Skip to content

Wizard v1.0.1

Choose a tag to compare

View all tags
@github-actions github-actions released this 25 Aug 15:14
· 29 commits to master since this release
v1.0.1
61f7a80
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Wizard v1.0.1 — analytics, task routing, Arrow streaming & security hardening

Wizard is a local-first autonomous data analysis agent. Ask a real question about your data in plain language; it investigates — looking, computing, revising its approach when the data disagrees with it — then verifies the result and explains it, streaming its reasoning as it happens. Two small Ollama models are enough to run it, nothing is sent anywhere by default, and cloud providers are supported too, gated behind an explicit data mode.

Get started

Download the zip for your platform below, extract it, and run:

./cli/wizard init       # checks prerequisites, installs dependencies
./cli/wizard start      # launches everything, opens a browser

No git clone, no Go toolchain, no build step — the wizard binary and two sample datasets (workspace/dataset.csv, workspace/housing.csv) are already in the zip. Prefer containers? docker compose up --build -d works from the same extracted folder.

See the README for configuration, model choice and troubleshooting.

What's new in v1.0.1

  • Polars Engine Integration: Added Polars support alongside DuckDB and Pandas for fast, multi-threaded DataFrame processing on large datasets (#113).
  • Smart Tiered Task Router: Deterministic classifier for lightweight, standard, and reasoning-heavy turns with safe downscaling to smaller installed models to reduce latency and API quota usage (#117).
  • Zero-Copy Apache Arrow Streaming: Binary Arrow IPC streaming endpoint (/api/workspace/stream-arrow) and frontend decoder for fast, memory-efficient dataset previews (#114).
  • Security & Fuzzing: Added continuous property-based fuzz testing for AST code guards and file ingest headers, plus OpenSSF Scorecard token permissions (#104).
  • Performance & Caching: Full integration of Semantic Result Cache (#115) and Dynamic Skill RAG retrieval (#116).

Known caveat

On macOS, xgboost and lightgbm need the OpenMP runtime, which macOS does not ship: brew install libomp if modelling code fails to import either.

Full source, tests, CI and contributor docs live in this same repository — see CONTRIBUTING.md.

Assets 8
Loading