[fix] braveapi: strip HTML tags and filter favicon thumbnails (#6381)

Heexy now passes the cacheft token via cookies and no
longer via HTTP headers.

Hence, the engine is broken without that change.

searx/autocomplete.py

@@ -21,6 +21,8 @@ from searx.engines import (
 from searx.network import get as http_get, post as http_post
 from searx.exceptions import SearxEngineResponseException
 from searx.utils import extr, gen_useragent
+from searx.data import ENGINE_TRAITS
+from searx.enginelib.traits import EngineTraits
 
 if t.TYPE_CHECKING:
     from searx.extended_types import SXNG_Response
@@ -133,7 +135,9 @@ def google_complete(query: str, sxng_locale: str) -> list[str]:
 
     """
 
-    google_info: dict[str, t.Any] = google.get_google_info({'searxng_locale': sxng_locale}, engines['google'].traits)
+    data = ENGINE_TRAITS.get("google") or {}
+    traits = EngineTraits(**data)
+    google_info: dict[str, t.Any] = google.get_google_info({'searxng_locale': sxng_locale}, traits)
     url = 'https://{subdomain}/complete/search?{args}'
     args = urlencode(
         {

searx/engines/braveapi.py

@@ -31,6 +31,7 @@ from dateutil import parser
 
 from searx.exceptions import SearxEngineAPIException
 from searx.result_types import EngineResults
+from searx.utils import html_to_text
 
 if t.TYPE_CHECKING:
     from searx.extended_types import SXNG_Response
@@ -75,6 +76,7 @@ def request(query: str, params: "OnlineParams") -> None:
         "q": query,
         "count": results_per_page,
         "offset": (params["pageno"] - 1) * results_per_page,
+        "text_decorations": False,
     }
 
     # Apply time filter if specified
@@ -112,14 +114,19 @@ def response(resp: "SXNG_Response") -> EngineResults:
     res = EngineResults()
     data = resp.json()
 
-    for result in data.get("web", {}).get("results", []):
+    for result in (data.get("web") or {}).get("results", []):
+        thumbnail_obj = result.get("thumbnail")
+        thumbnail = ""
+        if thumbnail_obj and not thumbnail_obj.get("logo", False):
+            thumbnail = thumbnail_obj.get("src") or ""
+
         res.add(
             res.types.MainResult(
                 url=result["url"],
-                title=result["title"],
+                title=html_to_text(result["title"]),
-                content=result.get("description", ""),
+                content=html_to_text(result.get("description", "")),
                 publishedDate=_extract_published_date(result.get("age")),
-                thumbnail=result.get("thumbnail", {}).get("src"),
+                thumbnail=thumbnail,
             ),
         )
 

searx/engines/heexy.py

@@ -13,13 +13,12 @@ It seems to use Bing internally, as the image thumbnails are loaded from Bing.
 from urllib.parse import urlencode
 
 import typing as t
-from lxml import html
 
 from searx.enginelib import EngineCache
 from searx.network import get
 from searx.exceptions import SearxEngineAPIException, SearxEngineAccessDeniedException
 from searx.result_types import EngineResults
-from searx.utils import eval_xpath, extract_text, gen_useragent
+from searx.utils import gen_useragent
 
 if t.TYPE_CHECKING:
     from searx.extended_types import SXNG_Response
@@ -73,8 +72,7 @@ def _get_api_token(query: str) -> str:
     if not resp.ok:
         raise SearxEngineAPIException("failed to obtain request token: invalid response code")
 
-    doc = html.fromstring(resp.text)
+    token = resp.cookies["cacheft"]
-    token = extract_text(eval_xpath(doc, "//html/@data-cacheft"))
     if not token:
         raise SearxEngineAPIException("failed to obtain request token: no token found")
 
@@ -92,8 +90,9 @@ def request(query: str, params: "OnlineParams") -> None:
         args["lang"] = params["searxng_locale"].split("-")[0]
 
     params["url"] = f"{api_url}/search/{heexy_categ}?{urlencode(args)}"
-    params["headers"]["X-Data-Cacheft"] = _get_api_token(query)
+
     params["headers"]["Origin"] = api_url
+    params["cookies"]["cacheft"] = _get_api_token(query)
 
 
 def response(resp: "SXNG_Response"):