What the pipeline actually produced
FoodSafe India ingests food-safety enforcement records from public sources and computes district- and brand-level risk scores from them. Everything below is the real, unedited output of running that pipeline once — not a mockup or a design placeholder.
Real US FDA recalls ingested
Fetched live from accessdata.fda.gov (no API key) by
pipeline/sources/openfda.py and written to the
database by the same code path production ingestion uses. Every source
link below is real and dereferenceable.
Real Indian districts added
Fetched live from data.gov.in's AGMARKNET commodity-price API by
pipeline/sources/agmarknet.py. Geographic
reference data, not enforcement records — AGMARKNET carries commodity
prices, not safety violations.
Computed district risk scores
Output of models.aggregate run over the
database above — real statistical computation (fail rate, Wilson
confidence interval), not hand-entered numbers. Some rows reflect a
handful of old placeholder records already present in the local
database before this run, not exclusively the openFDA data above; see
data/README.md.
Pipeline run log
Actual rows from the pipeline_runs table,
written by pipeline/run_and_log.py — the same
logging path the (currently broken) cloud ingestion uses.