FDA Recall & Enforcement Delta Monitor

Monitors the openFDA enforcement API (drugs, medical devices, food) for a user-maintained watchlist of firms / products / product types and emits one dataset item (alert) per newly detected or newly amended enforcement action since the last run — "free when nothing changes."

Who it is for

Compliance, quality-assurance, supply-chain, pharmacy and hospital-procurement teams obliged to track recalls affecting products they carry, stock, dispense or regulate. Instead of manually re-polling openFDA enforcement reports on a schedule, run this Actor on a cadence and it only alerts on what is new or amended.

What makes it different

Existing Apify Store coverage of FDA recalls is tiny and all pay-per-record: mediapro-tools/fda-recalls-monitor ($0.002 per recall record + $0.01 per new recall alert), thoob/fda-food-recall-feed ($0.005 per record), veridra/ fda-recalls-monitor ($0.05 per record), plus factpipe/fda-recalls-monitor, nexgendata/fda-recall-monitor and cloud9_ai/openfda-drug-scraper. Every listing has 0 reviews, none combine a watchlist with a per-query checkpoint and delta alerts ("new since last run"), and pay-per-event pricing bills even on empty scans.

This Actor keeps a watchlist and remembers what it already reported, so repeat runs emit only genuinely new or amended enforcement actions — free when nothing changes.

Data source

Public openFDA enforcement data only, no API key required:

• https://api.fda.gov/drug/enforcement.json
• https://api.fda.gov/device/enforcement.json
• https://api.fda.gov/food/enforcement.json

No credentials, no proxies, no scraping of anti-bot pages, no commercial inference (no LLMs) inside the product.

Input

See actor/.actor/input_schema.json. All fields optional; the default input is:

{ "productType": "Drugs", "lookbackDays": 30, "maxResults": 25 }

• productType: "Drugs" (default), "Medical Devices" or "Food" — picks the openFDA enforcement domain.
• watchlist: optional array of firm/product strings matched case-insensitively against the recalling firm, product description, product type, recall number and recall reason. Empty means every enforcement action in the chosen domain is in scope.
• lookbackDays (1-365, default 30): how far back to look on the first run; repeat runs only emit actions newer than the last checked state.
• maxResults (1-200, default 25): cap on alerts emitted in a single run.

Output

One dataset item per new or amended enforcement action:

{
  "recallNumber": "D-0041-2027",
  "reportDate": "20261021",
  "productType": "Drugs",
  "endpoint": "https://api.fda.gov/drug/enforcement.json",
  "recallingFirm": "Houston Rx LLC",
  "productDescription": null,
  "reasonForRecall": null,
  "classification": null,
  "status": "Ongoing",
  "newSinceLastRun": true,
  "changeType": "new",
  "watchlistMatched": null,
  "source": null
}

source is null in emitted items — the Actor never guesses a value it did not receive from openFDA. The queried openFDA enforcement endpoint is carried in the endpoint field instead.

changeType is "new" for a previously unseen recall number and "amended" for a known recall number whose published content changed. Every field is null when it could not be parsed from openFDA — the Actor never guesses.

Checkpoint / delta behaviour

State (the last-checked reportDate high-water mark plus a content fingerprint per recall number) is kept in the Actor default key-value store. On each run the Actor queries openFDA for report_date >= checkpoint, sorted newest first, deduplicates by recall number, and only pushes previously unseen or content-changed actions. Resetting the store re-emits actions within the lookback window.

Limits

• Max 50 watchlist terms, 200 alerts per run, 365-day lookback.
• Up to 1,000 openFDA records scanned per run (sorted newest first, stops at the checkpoint).
• Checkpoints persist per productType + watchlist in the Actor key-value store.

Verification

Run pytest actor/tests before release (builder acceptance criterion).