[fix] engines: duckduckgo web bypass botdetection

searx/engines/duckduckgo_web.py

@@ -14,11 +14,12 @@ can't build it ourselves and must scrape it from the HTML pages.
 """
 
 import typing as t
+import re
 
-from urllib.parse import quote_plus
+from urllib.parse import quote_plus, urljoin
 from lxml import html
 
-from searx.utils import html_to_text, gen_useragent, extract_text, eval_xpath
+from searx.utils import html_to_text, extract_text, eval_xpath
 from searx.result_types import EngineResults
 from searx.enginelib import EngineCache
 from searx.network import get
@@ -38,7 +39,6 @@ about = {
 # engine dependent config
 categories = ["general"]
 paging = True
-_HTTP_User_Agent: str = gen_useragent()
 
 base_url = "https://duckduckgo.com"
 
@@ -73,6 +73,8 @@ def _fetch_first_page_link(
     resp = get(
         url=f"{base_url}/?q={quote_plus(query)}&t=h_&ia=web",
         headers=headers,
+        impersonate="firefox",
+        default_headers=False,
         timeout=2,
     )
 
@@ -96,6 +98,43 @@ def _cache_key(query: str, pageno: int) -> str:
     return f"nextpage_url|{query}|{pageno}"
 
 
+def _solve_jsa(resp: "SXNG_Response") -> "SXNG_Response":
+    """Duckduckgo sometimes issues a challenge instead of json."""
+
+    # length that a real browser would report for where the broken snippet is
+    html_len = {
+        "<p><div></p><p></div": 32,
+        "<li><div></li><li></div": 29,
+        "<div><div></div><div></div": 33,
+        "<br><div></br><br></div": 23,
+    }
+
+    js = resp.text or ""
+    jsa_match = re.search(r"let jsa = (\d+);.*?DDG\.deep\.initialize\('([^']+)'", js, re.S)
+    if not jsa_match:
+        return resp
+
+    js_functions = dict(re.findall(r"let (\w+) = function\(num\) \{([^}]*)\};", js))
+    jsa = int(jsa_match.group(1))
+    try:
+        for name in re.findall(r"jsa = (\w+)\(jsa\);", js):
+            body = js_functions[name]
+            mul = re.search(r"num \* (\d+)", body)
+            jsa = jsa * int(mul.group(1)) if mul else jsa + html_len[re.search(r"`([^`]+)`", body).group(1)]
+    except (KeyError, AttributeError):
+        return resp
+
+    params = resp.search_params
+    follow = get(
+        urljoin("https://links.duckduckgo.com", jsa_match.group(2) + str(jsa)),
+        headers=params["headers"],
+        impersonate="firefox",
+        default_headers=False,
+    )
+    follow.search_params = params
+    return follow
+
+
 def request(query: str, params: "OnlineParams") -> None:
 
     if len(query) >= 500:
@@ -103,25 +142,15 @@ def request(query: str, params: "OnlineParams") -> None:
         params["url"] = None
         return
 
-    headers = params["headers"]
-
-    # The vqd value is generated from the query and the UA header.  To be able
-    # to reuse the vqd value, the UA header must be static.
-    headers["User-Agent"] = _HTTP_User_Agent
-    headers["Accept"] = "*/*"
-    headers["Referer"] = f"{base_url}/"
-    headers["Host"] = "duckduckgo.com"
-
-    # Sec-Fetch headers are required to not get blocked when sending a Firefox user agent
-    headers["Sec-Fetch-Dest"] = "script"
-    headers["Sec-Fetch-Mode"] = "no-cors"
-    headers["Sec-Fetch-Site"] = "same-site"
+    # firefox TLS only
+    params["impersonate"] = "firefox"
+    params["default_headers"] = False
 
     api_url = ""
     if params["pageno"] > 1:
         api_url = CACHE.get(_cache_key(query, params["pageno"]))
     else:
-        api_url = _fetch_first_page_link(query, headers)
+        api_url = _fetch_first_page_link(query, params["headers"])
 
     if not api_url:
         params["url"] = None
@@ -129,14 +158,27 @@ def request(query: str, params: "OnlineParams") -> None:
 
     params["url"] = api_url.replace("/d.js?", "/d.js?o=json&")
 
+    # loads as a script
+    headers = params["headers"]
+    headers["Accept"] = "*/*"
+    headers["Sec-Fetch-Dest"] = "script"
+    headers["Sec-Fetch-Mode"] = "no-cors"
+    headers["Sec-Fetch-Site"] = "same-site"
+    headers["Referer"] = f"{base_url}/"
+
     # TODO: support safesearch, timerange and engine traits  # pylint:disable=fixme
 
 
 def response(resp: "SXNG_Response"):
     res = EngineResults()
-    res_json = resp.json()
 
-    for result in res_json["results"]:
+    # check if ddg returns a challenge
+    # e.g. 'site:github.com searxng'
+    if "let jsa =" in (resp.text or ""):
+        resp = _solve_jsa(resp)
+
+    results = resp.json()["results"]
+    for result in results:
         if "u" not in result:
             continue
 
@@ -144,8 +186,8 @@ def response(resp: "SXNG_Response"):
             res.types.MainResult(url=result["u"], title=html_to_text(result["t"]), content=html_to_text(result["a"]))
         )
 
-    # link to next page
-    next_page_path = res_json["results"][-1].get("n")
+    if results:
+        next_page_path = results[-1].get("n")
         if next_page_path:
             CACHE.set(
                 _cache_key(resp.search_params["query"], resp.search_params["pageno"] + 1),

searx/search/processors/online.py

@@ -88,6 +88,9 @@ class HTTPParams(t.TypedDict):
     impersonate: t.NotRequired[str]
     """curl_cffi impersonate target. Default: ``chrome``."""
 
+    default_headers: t.NotRequired[bool]
+    """Ask curl_cffi to skip setting the impersonated browser's default headers."""
+
     curl_options: t.NotRequired[dict[int, t.Any]]
     """Any extra libcurl options for the request."""
 
@@ -169,8 +172,8 @@ class OnlineProcessor(EngineProcessor):
             "cookies": params["cookies"],
             "auth": params["auth"],
         }
-        for key in ("curl_options", "impersonate"):
-            if params.get(key):
+        for key in ("curl_options", "impersonate", "default_headers"):
+            if params.get(key) is not None:
                 request_args[key] = params[key]
 
         verify = params.get("verify")