GTFS Feed Extractor

Point at ANY GTFS-static feed URL and get flat, structured dataset rows — one actor for every transit agency on Earth.

GTFS (the General Transit Feed Specification, originally Google Transit) is the single global standard for public-transit schedule data. Thousands of agencies on every continent — subways, buses, trams, ferries, rail — publish the identical ZIP-of-CSVs: agency.txt, stops.txt, routes.txt, trips.txt, stop_times.txt, calendar.txt, calendar_dates.txt, shapes.txt, fare_*.txt, feed_info.txt and more.

Every other "transit" scraper wraps one agency. This actor is the generic runner: give it any GTFS-static ZIP URL and it turns the tables you ask for into dataset rows — one row per CSV record, the columns preserved exactly, plus a lossless copy of each original record.

What it does

• Extract mode (default): downloads the feed ZIP and parses the tables you choose into flat rows. Each row carries the CSV columns as-is, plus meta columns _table (which file), _feed_url, _row (0-based index in the table), and _raw (the verbatim original record).
• Discovery mode (listTablesOnly): emits one row per table the feed contains — its name, row count and column names — so you can see what a feed holds before pulling the big tables.

Robust to real-world feed variation: it matches table names case-insensitively against whatever the ZIP actually contains, parses with a proper CSV reader (quoted fields, embedded commas), strips UTF-8 BOMs, tolerates non-UTF-8 feeds and non-standard extension columns, and skips missing tables silently instead of crashing.

Input

Field	Type	Description
gtfs_url (required)	string	Direct URL to a GTFS-static ZIP feed, e.g. https://www.bart.gov/dev/schedules/google_transit.zip.
tables	array	Tables to extract, by name (with or without .txt): stops, routes, trips, calendar… Leave empty for a sensible default set (agency, feed_info, calendar, calendar_dates, routes, stops, trips — not stop_times, which is often 100k+ rows). Use ["all"] to extract every table including stop_times/shapes.
maxRecords	integer	Optional global cap on rows emitted across all requested tables. Keeps a sample cheap.
listTablesOnly	boolean	Discovery mode — list the feed's tables (name, row count, columns) instead of extracting.
bearer	string (secret)	Optional bearer token for gated feeds. Never logged.
extraHeaders	object	Optional extra HTTP headers for gated feeds, e.g. {"x-api-key": "…"}. Never logged.

Output

One dataset item per CSV record. For a stops row from BART, for example:

{
  "_table": "stops",
  "_feed_url": "https://www.bart.gov/dev/schedules/google_transit.zip",
  "_row": 0,
  "stop_id": "A10-1",
  "stop_code": "902101",
  "stop_name": "Lake Merritt",
  "stop_lat": "37.797296",
  "stop_lon": "-122.265287",
  "zone_id": "LAKE",
  "location_type": "0",
  "parent_station": "LAKE",
  "_raw": { "stop_id": "A10-1", "stop_code": "902101", "stop_name": "Lake Merritt", "…": "…" }
}

Empty CSV cells become null in the flat columns; _raw preserves the verbatim record (including any values beyond the header).

Finding feeds

Almost every public agency's GTFS feed is cataloged in the MobilityData Mobility Database (https://mobilitydatabase.org), the ~2,000-feed successor to TransitFeeds. Point this actor at any feed's direct download URL.

Pricing

Pay-per-event: $0.0005 per record (one emitted dataset row). Use maxRecords and a focused tables list to control cost — you rarely need every row of stop_times.

Notes & limits

• GTFS-static only. GTFS-Realtime (the protobuf .pb trip-update/vehicle-position feeds) is a different format and is not covered here.
• Feeds up to 512 MB are supported; larger feeds should be sampled with maxRecords / tables.
• A URL that 404s or isn't a ZIP returns a clear error; a feed missing an optional table simply skips it.