setup

name	setup
description	Set up everything needed to participate in the Viktor Challenge — environment, dataset, sanity checks, first baseline.

/setup — get a team from zero to a running baseline

Do these steps in order. Announce each result in one line.

1. Environment

• Confirm Python 3.10+ (python --version).
• Install the project dependencies with python -m pip install -r requirements.txt.
• No GPU or API key is needed for the analysis-only workflow.

2. Dataset

• No dataset link yet? python scripts/make_synthetic_sample.py writes a shape-identical synthetic sample to ./export/. Build your pipeline on it.
• Real dataset — obtain the current private download link and published checksum through the official challenge channel. Never store a signed URL in Git.
  1. Extract into ./export/: mkdir -p export && tar xzf trajectories_v1_01.jsonl.tar.gz -C export/ (scripts read any export/*.jsonl; the archive also contains the LICENSE — read it).
  2. Challenge use only, no redistribution. Never commit export/ or the archives to git (both are in .gitignore — keep them there).

3. Sanity checks

• python scripts/load_trajectories.py export/ — prints request counts, per-model distribution, reconstructed trajectory stats, and a schema check.
• Open one line yourself and confirm: exactly model, input, tools — no output, no usage. Models are anonymized (families , , across generations — ~9 ids); PII shows as named placeholders (, , ); images are redacted placeholders.