The simulation API needs to pass GCS dataset URIs for runtime dataset loading while keeping policyengine.py release manifests centered on canonical Hugging Face artifact references.
Today, policyengine.py resolves bundled datasets to HF URIs and passes remote strings directly into country package microsimulation constructors. For the US path, policyengine-us intercepts the dataset string before policyengine-core can handle gs://, so gs://policyengine-us-data/... fails instead of materializing to a local file.
Expected behavior:
- policyengine.py can materialize
gs:// dataset URIs to local files before country package construction.
- policyengine.py can still materialize
hf:// dataset URIs.
- US and UK dataset creation and managed microsimulation helpers pass local runtime sources to country packages while preserving canonical bundle provenance.
- Sim API can normalize manifest-derived HF URIs to GCS runtime URIs once this support is released.
The simulation API needs to pass GCS dataset URIs for runtime dataset loading while keeping policyengine.py release manifests centered on canonical Hugging Face artifact references.
Today, policyengine.py resolves bundled datasets to HF URIs and passes remote strings directly into country package microsimulation constructors. For the US path, policyengine-us intercepts the dataset string before policyengine-core can handle
gs://, sogs://policyengine-us-data/...fails instead of materializing to a local file.Expected behavior:
gs://dataset URIs to local files before country package construction.hf://dataset URIs.