FetchFlight

Elixir port of the Python fast_flights library. Scrapes Google Flights by encoding a query as a protobuf binary, fetching the results page, and returning structured flight data — no API key required.

Two entry points:

• get_flights/2 — fetch specific itineraries for a given date
• get_price_graph/1 — fetch a calendar of cheapest prices across a date range

Installation

def deps do
  [
    {:fetch_flight, "~> 0.2"}
  ]
end

Usage

get_flights/2 — flight search

query = %{
  data: [
    %{
      date: "2026-05-01",
      from_airport: %{code: "SFO"},
      to_airport: %{code: "JFK"},
      max_stops: nil,
      airlines: []
    }
  ],
  seat: :economy,
  trip: :one_way,
  passengers: [:adult]
}

{:ok, {metadata, flights}} = FetchFlight.get_flights(query)

Options

FetchFlight.get_flights(query, language: "en", currency: "USD")

Query fields

Top-level

Per leg (data list)

Multi-city example

query = %{
  data: [
    %{date: "2026-06-01", from_airport: %{code: "SFO"}, to_airport: %{code: "LHR"}, max_stops: nil, airlines: []},
    %{date: "2026-06-10", from_airport: %{code: "LHR"}, to_airport: %{code: "CDG"}, max_stops: 0,   airlines: []}
  ],
  seat: :business,
  trip: :multi_city,
  passengers: [:adult, :adult]
}

get_price_graph/1 — price calendar

Returns the cheapest price for each departure date in a range — useful for finding the best days to fly.

query = %{
  range_start_date: "2026-05-01",
  range_end_date:   "2026-05-31",
  trip_length:      7,
  src_airports:     ["SFO"],
  dst_airports:     ["JFK"]
}

{:ok, offers} = FetchFlight.get_price_graph(query)

# With a different currency
{:ok, offers} = FetchFlight.get_price_graph(query, currency: "EUR")

# With time filters (afternoon departures, evening arrivals)
{:ok, offers} = FetchFlight.get_price_graph(%{query | departure_time: {12, 18}, arrival_time: {14, 23}})

Options

FetchFlight.get_price_graph(query, currency: "EUR")

Results are sorted by start_date ascending:

[
  %FetchFlight.PriceGraphOffer{start_date: "2026-04-17", return_date: "2026-04-24", price: 219.0},
  %FetchFlight.PriceGraphOffer{start_date: "2026-04-18", return_date: "2026-04-25", price: 246.0},
  %FetchFlight.PriceGraphOffer{start_date: "2026-05-01", return_date: "2026-05-08", price: 231.0},
  ...
]

Note: Google's Price graph covers roughly ±5 weeks around the search date, so results may extend slightly outside the requested range.

Price graph query fields

Output

get_flights/2

{:ok, {%FetchFlight.JsMetadata{airlines: [...], alliances: [...]}, flights}}

flights is a list of %FetchFlight.Flights{}, sorted cheapest first:

%FetchFlight.Flights{
  type:     "DL",                      # IATA carrier code
  price:    344,                       # in the requested currency
  airlines: ["Delta"],
  carbon: %FetchFlight.CarbonEmission{
    emission_grams:          284_000,
    typical_on_route_grams:  354_000
  },
  flights: [
    %FetchFlight.SingleFlight{
      from_airport: %FetchFlight.Airport{code: "SFO", name: "San Francisco International Airport"},
      to_airport:   %FetchFlight.Airport{code: "JFK", name: "John F. Kennedy International Airport"},
      departure: %FetchFlight.SimpleDatetime{date: [2026, 5, 1], time: [7, 0]},
      arrival:   %FetchFlight.SimpleDatetime{date: [2026, 5, 1], time: [15, 28]},
      duration_minutes: 328,
      plane_type: "Boeing 767"
    }
  ]
}

SimpleDatetime format

• :date — [year, month, day]
• :time — [hour, minute], or [hour] when minutes = 0

get_price_graph/1

{:ok, [%FetchFlight.PriceGraphOffer{}, ...]}

Each %FetchFlight.PriceGraphOffer{} has:

How it works

get_flights/2

1. Encode — the query is serialized as a protobuf binary (manually encoded, no generated code) and Base64-encoded into the tfs URL parameter.
2. Fetch — GET https://www.google.com/travel/flights?tfs=...&hl=...&curr=... with realistic browser headers via req.
3. Parse — floki locates the <script class="ds:1"> tag; jason decodes the embedded JSON payload; specific array indices are navigated to extract flights, prices, and carbon data.

get_price_graph/1

1. Encode — the query (both outbound and return legs) is serialized as a protobuf binary and Base64-encoded into the tfs URL parameter, same as get_flights/2. Including both legs tells Google the intended trip length.
2. Navigate — a headless Chromium browser (via playwright) loads the Google Flights results page, dismisses any deal popup, scrolls to the "Prices for nearby dates" section, and clicks the "Price graph" tab.
3. Intercept — a response listener captures the GetCalendarGraph XHR that fires when the Price graph tab is clicked. This endpoint returns ~10 weeks of daily prices, one entry per departure date at exactly the requested trip length.
4. Parse — the response (prefixed with )]}'\n) is decoded with jason. The outer envelope is unwrapped, the inner JSON string is decoded, and [departure_date, return_date, [[null, price], token], 1] entries are filtered to the requested date range and sorted by departure date.

Disclaimer

This library scrapes a public web interface. Google may change the page structure without notice, which could break parsing. It is intended for personal and research use — check Google's Terms of Service before deploying at scale.

License

MIT