Source sets

When an environment's source database backend is CSV, you don't configure a database — you point at a source set, which is a named collection of source CSV files attached to the pipeline.

You can have multiple source sets per pipeline (e.g. synthea-1k-2024-01, synthea-10k-2024-06), and pick which one an env reads from on its environment page. Switching cohorts is a one-click operation.

Uploading

Pipelines → \<your pipeline> → Source sets → New → name the set → upload the CSV files. The orchestrator content-addresses each file by sha256 — re-uploading identical content costs no extra disk.

Each file shows up with size, upload time, and a delete button. You can update individual files later without re-uploading the whole set.

Filename conventions

The engine reads files by name. For Synthea, the orchestrator expects the standard names (patients.csv, encounters.csv, organizations.csv, etc.). If your upstream uses different names, either rename before upload or adjust the RiaH's source table names to match what you've got.

Re-runs

A source set is a per-pipeline reusable asset. Switching the env's source set between runs lets you re-run the pipeline on a different cohort without rebuilding any of the mapping config — useful for "does it still load cleanly when I throw 10× the patients at it" tests.