Maritime Safety Mesh Simulation

Reproducible multi-agent maritime weather mesh simulation package.

Quick Start

make clean
make install
make test
make gui

# Run the full experiment suite
uv run python -c "from maritime_mesh.experiment.runner import ExperimentRunner; ExperimentRunner(n_seeds=30, max_workers=8, process_max_tasks_per_child=1).run_all()"

First Successful Run Checklist

1. Install and verify:
   • make clean
   • make install
   • make test
2. Generate baseline experiment outputs:
   • uv run python -c "from maritime_mesh.experiment.runner import ExperimentRunner; ExperimentRunner(n_seeds=30, max_workers=8, process_max_tasks_per_child=1).run_all()"
3. Confirm artifacts in your output directory (default: outputs/maritime_mesh/):
   • summary.csv (required for Results/Hypothesis pages)
   • *.parquet run logs (required for Map Playback)
   • *.manifest.json run manifests (run metadata and schema checks)
4. Launch dashboard:
   • make gui
5. In the dashboard:
   • verify the selected output directory points to the same run artifacts,
   • open Results to confirm KPI tables/charts are visible.

CLI and GUI Run Parity

• The CLI runner (maritime_mesh.experiment.runner.ExperimentRunner) and dashboard Run page write the same artifact contract (summary.csv, parquet logs, manifests).
• Prefer CLI for scripted/reproducible batch runs and CI workflows.
• Prefer GUI for iterative setup tuning (lanes, shore stations, weather/comms settings) and visual preview before launch.

Project Structure

• src/maritime_mesh/ - simulation package
• tests/unit/ - module-level unit tests
• tests/integration/ - scenario and runner integration tests
• outputs/maritime_mesh/ - generated summaries and per-run parquet logs