Deploying SearXNG in Docker: Giving Local AI Agents Private Web Search Capabilities

Overview

Local AI agents (via Open-WebUI or LangChain) often need real-time web search capabilities to ground their responses. Relying on commercial search APIs (like Google Custom Search or Bing API) leaks agent queries and user prompts to third parties.

SearXNG is a privacy-respecting, open-source metasearch engine. Deploying SearXNG in Docker provides local AI agents with a JSON search API while stripping tracking cookies and user logs.

---

Docker Compose Setup

Create /opt/searxng-stack/docker-compose.yml:

version: "3.8"

services:
  searxng:
    image: searxng/searxng:latest
    container_name: searxng_search
    restart: unless-stopped
    ports:
      - "8080:8080"
    volumes:
      - ./searxng:/etc/searxng:rw
    environment:
      - SEARXNG_BASE_URL=http://localhost:8080/

Configure JSON API output in ./searxng/settings.yml:

search:
  formats:
    - html
    - json

---

Verification & Open-WebUI Integration

Test SearXNG JSON API endpoint:

curl "http://localhost:8080/search?q=OpsNexusAI&format=json"

In Open-WebUI, set Web Search Engine to searxng and URL to http://searxng:8080/search.

---

Redis Caching Sidecar for High-Performance Search Responses

To prevent rate-limiting from upstream search engines (Google, DuckDuckGo, Wikipedia), deploy a Redis caching sidecar alongside SearXNG.

Complete docker-compose.yml with Redis Cache

version: "3.8"

services:
  searxng:
    image: searxng/searxng:latest
    container_name: searxng_search
    restart: unless-stopped
    ports:
      - "8080:8080"
    volumes:
      - ./searxng:/etc/searxng:rw
    environment:
      - SEARXNG_BASE_URL=http://localhost:8080/
    depends_on:
      - redis

  redis:
    image: redis:7-alpine
    container_name: searxng_redis
    restart: unless-stopped
    command: redis-server --save 30 1 --loglevel warning
    volumes:
      - redis_data:/data

volumes:
  redis_data:

Architecture: Privacy-Preserving Agentic RAG

AI agents require web search access for real-time information retrieval. However, routing search queries directly to commercial search APIs (Google, Bing) exposes internal prompt contexts, proprietary tokens, and user IP addresses to third parties.

SearXNG acts as an aggressive privacy-preserving meta-search aggregator. It strips tracking IDs, rotates user-agents, pools results from 70+ search engines, and exposes a clean JSON REST API for frameworks like LangChain, LlamaIndex, and AutoGen.

+-------------------------------------------------------------------------+
|                         Local AI Agent Stack                            |
|  +-------------------------------------------------------------------+  |
|  | LangChain / LlamaIndex / AutoGen Agent Framework                  |  |
|  +-------------------------------------------------------------------+  |
|                                   | JSON API (SearXNG API Endpoint)     |
|                                   \/                                    |
|  +-------------------------------------------------------------------+  |
|  | SearXNG Aggregator Engine (Local Docker Container)                |  |
|  +-------------------------------------------------------------------+  |
|         ||                        ||                       ||           |
|         \/                        \/                       \/           |
|   +-----------+            +--------------+          +-------------+    |
|   | DuckDuckGo|            | Wikipedia    |          | Qwant       |    |
|   +-----------+            +--------------+          +-------------+    |
+-------------------------------------------------------------------------+

Step 1: Production SearXNG Deployment Manifest (docker-compose.yml)

version: "3.8"

services:
  searxng:
    image: searxng/searxng:latest
    container_name: private-searxng
    restart: unless-stopped
    ports:
      - "8080:8080"
    volumes:
      - ./searxng:/etc/searxng:rw
    environment:
      - SEARXNG_BASE_URL=http://localhost:8080/
      - INSTANCE_NAME=OpsNexus-Private-Search
    cap_drop:
      - ALL
    cap_add:
      - CHOWN
      - SETGID
      - SETUID

  redis:
    image: redis:7-alpine
    container_name: searxng-redis
    restart: unless-stopped
    command: redis-server --save "" --appendonly no

Step 2: SearXNG Configuration (searxng/settings.yml)

Enable JSON output format for AI agent integration:

use_default_settings: true
server:
  port: 8080
  bind_address: "0.0.0.0"
  secret_key: "GENERATE_SECURE_RANDOM_SECRET_KEY_HERE"
  limiter: false

search:
  safe_search: 0
  autocomplete: ""
  formats:
    - html
    - json

engines:
  - name: duckduckgo
    engine: duckduckgo
    shortcut: ddg
    disabled: false

  - name: wikipedia
    engine: wikipedia
    shortcut: wp
    disabled: false

Step 3: Python AI Agent Integration Script

Connect SearXNG to a Python AI Agent:

import requests

def searxng_search(query: str, limit: int = 5):
    url = "http://localhost:8080/search"
    params = {
        "q": query,
        "format": "json",
        "categories": "general"
    }
    response = requests.get(url, params=params, timeout=5)
    data = response.json()
    
    results = []
    for item in data.get("results", [])[:limit]:
        results.append({
            "title": item.get("title"),
            "url": item.get("url"),
            "snippet": item.get("content")
        })
    return results

# Example call
search_results = searxng_search("Latest CUDA 12.4 container optimization techniques")
print(search_results)