Skip to main content

Operations

RecallRx is designed to run as a scheduled static data pipeline.

Daily update​

The dataset update workflow runs the Python collector, validates the generated JSON, and commits changes only when the data/ directory changes.

python -m recallrx build --output data
python -m recallrx validate data

The collectors use Python HTTP requests rather than wget, so they can set a clear user agent, retry transient failures, page through authority search results, parse HTML, and fall back to PDFs when needed.

Deployment​

The application deployment workflow builds dist/ from site/ and data/. The documentation workflow builds docs-site/ separately so docs can be validated without replacing the search app deployment.

Manual checks​

When a scheduled update changes many records, inspect:

  • data/countries/<country>/build-report.json
  • rejected candidate counts
  • parser warnings on accepted records
  • changes to recalls-summary.json
  • any new missing product-code, lot, date, or reason fields