Use public SearXNG instances instead of bundling

app.py

@@ -3,8 +3,6 @@ import platform
 import sys
 import shutil
 import tempfile
-import subprocess
-import time
 import threading
 import torch
 import spaces
@@ -208,58 +206,50 @@ def _get_runtime_specs():
 
 
 # ---------------------------------------------------------------------------
-# SearXNG local instance
+# SearXNG search (uses public instances with JSON API)
 # ---------------------------------------------------------------------------
 
-SEARXNG_PORT = 8888
-SEARXNG_BASE = f"http://127.0.0.1:{SEARXNG_PORT}"
-_searxng_process = None
+SEARXNG_INSTANCES = [
+    "https://search.ononoki.org",
+    "https://searx.be",
+    "https://search.sapti.me",
+    "https://searxng.site",
+]
+_searxng_active_instance = None
 
 
-def _start_searxng():
-    """Start SearXNG as a background subprocess."""
-    global _searxng_process
-    settings_path = os.path.join(os.path.dirname(__file__), "searxng_settings.yml")
-    if not os.path.exists(settings_path):
-        _log("SearXNG settings not found, search disabled")
-        return False
+def _probe_searxng_instances():
+    """Find a working SearXNG instance."""
+    global _searxng_active_instance
+    for instance in SEARXNG_INSTANCES:
+        try:
+            url = f"{instance}/search?q=test&format=json"
+            req = urllib.request.Request(url, headers={"User-Agent": "RespiteAPI/1.0"})
+            resp = urllib.request.urlopen(req, timeout=5)
+            _json.loads(resp.read())
+            _searxng_active_instance = instance
+            _log(f"SearXNG instance ready: {instance}")
+            return
+        except Exception:
+            continue
+    _log("Warning: No working SearXNG instance found")
 
-    try:
-        _searxng_process = subprocess.Popen(
-            [sys.executable, "-m", "searx.webapp"],
-            env={**os.environ, "SEARXNG_SETTINGS_PATH": settings_path},
-            stdout=subprocess.PIPE,
-            stderr=subprocess.PIPE,
-        )
-        # Wait for SearXNG to start
-        for _ in range(30):
-            time.sleep(1)
-            try:
-                req = urllib.request.Request(
-                    f"{SEARXNG_BASE}/search?q=test&format=json",
-                    headers={"User-Agent": "RespiteAPI/1.0"},
-                )
-                urllib.request.urlopen(req, timeout=3)
-                _log("SearXNG started successfully")
-                return True
-            except Exception:
-                continue
-        _log("SearXNG failed to start within 30s")
-        return False
-    except Exception as e:
-        _log(f"SearXNG start failed: {e}")
-        return False
+
+# Probe instances in background so it doesn't block startup
+threading.Thread(target=_probe_searxng_instances, daemon=True).start()
 
 
 def _search_web(query, categories="general", max_results=10):
-    """Query the local SearXNG instance."""
+    """Query a public SearXNG instance."""
+    if not _searxng_active_instance:
+        raise RuntimeError("No SearXNG instance available")
     params = {
         "q": query,
         "format": "json",
         "categories": categories,
         "language": "en",
     }
-    url = f"{SEARXNG_BASE}/search?{urllib.parse.urlencode(params)}"
+    url = f"{_searxng_active_instance}/search?{urllib.parse.urlencode(params)}"
     req = urllib.request.Request(url, headers={"User-Agent": "RespiteAPI/1.0"})
     resp = urllib.request.urlopen(req, timeout=10)
     data = _json.loads(resp.read())
@@ -274,17 +264,12 @@ def _search_web(query, categories="general", max_results=10):
         })
     return {
         "query": query,
+        "instance": _searxng_active_instance,
         "results": results,
         "total_results": len(data.get("results", [])),
-        "categories": data.get("infoboxes", []),
     }
 
 
-# Start SearXNG in background thread so it doesn't block Gradio startup
-_searxng_thread = threading.Thread(target=_start_searxng, daemon=True)
-_searxng_thread.start()
-
-
 # ---------------------------------------------------------------------------
 # Model cache
 # ---------------------------------------------------------------------------