searxng/searx/engines/mullvad_leta.py

208 lines
6.5 KiB
Python
Raw Permalink Normal View History

# SPDX-License-Identifier: AGPL-3.0-or-later
"""This is the implementation of the Mullvad-Leta meta-search engine.
This engine **REQUIRES** that searxng operate within a Mullvad VPN
If using docker, consider using gluetun for easily connecting to the Mullvad
- https://github.com/qdm12/gluetun
Otherwise, follow instructions provided by Mullvad for enabling the VPN on Linux
- https://mullvad.net/en/help/install-mullvad-app-linux
.. hint::
The :py:obj:`EngineTraits` is empty by default. Maintainers have to run
``make data.traits`` (in the Mullvad VPN / :py:obj:`fetch_traits`) and rebase
the modified JSON file ``searx/data/engine_traits.json`` on every single
update of SearXNG!
"""
from __future__ import annotations
from typing import TYPE_CHECKING
from httpx import Response
from lxml import html
from searx.enginelib.traits import EngineTraits
from searx.locales import region_tag, get_official_locales
from searx.utils import eval_xpath, extract_text, eval_xpath_list
from searx.exceptions import SearxEngineResponseException
if TYPE_CHECKING:
import logging
logger = logging.getLogger()
traits: EngineTraits
use_cache: bool = True # non-cache use only has 100 searches per day!
leta_engine: str = 'google'
search_url = "https://leta.mullvad.net"
# about
about = {
"website": search_url,
"wikidata_id": 'Q47008412', # the Mullvad id - not leta, but related
"official_api_documentation": 'https://leta.mullvad.net/faq',
"use_official_api": False,
"require_api_key": False,
"results": 'HTML',
}
# engine dependent config
categories = ['general', 'web']
paging = True
max_page = 50
time_range_support = True
time_range_dict = {
"day": "d1",
"week": "w1",
"month": "m1",
"year": "y1",
}
available_leta_engines = [
'google', # first will be default if provided engine is invalid
'brave',
]
def is_vpn_connected(dom: html.HtmlElement) -> bool:
"""Returns true if the VPN is connected, False otherwise"""
connected_text = extract_text(eval_xpath(dom, '//main/div/p[1]'))
return connected_text != 'You are not connected to Mullvad VPN.'
def assign_headers(headers: dict) -> dict:
"""Assigns the headers to make a request to Mullvad Leta"""
headers['Accept'] = "text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,*/*;q=0.8"
headers['Content-Type'] = "application/x-www-form-urlencoded"
headers['Host'] = "leta.mullvad.net"
headers['Origin'] = "https://leta.mullvad.net"
return headers
def request(query: str, params: dict):
country = traits.get_region(params.get('searxng_locale', 'all'), traits.all_locale) # type: ignore
result_engine = leta_engine
if leta_engine not in available_leta_engines:
result_engine = available_leta_engines[0]
logger.warning(
'Configured engine "%s" not one of the available engines %s, defaulting to "%s"',
leta_engine,
available_leta_engines,
result_engine,
)
params['url'] = search_url
params['method'] = 'POST'
params['data'] = {
"q": query,
"gl": country if country is str else '',
'engine': result_engine,
}
# pylint: disable=undefined-variable
if use_cache:
params['data']['oc'] = "on"
# pylint: enable=undefined-variable
if params['time_range'] in time_range_dict:
params['dateRestrict'] = time_range_dict[params['time_range']]
else:
params['dateRestrict'] = ''
if params['pageno'] > 1:
# Page 1 is n/a, Page 2 is 11, page 3 is 21, ...
params['data']['start'] = ''.join([str(params['pageno'] - 1), "1"])
if params['headers'] is None:
params['headers'] = {}
assign_headers(params['headers'])
return params
def extract_result(dom_result: list[html.HtmlElement]):
# Infoboxes sometimes appear in the beginning and will have a length of 0
if len(dom_result) == 3:
[a_elem, h3_elem, p_elem] = dom_result
elif len(dom_result) == 4:
[_, a_elem, h3_elem, p_elem] = dom_result
else:
return None
return {
'url': extract_text(a_elem.text),
'title': extract_text(h3_elem),
'content': extract_text(p_elem),
}
def extract_results(search_results: html.HtmlElement):
for search_result in search_results:
dom_result = eval_xpath_list(search_result, 'div/div/*')
result = extract_result(dom_result)
if result is not None:
yield result
def response(resp: Response):
"""Checks if connected to Mullvad VPN, then extracts the search results from
the DOM resp: requests response object"""
dom = html.fromstring(resp.text)
if not is_vpn_connected(dom):
raise SearxEngineResponseException('Not connected to Mullvad VPN')
search_results = eval_xpath(dom.body, '//main/div[2]/div')
return list(extract_results(search_results))
def fetch_traits(engine_traits: EngineTraits):
"""Fetch languages and regions from Mullvad-Leta
.. warning::
Fetching the engine traits also requires a Mullvad VPN connection. If
not connected, then an error message will print and no traits will be
updated.
"""
# pylint: disable=import-outside-toplevel
# see https://github.com/searxng/searxng/issues/762
from searx.network import post as http_post
# pylint: enable=import-outside-toplevel
resp = http_post(search_url, headers=assign_headers({}))
if not isinstance(resp, Response):
print("ERROR: failed to get response from mullvad-leta. Are you connected to the VPN?")
return
if not resp.ok:
print("ERROR: response from mullvad-leta is not OK. Are you connected to the VPN?")
return
dom = html.fromstring(resp.text)
if not is_vpn_connected(dom):
print('ERROR: Not connected to Mullvad VPN')
return
# supported region codes
options = eval_xpath_list(dom.body, '//main/div/form/div[2]/div/select[1]/option')
if options is None or len(options) <= 0:
print('ERROR: could not find any results. Are you connected to the VPN?')
for x in options:
eng_country = x.get("value")
sxng_locales = get_official_locales(eng_country, engine_traits.languages.keys(), regional=True)
if not sxng_locales:
print(
"ERROR: can't map from Mullvad-Leta country %s (%s) to a babel region."
% (x.get('data-name'), eng_country)
)
continue
for sxng_locale in sxng_locales:
engine_traits.regions[region_tag(sxng_locale)] = eng_country