status/doctor now also report which provider and data mode are configured.

Known caveat

On macOS, xgboost/lightgbm need Homebrew's OpenMP runtime: brew install libomp if you hit an import error running a modelling analysis. See the README's Troubleshooting section.

For the full source, tests, and architecture docs, see the main repository and CONTRIBUTING.md.