faster-flights 3.5.5

The fast, robust, strongly-typed Google Flights scraper (API) implemented in Python.

✈️ faster-flights

Fast, strongly-typed Google Flights scraping for Python.

Documentation • Issues • PyPI

pip install faster-flights

Quick start

from fast_flights import FlightQuery, Passengers, create_query, get_flights

query = create_query(
    flights=[
        FlightQuery(
            date="2026-03-31",
            from_airport="TPE",
            to_airport="NRT",
        ),
    ],
    trip="one-way",
    seat="economy",  # economy / premium-economy / business / first
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

results = get_flights(query)

for flight in results[:3]:
    print(f"{flight.airlines} - ${flight.price}")

Current public API

The supported top-level API is the set exported from fast_flights:

• FlightQuery
• Passengers
• create_query() and create_filter() (create_filter is a compatibility alias)
• SearchSession
• build_booking_tfs() / build_booking_url()
• get_flights()
• select_flight()
• get_return_flights()
• get_flights_multicity()
• get_flights_multicity_chained()
• get_selected_flight_page()
• PlaywrightBrowserProvider (experimental)
• fetch_flights_html()

Use IATA airport codes like "TPE", "NRT", or "JFK" in FlightQuery. The repository does not currently export a public airport search helper.

Round-trip searches

Round-trip search is a two-step flow:

1. Query outbound options with get_flights()
2. Select one outbound option and fetch return options with get_return_flights()

from fast_flights import (
    FlightQuery,
    Passengers,
    create_query,
    get_flights,
    get_return_flights,
    select_flight,
)

query = create_query(
    flights=[
        FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="NRT"),
        FlightQuery(date="2026-04-05", from_airport="NRT", to_airport="TPE"),
    ],
    trip="round-trip",
    seat="business",
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

outbound = get_flights(query)
return_query = select_flight(query, outbound[0])
returning = get_return_flights(return_query)

Each outbound result carries a hidden select_token. select_flight() wraps that token into a ReturnQuery.

get_return_flights() first tries Google's selected-flight HTML and verifies that the returned itinerary direction matches the requested return leg. If Google still serves the wrong payload there, the library rebuilds the browser-style selected tfs state from the chosen outbound segments and fetches the bundled return page again.

Only if both HTML paths fail does the library fall back to an independent one-way query for the requested return leg. In that last-resort mode, the returned Flights.price values are per-leg one-way prices rather than Google's selected round-trip total.

Session workflow

Use SearchSession when you want one API for chained search state, current search links, and the final booking link:

from fast_flights import FlightQuery, Passengers, SearchSession, create_query

query = create_query(
    flights=[
        FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="NRT"),
        FlightQuery(date="2026-04-05", from_airport="NRT", to_airport="TPE"),
    ],
    trip="round-trip",
    seat="economy",
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

session = SearchSession(query, mode="rpc-first")
outbound = session.results()
session = session.select(outbound[0])
returning = session.results()
session = session.select(returning[0])

print(session.current_search_tfs)   # selected-page / search-state tfs
print(session.final_booking_tfs)    # final booking itinerary tfs
print(session.booking_url())        # https://www.google.com/travel/flights/booking?...

build_booking_tfs() and build_booking_url() expose the same final booking-state builder directly if you already have selected_legs.

Experimental browser parity

If later-leg results do not match what Google shows in a real browser, you can let SearchSession fall back to a browser-captured shopping response:

from fast_flights import (
    FlightQuery,
    Passengers,
    PlaywrightBrowserProvider,
    SearchSession,
    create_query,
)

query = create_query(
    flights=[
        FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="LHR"),
        FlightQuery(date="2026-04-14", from_airport="LHR", to_airport="TPE"),
    ],
    trip="round-trip",
    seat="economy",
    passengers=Passengers(adults=1),
    language="en-GB",
    currency="GBP",
)

session = SearchSession(
    query,
    mode="rpc-first",
    browser_fallback=True,
    browser_provider=PlaywrightBrowserProvider(),
)

This path is experimental and only applies to later legs. The normal order stays:

• rpc-first: shopping RPC -> selected/bundled HTML -> browser capture -> directional fallback
• ssr-first: selected/bundled HTML -> shopping RPC -> browser capture -> directional fallback

The browser provider captures the real GetShoppingResults response and feeds that response into the existing parser. It does not replay the request body in v1.

Multi-city searches

There are three supported multi-city workflows.

1. Total trip price + directional per-leg options

Use get_flights_multicity_chained() when you want Google's bundled multi-city pricing plus correctly directed parsed options for each leg.

from fast_flights import FlightQuery, get_flights_multicity_chained

legs = [
    FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="NRT"),
    FlightQuery(date="2026-04-05", from_airport="NRT", to_airport="HKG"),
    FlightQuery(date="2026-04-10", from_airport="HKG", to_airport="TPE"),
]

result = get_flights_multicity_chained(
    legs,
    seat="economy",
    language="en-US",
    currency="USD",
)

print(result[0].total_price)
print(result[1].flights[0].flights[-1].to_airport.code)  # per-leg directional results

total_price is Google's bundled multi-city price from the RPC response. The per-leg flights lists are populated with independent one-way searches for each leg so the parsed routes always match the requested leg direction. That means later-leg Flights.price values are per-leg one-way prices, not re-priced bundled totals.

2. Per-leg flight details

Use get_flights_multicity() when you want each leg as an independent one-way search.

from fast_flights import FlightQuery, Passengers, get_flights_multicity

legs = get_flights_multicity(
    flights=[
        FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="NRT"),
        FlightQuery(date="2026-04-05", from_airport="NRT", to_airport="HKG"),
        FlightQuery(date="2026-04-10", from_airport="HKG", to_airport="TPE"),
    ],
    seat="economy",
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

This returns per-leg one-way pricing, not Google's bundled total trip price.

3. Hybrid workflow

Use get_flights(create_query(..., trip="multi-city")) for Google's bundled first-leg options, then get_flights_multicity() for detailed options on legs 2+.

from fast_flights import (
    FlightQuery,
    Passengers,
    create_query,
    get_flights,
    get_flights_multicity,
)

legs = [
    FlightQuery(date="2026-03-31", from_airport="TPE", to_airport="NRT"),
    FlightQuery(date="2026-04-05", from_airport="NRT", to_airport="HKG"),
    FlightQuery(date="2026-04-10", from_airport="HKG", to_airport="TPE"),
]

query = create_query(
    flights=legs,
    trip="multi-city",
    seat="economy",
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

leg1 = get_flights(query)
remaining = get_flights_multicity(
    flights=legs[1:],
    seat="economy",
    passengers=Passengers(adults=1),
    language="en-US",
    currency="USD",
)

4. Direct Google Flights selected page

Use get_selected_flight_page() when a user picks an option and you want the direct Google Flights page for that selection.

from fast_flights import get_selected_flight_page

selected = get_selected_flight_page(query, leg1[0])

print(selected.url)
print(selected.f_sid)
print(selected.bl)
print(selected.data_service_requests["ds:1"].rpc_id)

For multi-city searches, this gives you the real Google Flights selected page plus the client-side data-service requests embedded in that page. Google still does not expose later-leg bundled options as parsed Flights objects through this library's public API.

The round-trip return-flight API is not the supported way to fetch legs 2+ for multi-city itineraries. Google's HTML response for that path does not provide the required data.

Integrations and proxies

Fetching customization is split between integration= for HTML fetches and browser_provider= for experimental browser-assisted parity in SearchSession.

Bright Data

from fast_flights import get_flights
from fast_flights.integrations import BrightData

integration = BrightData(api_key="...")
results = get_flights(query, integration=integration)

Custom integrations

Subclass fast_flights.integrations.base.Integration and implement fetch_html(q) -> str.

from fast_flights.integrations.base import Integration


class MyIntegration(Integration):
    def fetch_html(self, q):
        return "...html..."

Browser providers

Browser providers are separate from integrations. They do not replace fetch_html(). They only provide an optional network-capture fallback for chained SearchSession searches.

from fast_flights import PlaywrightBrowserProvider, SearchSession

session = SearchSession(
    query,
    browser_fallback=True,
    browser_provider=PlaywrightBrowserProvider(),
)

Proxy support

results = get_flights(query, proxy="http://user:pass@proxy:8080")

get_flights_multicity() also forwards integration= and proxy= to each per-leg one-way search. RPC-based multi-city flows (get_flights(...trip="multi-city") and get_flights_multicity_chained()) support proxy= but not integration overrides.

Documentation

• Getting started
• Query building
• Return flights
• Multi-city
• Integrations and proxies

Contributing

Contributing is welcomed. A few notes though:

1. please no ai slop. i am not reading all that.
2. one change at a time. what your title says is what you've changed.
3. no new dependencies unless it's related to the core parsing.
4. really, i cant finish reading all of them, i have other projects and life to do. really sorry