Investigate GitHub Release assets for pre-built DSSTox and ECOTOX databases

[seanthimons]

Summary

Both dss_install() and eco_install() currently build their DuckDB databases from scratch (DSSTox from CompTox API, ECOTOX from EPA FTP ASCII dump). This works but is slow and requires API keys / large downloads.

GitHub Release assets support files up to 2 GB. Both databases fit comfortably:

• DSSTox: ~15M rows, estimated moderate size
• ECOTOX: ~424 MB

Proposal

Investigate hosting pre-built .duckdb files as GitHub Release assets so dss_install() / eco_install() can offer a fast-path download instead of building from source every time.

Questions to resolve

• What's the actual built size of dsstox.duckdb?
• Can we version the databases by EPA release date (e.g., ecotox-2026-03.duckdb)?
• Should *_install() default to download (fast) with a build = TRUE flag for from-source, or the other way around?
• Any licensing concerns with redistributing EPA-derived databases as release assets?
• CI workflow to automate building and attaching assets on release?

References

• DSSTOX_MERGE_PLAN.md — Section 3 (Database Build Pipeline), Option C
• ECOTOX_MERGE_PLAN.md — Section 3 (Database Build Pipeline), Option B