FDA Food Recall Supplier Match Feed

Match official openFDA food recalls to supplier and product watchlists with explicit candidate evidence and saved change tracking.

Quick start

Run with the default input for a working demonstration. The default uses the identified open data preset. Demonstrations are owner tests, not customer revenue.

{
  "monitorKey": "default"
}

For your own data, set rightsConfirmed to true only if you have permission to process it. Use the input descriptions for object fields and column mappings. URLs must be public HTTPS without credentials or secret query parameters. Website/feed requests respect robots.txt; APIs use their documented endpoints. No authenticated pages, anti-bot bypass or paid API is used.

Results and saved comparisons

OUTPUT in the run key-value store is authoritative JSON (records, diagnostics, summary and billing checkpoints). RESULTS.csv is a formula-protected export; DIAGNOSTICS.json explains failures and incomplete input. The dataset includes evidence records and nonbillable source diagnostics. Operator-stop and below-budget diagnostics are available in OUTPUT and DIAGNOSTICS.json. A green Apify run alone does not prove a source was healthy: inspect OUTPUT.status and diagnostics. Reuse monitorKey for comparisons, change it for an independent baseline. Customer state is stored in the Actor-specific named store in the customer's Apify account, bounded to 5,000 state entries / 4 MiB. Avoid concurrent runs with the same monitorKey. Source records outside the retained window may appear as baseline again.

Official food enforcement API, CC0. Text matches are candidates and can be ambiguous. Not a safety, legal-compliance or recall-completeness guarantee. A capped query is explicitly marked incomplete and never infers recall resolution from disappearance. FDA can update or delay source records. Before a recall number is assigned, a pending identity uses the official event ID and product-description hash; assignment of the recall number can appear as a new source record.

Price

Actor start: $0.01, charged by Apify per GB of assigned memory with a one-event minimum. Complete recall query: $0.030 for each successfully completed recall query. Completed healthy checks include zero matches or no changes. Rejected/incomplete source units are not charged the custom event; the start event still applies. For standard runs, the Actor pricing page states whether platform usage during the run is included in event prices. After the run, interactions with the dataset can incur standard Apify platform usage charges. Check Apify pricing and your account usage for details. Use maxTotalChargeUsd in the run options to bound your expense. The Actor stops before starting a unit that does not fit the remaining event budget. Results are saved before a custom charge. Uncertain charge/dataset submissions are not retried, preventing duplicate submissions; inspect OUTPUT and chargedEventCounts when a response is uncertain.

Limits and operation

Each run is bounded by the documented input limits and 180-second default timeout. Memory is capped at 256 MB. This is an evidence utility; review results before acting. Original INPUT and run outputs remain in your Apify account under its retention settings. Remove customer data according to your own policy; no external analytics or customer-data forwarding is implemented. Built-in price and compute assumptions are hypotheses until actual customer usage is observed.

Operator pause/recovery: build a version with PORTFOLIO_STOP=1 and switch latest to that build; a stopped run returns diagnostics and no custom outcome charge. For recovery, rebuild with PORTFOLIO_STOP=0, verify a private validation tag, then switch latest. Editing an environment variable alone does not modify an existing build.

Sources and licenses

• https://open.fda.gov/apis/food/enforcement/
• https://open.fda.gov/license/

Dependencies: cheerio (MIT), csv-parse (MIT), fast-xml-parser (MIT), robots-parser (MIT).

Actual owner cloud example

Validated 2026-10-26T02:24:06.375Z. Run vZnFZBDiqHyKrODIf completed with 1 processing units and 101 records. This is an owner test, not a customer sale.

Input:

{
  "monitorKey": "owner-validation-v1"
}

Selected actual output records (full JSON/CSV available in your run storage):

[
  {
    "recordType": "recall_query_health",
    "status": "completed",
    "sourceUrl": "https://api.fda.gov/food/enforcement.json?search=report_date%3A%5B20260728%20TO%2020261026%5D&sort=report_date:desc&limit=100",
    "observedAt": "2026-10-26T02:20:21.715Z",
    "scanned": 100,
    "totalAvailable": 240,
    "complete": false,
    "capped": true,
    "matched": 100,
    "deltaRecords": 100,
    "billable": true
  },
  {
    "recordType": "recall_evidence",
    "sourceRecordId": "H-0015-2027",
    "recallNumber": "H-0015-2027",
    "pendingRecallNumber": false,
    "eventId": "99933",
    "recallingFirm": "QUESO PALMITA GP LLC",
    "productDescription": "QUESO PALMITA. Generic Name: Pastel de Queso (Cheese Pie). Intended use: Human consumption. Condition: Frozen. Shelf life or expected life: 90 days. Type of packaging: Vacuum, resealable package. Net weight: 1.2 lbs. UPC: 094922980045.",
    "reason": "Undeclared milk, cheese declared",
    "classification": "Class II",
    "recallStatus": "Ongoing",
    "distribution": "Texas",
    "state": "TX",
    "country": "United States",
    "reportDate": "20261014",
    "initiationDate": "20260915",
    "terminationDate": null,
    "sourceUrl": "https://api.fda.gov/food/enforcement.json?search=recall_number%3A%22H-0015-2027%22",
    "matches": [],
    "matchStatus": "recent_recall",
    "changeType": "baseline",
    "observedAt": "2026-10-26T02:20:21.715Z",
    "attribution": "Data provided by the U.S. Food and Drug Administration",
    "license": "CC0-1.0",
    "humanReviewRequired": true
  }
]