| English | 中文 |
This offline example uses the existing gcam_aggregate_provinces tool and unchanged calculation. It does not download data or start a model. Prepare a private assets/config/auth root using the configuration guide; even the toy example writes generated files outside Git.
From the repository root:
python -m pip install -e '.[dev,tools]'
python -m runtime.cli run-tool gcam_aggregate_provinces --project-id toy_example --stage 01_socioeconomic_natural_change --params-file config/examples/run-tool/toy-aggregate.json --dry-run
python -m runtime.cli run-tool gcam_aggregate_provinces --project-id toy_example --stage 01_socioeconomic_natural_change --params-file config/examples/run-tool/toy-aggregate.json
The public input contains four synthetic rows. The reference has one row: weighted_price = 40/3, quantity_sum = 3 and province_count = 2. Compare parsed numerical values; CSV encoding and float formatting may differ.
Your result is under ${AIRMOSAIC_ASSETS_DIR}/results/toy_example/<run_id>/01_socioeconomic_natural_change/gcam_aggregate_provinces/. Its process receipt is under the same run’s records/ directory. Each invocation uses a fresh run ID. The runner does not overwrite public input/reference files or register its result as reusable data. Human review and registration remain separate.