v0.2.0

Rewritten as a practical A/B test checker.

• causalpilot analyze: sample ratio test, difference in means, CUPED, always-valid
  intervals for peeking, guardrails, and a ship / hold / extend call with reasons.
• causalpilot hillstrom: downloads and analyzes the MineThatData e-mail experiment.
• causalpilot simulate and benchmark/run_benchmark.py: synthetic experiments with a
  known effect, built on the four-segment simulator from 0.1.0.
• Browser version on GitHub Pages, checked against the Python numbers in CI.
• Removed the FastAPI state machine, the Next.js 3D control room, the paper and the
  rule-based "LLM" baselines.

Install: pip install git+https://github.com/appleweiping/causalpilot@v0.2.0
Browser version: https://appleweiping.github.io/causalpilot/

v0.1.0 (old simulator version, superseded)

First version: a simulated experimentation agent with a FastAPI backend and a 3D web control room. All of its numbers came from a simulator. It was replaced in 0.2.0 by a command-line and browser A/B test checker; see the v0.2.0 release and the README on main. The attached video and PDF describe this old version only.