[mod] improve 'Autodetect search language' plugin

- Add documentation to the plugin
- Harmonize FastText language model with SearXNG's language model

Reosurces::

    import fasttext                                    # --> +10 MB
    fasttext.load_model(str(data_dir / 'lid.176.ftz')) # --> +4MB

Suggested-by: @dalf

- To speed up and simplify the deployment use fasttext-wheel instead of fasttext
- Building numpy on the Alpine Linux of docker-images takes ages --> install
  py3-numpy from Alpines package manager (apk)
- Alpine Linux on docker-images (musl libc) do not support fasttext-wheel (gnu
  libc) --> patch Dockerfile and build from fastetxt:

     sed -i s/fasttext-wheel/fasttext/ requirements.txt

Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>

Dockerfile

@@ -36,6 +36,7 @@ RUN apk add --no-cache -t build-dependencies \
     su-exec \
     python3 \
     py3-pip \
+    py3-numpy \
     libxml2 \
     libxslt \
     openssl \
@@ -43,6 +44,8 @@ RUN apk add --no-cache -t build-dependencies \
     uwsgi \
     uwsgi-python3 \
     brotli \
+ && pip3 install --no-cache setuptools wheel \
+ && sed -i s/fasttext-wheel/fasttext/ requirements.txt \
  && pip3 install --no-cache -r requirements.txt \
  && apk del build-dependencies \
  && rm -rf /root/.cache

docs/src/searx.plugins.autodetect_search_language.rst

@@ -0,0 +1,8 @@
+.. _autodetect search language:
+
+======================
+Search language plugin
+======================
+
+.. automodule:: searx.plugins.autodetect_search_language
+  :members:

requirements.txt

@@ -16,4 +16,4 @@ setproctitle==1.3.2
 redis==4.4.0
 markdown-it-py==2.1.0
 typing_extensions==4.4.0
-fasttext==0.9.2
+fasttext-wheel==0.9.2

searx/plugins/autodetect_search_language.py

@@ -1,19 +1,115 @@
-import fasttext
-import os
+# SPDX-License-Identifier: AGPL-3.0-or-later
+# lint: pylint
+"""Plugin to detect the search language from the search query.
+
+The language detection is done by using the fastText_ library (`python
+fasttext`_). fastText_ distributes the `language identification model`_, for
+reference:
+
+- `FastText.zip: Compressing text classification models`_
+- `Bag of Tricks for Efficient Text Classification`_
+
+The `language identification model`_ support the language codes (ISO-639-3)::
+
+   af als am an ar arz as ast av az azb ba bar bcl be bg bh bn bo bpy br bs bxr
+   ca cbk ce ceb ckb co cs cv cy da de diq dsb dty dv el eml en eo es et eu fa
+   fi fr frr fy ga gd gl gn gom gu gv he hi hif hr hsb ht hu hy ia id ie ilo io
+   is it ja jbo jv ka kk km kn ko krc ku kv kw ky la lb lez li lmo lo lrc lt lv
+   mai mg mhr min mk ml mn mr mrj ms mt mwl my myv mzn nah nap nds ne new nl nn
+   no oc or os pa pam pfl pl pms pnb ps pt qu rm ro ru rue sa sah sc scn sco sd
+   sh si sk sl so sq sr su sv sw ta te tg th tk tl tr tt tyv ug uk ur uz vec vep
+   vi vls vo wa war wuu xal xmf yi yo yue zh
+
+The `language identification model`_ is harmonized with the SearXNG's language
+(locale) model.  General conditions of SearXNG's locale model are:
+
+a. SearXNG's locale of a query is passed to the
+   :py:obj:`searx.locales.get_engine_locale` to get a language and/or region
+   code that is used by an engine.
+
+b. SearXNG and most of the engines do not support all the languages from
+   language model and there might be also a discrepancy in the ISO-639-3 and
+   ISO-639-2 handling (:py:obj:`searx.locales.get_engine_locale`).  Further
+   more, in SearXNG the locales like ``zh-TH`` (``zh-CN``) are mapped to
+   ``zh_Hant`` (``zh_Hans``).
+
+Conclusion: This plugin does only auto-detect the languages a user can select in
+the language menu (:py:obj:`supported_langs`).
+
+SearXNG's locale of a query comes from (*highest wins*):
+
+1. The ``Accept-Language`` header from user's HTTP client.
+2. The user select a locale in the preferences.
+3. The user select a locale from the menu in the query form (e.g. ``:zh-TW``)
+4. This plugin is activated in the preferences and the locale (only the language
+   code / none region code) comes from the fastText's language detection.
+
+Conclusion: There is a conflict between the language selected by the user and
+the language from language detection of this plugin.  For example, the user
+explicitly selects the German locale via the search syntax to search for a term
+that is identified as an English term (try ``:de-DE thermomix``, for example).
+
+.. hint::
+
+   To SearXNG maintainers; please take into account: under some circumstances
+   the auto-detection of the language of this plugin could be detrimental to
+   users expectations.  Its not recommended to activate this plugin by
+   default. It should always be the user's decision whether to activate this
+   plugin or not.
+
+.. _fastText: https://fasttext.cc/
+.. _python fasttext: https://pypi.org/project/fasttext/
+.. _language identification model: https://fasttext.cc/docs/en/language-identification.html
+.. _Bag of Tricks for Efficient Text Classification: https://arxiv.org/abs/1607.01759
+.. _`FastText.zip: Compressing text classification models`: https://arxiv.org/abs/1612.03651
+
+"""
+
 from flask_babel import gettext
+import fasttext
+import babel
+
+from searx.data import data_dir
+from searx.languages import language_codes
+
+# Monkey patch: prevent fasttext from showing a (useless) warning when loading a
+# model.
+fasttext.FastText.eprint = lambda x: None
 
 name = gettext('Autodetect search language')
 description = gettext('Automatically detect the query search language and switch to it.')
 preference_section = 'general'
 default_on = False
 
+lang_model: fasttext.FastText._FastText = None
+"""fasttext model to predict laguage of a search term"""
 
-fasttext.FastText.eprint = lambda x: None
-model = fasttext.load_model(os.path.dirname(os.path.realpath(__file__)) + '/../data/lid.176.ftz')
+supported_langs = set()
+"""Languages supported by most searxng engines (:py:obj:`searx.languages.language_codes`)."""
+
+
+def get_model():
+    # lazy load, in order to to save memory
+    global lang_model  # pylint: disable=global-statement
+    if lang_model is None:
+        lang_model = fasttext.load_model(str(data_dir / 'lid.176.ftz'))
+    return lang_model
+
+
+def pre_search(request, search):  # pylint: disable=unused-argument
+    prediction = get_model().predict(search.search_query.query, k=1, threshold=0.3)
+    if prediction:
+        lang = prediction[0][0].split('__label__')[1]
+        if lang in supported_langs:
+            search.search_query.lang = lang
+            try:
+                search.search_query.locale = babel.Locale.parse(lang)
+            except babel.core.UnknownLocaleError:
+                pass
+    return True
 
 
-def pre_search(request, search):
-    lang = model.predict(search.search_query.query, k=1)
-    if lang[1][0] >= 0.3:
-        search.search_query.lang = lang[0][0].split('__label__')[1]
+def init(app, settings):  # pylint: disable=unused-argument
+    for searxng_locale in language_codes:
+        supported_langs.add(searxng_locale[0].split('-')[0])
     return True

searx/settings.yml

@@ -196,6 +196,9 @@ outgoing:
 #   - 'Open Access DOI rewrite'
 #   - 'Vim-like hotkeys'
 #   - 'Tor check plugin'
+#   # Read the docs before activate: auto-detection of the language could be
+#   # detrimental to users expectations / users can activate the plugin in the
+#   # preferences if they want.
 #   - 'Autodetect search language'
 
 # Configuration of the "Hostname replace" plugin: