Decision System Review Studio

This repository contains the deployable dashboard version of Decision System Review Studio, a user-facing visual analytics project built on top of RLVA.

The dashboard helps users review controller behavior across five scenarios:

• Traffic signal control
• Inventory planning
• Service queue management
• CartPole control QA
• LunarLander robotics safety

The GitHub version includes:

• The dashboard source under rlva/src/
• The report assets and benchmark outputs required by the dashboard
• The full 1GB course dataset, stored in GitHub-safe split archive parts under rlva/outputs/course_dataset/

It does not include local virtual environments, caches, or unrelated course materials.

Local Deployment

Use Python 3.10+.

From a fresh clone:

git clone https://github.com/volowang/Playground.git
cd Playground
python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt
export PYTHONPATH=.
streamlit run rlva/src/app.py

If .venv already exists locally, start from the activation step:

source .venv/bin/activate
export PYTHONPATH=.
streamlit run rlva/src/app.py

Open the Streamlit URL shown in the terminal, typically http://localhost:8501.

What To Show In The Demo

1. Choose one of the five scenario buttons at the top.
2. Click a point in any chart to see what that point means and what system problem it may indicate.
3. Compare controllers in the lower charts.
4. Export a decision memo from the Save / Export section.

Dataset Delivery

GitHub does not allow a normal Git blob larger than 100MB, so the required 1GB dataset is committed as split archive parts:

• rlva/outputs/course_dataset/rlva_trace_corpus_close_to_1gb.csv.gz.part-*

The app automatically reassembles these parts into a local gzip archive on first use and reads the dataset from there. You do not need to manually unzip or rebuild the dataset before launching the dashboard.

Project Layout

rlva/
  src/                       dashboard and support code
  outputs/
    benchmark/               benchmark summaries, traces, figures, reports
    course_dataset/          1GB dataset archive parts and metadata
    figs/                    comparison HTML artifacts used by the dashboard
    report_assets/           supporting markdown/csv assets
    summaries/               precomputed summary tables
    tables/                  precomputed comparison tables

Notes

• The repository is optimized for opening the dashboard locally.
• Training environments, model-building toolchains, and checked-in virtual environments are intentionally excluded from the GitHub deployment version.