pi commited on
Commit
17b0658
·
1 Parent(s): b3596ff

fix(search): give engines room to finish, and add ones a datacenter IP is allowed

Browse files

Two separate problems shared one symptom, so they were being read as one.

SearXNG's outgoing.request_timeout was 5.0. DuckDuckGo from a datacenter IP
routinely takes 5-7s, so SearXNG gave up on it and listed it unresponsive — the
identical "engine is broken" report a block produces, for an engine that was
only slow. Raised to 9.0, with Ael's race budget raised to match, so an engine
is judged on the results it returns rather than on our impatience. The happy
path is unaffected: the race still finishes on the first viable set.

Google and Wikipedia answer 200-with-nothing from Hugging Face's shared egress
IP. That left Bing carrying every query with no redundancy, which is not
resilience — a single-IP block on the one working engine takes out search
entirely. Added mojeek, marginalia, rightdao, mwmbl, presearch and qwant: the
engines that still answer a datacenter IP without a CAPTCHA, and deliberately
redundant with each other.

The pre-existing google/duckduckgo/bing/wikipedia entries are kept. A rate limit
is not a ban, and an engine that recovers should be picked up without another
config edit.

Files changed (1) hide show
  1. searxng_settings.yml +43 -2
searxng_settings.yml CHANGED
@@ -24,8 +24,14 @@ ui:
24
  static_use_hash: true
25
 
26
  outgoing:
27
- request_timeout: 5.0
28
- max_request_timeout: 15.0
 
 
 
 
 
 
29
  useragent_suffix: ""
30
  pool_connections: 100
31
  pool_maxsize: 20
@@ -65,3 +71,38 @@ engines:
65
  engine: arxiv
66
  shortcut: ar
67
  disabled: false
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
  static_use_hash: true
25
 
26
  outgoing:
27
+ # 5.0 was killing engines that legitimately need longer. DuckDuckGo from a
28
+ # datacenter IP regularly takes 5-7s, so at 5.0 SearXNG abandoned it and
29
+ # reported it unresponsive — the same "engine is broken" symptom a block
30
+ # produces, for an engine that was merely slow. Ael's race budget is 12s, so
31
+ # there is headroom to let the engine finish and be judged on its results
32
+ # rather than on our patience.
33
+ request_timeout: 9.0
34
+ max_request_timeout: 20.0
35
  useragent_suffix: ""
36
  pool_connections: 100
37
  pool_maxsize: 20
 
71
  engine: arxiv
72
  shortcut: ar
73
  disabled: false
74
+
75
+ # Added 2026-09-30. Google and Wikipedia both answer 200-with-nothing from
76
+ # Hugging Face's shared egress IP, which left Bing carrying every query. These
77
+ # are the engines that still answer a datacenter IP without a CAPTCHA, and
78
+ # they are redundant with each other on purpose: the failure being fixed is
79
+ # a single-IP block, and one working engine is not resilience.
80
+ - name: mojeek
81
+ engine: mojeek
82
+ shortcut: mjk
83
+ disabled: false
84
+
85
+ - name: marginalia
86
+ engine: marginalia
87
+ shortcut: mg
88
+ disabled: false
89
+
90
+ - name: rightdao
91
+ engine: rightdao
92
+ shortcut: rdao
93
+ disabled: false
94
+
95
+ - name: mwmbl
96
+ engine: mwmbl
97
+ shortcut: mw
98
+ disabled: false
99
+
100
+ - name: presearch
101
+ engine: presearch
102
+ shortcut: ps
103
+ disabled: false
104
+
105
+ - name: qwant
106
+ engine: qwant
107
+ shortcut: qw
108
+ disabled: false