How to Empower LLMs with a Private SearXNG Search Engine for Real‑Time Knowledge

SearXNG: An Ideal Foundation for Private Search Engines

SearXNG is an open‑source meta‑search engine that aggregates results from Google, Bing, Baidu and others while protecting user privacy. It offers a meta‑search mechanism, high customizability, open‑source transparency (AGPL‑3.0), and lightweight Docker deployment.

Deploying the Private Search Infrastructure

Docker is the simplest deployment method; the official SearXNG Docker image supports multiple architectures.

<code>git clone [email protected]:pig-mesh/searxng-docker.git
cd searxng-docker

docker run \
    -d \
    -p 8080:8080 \
    -v "${PWD}/searxng:/etc/searxng" \
    -e "BASE_URL=http://0.0.0.0:8080/" \
    -e "INSTANCE_NAME=searxng" \
    registry.cn-hangzhou.aliyuncs.com/dockerhub_mirror/searxng:2025.2.20-28d1240fc</code>

After deployment, access the engine at

http://localhost:8080

.

LLM Online Search Integration Options

Once the private engine is running, integrate it with large language models. The following are common solutions.

LangChain4j + SearXNG Integration

Adding the Dependency

<code>&lt;dependency&gt;
    &lt;groupId&gt;dev.langchain4j&lt;/groupId&gt;
    &lt;artifactId&gt;langchain4j-community-web-search-engine-searxng&lt;/artifactId&gt;
    &lt;version&gt;1.0.0-beta2&lt;/version&gt;
&lt;/dependency&gt;</code>

Basic Search Implementation

<code>WebSearchEngine searchEngine = SearXNGWebSearchEngine.builder()
        .baseUrl("http://127.0.0.1:8080")
        .build();
// Execute search
WebSearchResults searchResults = searchEngine.search("Who is the US president?");
searchResults.toTextSegments().forEach(System.out::println);</code>

RAG Mode Integration

<code>&lt;dependency&gt;
    &lt;groupId&gt;dev.langchain4j&lt;/groupId&gt;
    &lt;artifactId&gt;langchain4j-easy-rag&lt;/artifactId&gt;
    &lt;version&gt;1.0.0-beta2&lt;/version&gt;
&lt;/dependency&gt;</code>

Combine the search engine with a language model to enable retrieval‑augmented generation:

<code>// Create model client (replace with other LLM APIs)
OpenAiChatModel model = OpenAiChatModel.builder()
        .baseUrl("https://api.deepseek.com/v1")
        .modelName("deepseek-chat")
        .apiKey("sk-xxx")
        .build();

WebSearchEngine searchEngine = SearXNGWebSearchEngine.builder()
        .baseUrl("http://127.0.0.1:8080")
        .build();

WebSearchContentRetriever contentRetriever = WebSearchContentRetriever.builder()
        .webSearchEngine(searchEngine)
        .maxResults(3)
        .build();

SearchEnabledAssistant assistant = AiServices.builder(SearchEnabledAssistant.class)
        .contentRetriever(contentRetriever)
        .chatLanguageModel(model)
        .build();

String answer = assistant.answer("Who is the US president?");
System.out.println(answer);</code>

Advanced Optimizations

Further tuning can improve the search experience. SearXNG allows configuring the list of upstream engines and prioritizing Chinese search services to reduce latency and avoid restrictions.

Simplify the engine list : enable only necessary engines and disable slow or rarely used ones.

Support Chinese search engines : SearXNG includes 360search, Baidu, Quark, Sogou, etc.