123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264 |
- # SPDX-License-Identifier: AGPL-3.0-or-later
- """This module implements functions needed for the autocompleter.
- """
- # pylint: disable=use-dict-literal
- import json
- import html
- from urllib.parse import urlencode, quote_plus
- import lxml
- from httpx import HTTPError
- from searx import settings
- from searx.engines import (
- engines,
- google,
- )
- from searx.network import get as http_get, post as http_post
- from searx.exceptions import SearxEngineResponseException
- def update_kwargs(**kwargs):
- if 'timeout' not in kwargs:
- kwargs['timeout'] = settings['outgoing']['request_timeout']
- kwargs['raise_for_httperror'] = True
- def get(*args, **kwargs):
- update_kwargs(**kwargs)
- return http_get(*args, **kwargs)
- def post(*args, **kwargs):
- update_kwargs(**kwargs)
- return http_post(*args, **kwargs)
- def brave(query, _lang):
- # brave search autocompleter
- url = 'https://search.brave.com/api/suggest?'
- url += urlencode({'q': query})
- country = 'all'
- # if lang in _brave:
- # country = lang
- kwargs = {'cookies': {'country': country}}
- resp = get(url, **kwargs)
- results = []
- if resp.ok:
- data = resp.json()
- for item in data[1]:
- results.append(item)
- return results
- def dbpedia(query, _lang):
- # dbpedia autocompleter, no HTTPS
- autocomplete_url = 'https://lookup.dbpedia.org/api/search.asmx/KeywordSearch?'
- response = get(autocomplete_url + urlencode(dict(QueryString=query)))
- results = []
- if response.ok:
- dom = lxml.etree.fromstring(response.content)
- results = dom.xpath('//Result/Label//text()')
- return results
- def duckduckgo(query, sxng_locale):
- """Autocomplete from DuckDuckGo. Supports DuckDuckGo's languages"""
- traits = engines['duckduckgo'].traits
- args = {
- 'q': query,
- 'kl': traits.get_region(sxng_locale, traits.all_locale),
- }
- url = 'https://duckduckgo.com/ac/?type=list&' + urlencode(args)
- resp = get(url)
- ret_val = []
- if resp.ok:
- j = resp.json()
- if len(j) > 1:
- ret_val = j[1]
- return ret_val
- def google_complete(query, sxng_locale):
- """Autocomplete from Google. Supports Google's languages and subdomains
- (:py:obj:`searx.engines.google.get_google_info`) by using the async REST
- API::
- https://{subdomain}/complete/search?{args}
- """
- google_info = google.get_google_info({'searxng_locale': sxng_locale}, engines['google'].traits)
- url = 'https://{subdomain}/complete/search?{args}'
- args = urlencode(
- {
- 'q': query,
- 'client': 'gws-wiz',
- 'hl': google_info['params']['hl'],
- }
- )
- results = []
- resp = get(url.format(subdomain=google_info['subdomain'], args=args))
- if resp.ok:
- json_txt = resp.text[resp.text.find('[') : resp.text.find(']', -3) + 1]
- data = json.loads(json_txt)
- for item in data[0]:
- results.append(lxml.html.fromstring(item[0]).text_content())
- return results
- def mwmbl(query, _lang):
- """Autocomplete from Mwmbl_."""
- # mwmbl autocompleter
- url = 'https://api.mwmbl.org/search/complete?{query}'
- results = get(url.format(query=urlencode({'q': query}))).json()[1]
- # results starting with `go:` are direct urls and not useful for auto completion
- return [result for result in results if not result.startswith("go: ") and not result.startswith("search: ")]
- def seznam(query, _lang):
- # seznam search autocompleter
- url = 'https://suggest.seznam.cz/fulltext/cs?{query}'
- resp = get(
- url.format(
- query=urlencode(
- {'phrase': query, 'cursorPosition': len(query), 'format': 'json-2', 'highlight': '1', 'count': '6'}
- )
- )
- )
- if not resp.ok:
- return []
- data = resp.json()
- return [
- ''.join([part.get('text', '') for part in item.get('text', [])])
- for item in data.get('result', [])
- if item.get('itemType', None) == 'ItemType.TEXT'
- ]
- def stract(query, _lang):
- # stract autocompleter (beta)
- url = f"https://stract.com/beta/api/autosuggest?q={quote_plus(query)}"
- resp = post(url)
- if not resp.ok:
- return []
- return [html.unescape(suggestion['raw']) for suggestion in resp.json()]
- def startpage(query, sxng_locale):
- """Autocomplete from Startpage. Supports Startpage's languages"""
- lui = engines['startpage'].traits.get_language(sxng_locale, 'english')
- url = 'https://startpage.com/suggestions?{query}'
- resp = get(url.format(query=urlencode({'q': query, 'segment': 'startpage.udog', 'lui': lui})))
- data = resp.json()
- return [e['text'] for e in data.get('suggestions', []) if 'text' in e]
- def swisscows(query, _lang):
- # swisscows autocompleter
- url = 'https://swisscows.ch/api/suggest?{query}&itemsCount=5'
- resp = json.loads(get(url.format(query=urlencode({'query': query}))).text)
- return resp
- def qwant(query, sxng_locale):
- """Autocomplete from Qwant. Supports Qwant's regions."""
- results = []
- locale = engines['qwant'].traits.get_region(sxng_locale, 'en_US')
- url = 'https://api.qwant.com/v3/suggest?{query}'
- resp = get(url.format(query=urlencode({'q': query, 'locale': locale, 'version': '2'})))
- if resp.ok:
- data = resp.json()
- if data['status'] == 'success':
- for item in data['data']['items']:
- results.append(item['value'])
- return results
- def wikipedia(query, sxng_locale):
- """Autocomplete from Wikipedia. Supports Wikipedia's languages (aka netloc)."""
- results = []
- eng_traits = engines['wikipedia'].traits
- wiki_lang = eng_traits.get_language(sxng_locale, 'en')
- wiki_netloc = eng_traits.custom['wiki_netloc'].get(wiki_lang, 'en.wikipedia.org')
- url = 'https://{wiki_netloc}/w/api.php?{args}'
- args = urlencode(
- {
- 'action': 'opensearch',
- 'format': 'json',
- 'formatversion': '2',
- 'search': query,
- 'namespace': '0',
- 'limit': '10',
- }
- )
- resp = get(url.format(args=args, wiki_netloc=wiki_netloc))
- if resp.ok:
- data = resp.json()
- if len(data) > 1:
- results = data[1]
- return results
- def yandex(query, _lang):
- # yandex autocompleter
- url = "https://suggest.yandex.com/suggest-ff.cgi?{0}"
- resp = json.loads(get(url.format(urlencode(dict(part=query)))).text)
- if len(resp) > 1:
- return resp[1]
- return []
- backends = {
- 'dbpedia': dbpedia,
- 'duckduckgo': duckduckgo,
- 'google': google_complete,
- 'mwmbl': mwmbl,
- 'seznam': seznam,
- 'startpage': startpage,
- 'stract': stract,
- 'swisscows': swisscows,
- 'qwant': qwant,
- 'wikipedia': wikipedia,
- 'brave': brave,
- 'yandex': yandex,
- }
- def search_autocomplete(backend_name, query, sxng_locale):
- backend = backends.get(backend_name)
- if backend is None:
- return []
- try:
- return backend(query, sxng_locale)
- except (HTTPError, SearxEngineResponseException):
- return []
|