From cfd939318495821485c7b3bcc59c293afb29d94a Mon Sep 17 00:00:00 2001
From: matevip <matevip@mate.vip>
Date: Fri, 24 Apr 2026 23:25:03 +0800
Subject: [PATCH] fix(docker): bake searxng settings.yml into custom image

---
 docker-compose.yml          | 17 ++++-------
 docker/searxng/Dockerfile   |  9 ++++++
 docker/searxng/settings.yml | 61 +++++++++++++++++++++++++++++++++++++
 3 files changed, 76 insertions(+), 11 deletions(-)
 create mode 100644 docker/searxng/Dockerfile
 create mode 100644 docker/searxng/settings.yml

diff --git a/docker-compose.yml b/docker-compose.yml
index 82481181..b9ce2daf 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -33,12 +33,13 @@ services:
 
   # SearXNG 搜索引擎（keyless 搜索 provider）
   #
-  # ⚠️ deploy/searxng/settings.yml is bind-mounted because the upstream image
-  # ships with JSON format disabled and the anti-bot Limiter plugin enabled —
-  # both of which would silently break mateclaw's SearXNGSearchProvider.
-  # Do NOT remove that bind mount.
+  # The custom image at docker/searxng/ bakes in settings.yml so the sidecar
+  # works out of the box (upstream image ships JSON disabled + Limiter enabled,
+  # both of which silently break mateclaw's SearXNGSearchProvider).
+  # No host bind-mount — edit docker/searxng/settings.yml and rebuild.
   searxng:
-    image: searxng/searxng:latest
+    build:
+      context: ./docker/searxng
     container_name: mateclaw-searxng
     restart: unless-stopped
     environment:
@@ -46,11 +47,6 @@ services:
       - SEARXNG_SECRET=${SEARXNG_SECRET:-mateclaw-dev-searxng-secret-change-me}
       - UWSGI_WORKERS=2
       - UWSGI_THREADS=4
-    volumes:
-      - searxng_data:/etc/searxng
-      # This override MUST be mounted AFTER the named volume so it shadows
-      # the image's default settings.yml (enables JSON, disables limiter).
-      - ./deploy/searxng/settings.yml:/etc/searxng/settings.yml:ro
     ports:
       - "8088:8080"
     healthcheck:
@@ -103,4 +99,3 @@ services:
 volumes:
   mysql_data:
   server_data:
-  searxng_data:
diff --git a/docker/searxng/Dockerfile b/docker/searxng/Dockerfile
new file mode 100644
index 00000000..a5dbc4ad
--- /dev/null
+++ b/docker/searxng/Dockerfile
@@ -0,0 +1,9 @@
+# Custom SearXNG image for MateClaw.
+#
+# Bakes our settings.yml into /etc/searxng/settings.yml so the sidecar works
+# out of the box with no host bind-mount. The upstream image ships JSON output
+# disabled and the Limiter plugin enabled — both silently break mateclaw's
+# SearXNGSearchProvider, so this override is required.
+FROM searxng/searxng:latest
+
+COPY settings.yml /etc/searxng/settings.yml
diff --git a/docker/searxng/settings.yml b/docker/searxng/settings.yml
new file mode 100644
index 00000000..f159258b
--- /dev/null
+++ b/docker/searxng/settings.yml
@@ -0,0 +1,61 @@
+# SearXNG config for MateClaw's bundled search sidecar.
+#
+# Two things differ from the upstream default:
+#   1. JSON output format is enabled — mateclaw's SearXNGSearchProvider
+#      queries /search?format=json and fails silently without this.
+#   2. The anti-bot Limiter plugin is disabled — it otherwise rejects
+#      server-side HTTP calls (no JS, no cookies) with HTTP 429.
+#
+# This file is baked into the custom image via docker/searxng/Dockerfile —
+# do NOT bind-mount it from the host (prior host bind-mount broke deploys
+# where the host directory did not exist and Docker auto-created an empty
+# dir over the path).
+#
+# See https://docs.searxng.org/admin/settings/ for all knobs.
+
+use_default_settings: true
+
+general:
+  # Cosmetic only; shown in the UI title.
+  instance_name: "MateClaw Search"
+  # Keep this private — no outbound metrics.
+  donation_url: false
+  contact_url: false
+  enable_metrics: false
+
+search:
+  safe_search: 0
+  autocomplete: ""
+  default_lang: "auto"
+  formats:
+    - html
+    - json    # REQUIRED for mateclaw integration
+
+server:
+  # Override the default dev secret; docker-compose passes SEARXNG_SECRET in.
+  secret_key: "${SEARXNG_SECRET:-please-change-me-to-a-random-32-char-string}"
+  # Trust Docker's internal network — the reverse-proxy / rate-limit plugin
+  # uses this to know the caller's real IP.
+  limiter: false
+  image_proxy: false
+  # Bind address matches the container default.
+  bind_address: "0.0.0.0"
+  port: 8080
+
+ui:
+  static_use_hash: true
+
+# The default engine list is huge; keep a tight set of reliable ones.
+engines:
+  - name: duckduckgo
+    disabled: false
+  - name: bing
+    disabled: false
+  - name: brave
+    disabled: false
+  - name: wikipedia
+    disabled: false
+  - name: google
+    disabled: false
+  - name: startpage
+    disabled: false