What Search Backend Does Vane Use? SearXNG Integration Explained

Vane uses SearXNG as its meta-search backend, fetching web results via a configurable HTTP endpoint exposed at <searxngURL>/search?format=json.

The open-source Vane project (ItzCrazyKns/Vane) relies on SearXNG to power its web search capabilities rather than implementing a crawler directly. By leveraging SearXNG's aggregated search results, Vane provides users with privacy-focused, configurable web search while maintaining complete control over the data source through environment variables or JSON configuration.

How Vane Integrates with SearXNG

Vane's search architecture delegates all web queries to a user-configured SearXNG instance. The integration follows a clean separation between the search client, configuration management, and server registry.

The Search Client Implementation

The core search logic resides in src/lib/searxng.ts, which exports the searchSearxng function. This helper constructs and executes HTTP requests to the SearXNG API, appending ?format=json to ensure machine-readable responses. When invoked, it contacts the configured instance at <searxngURL>/search?format=json and returns parsed results including suggestions and result items.

Configuration Management

The SearXNG endpoint URL is defined in src/lib/config/index.ts under the search UI configuration section, specifically within the searxngURL key. At runtime, the server registry in src/lib/config/serverRegistry.ts exposes getSearxngURL(), which retrieves the current configuration value for use by the search client.

Configuring the SearXNG Backend

Vane supports two methods for specifying the SearXNG instance: environment variables or static JSON configuration.

Environment Variable Method

Set the SEARXNG_API_URL variable before starting the server:

export SEARXNG_API_URL="http://localhost:4000"

When the ConfigManager initializes, it reads this value and exposes it through getSearxngURL().

JSON Configuration Method

Alternatively, specify the URL in data/config.json:

{
  "search": {
    "searxngURL": "http://my-searxng-instance.example.com"
  }
}

Using the Search API

The backend exposes web search functionality through the /api/search route. When the request includes "sources": ["web"], the handler routes the query to searchSearxng.

Example API Request

A sample POST request body:

{
  "chatModel": { "providerId": "openai", "key": "gpt-4o-mini" },
  "embeddingModel": { "providerId": "openai", "key": "text-embedding-3-large" },
  "sources": ["web"],
  "query": "What is Vane?"
}

The system forwards the query to the configured SearXNG instance and returns aggregated web results alongside AI-generated responses.

Performing Searches Programmatically

Once configured, the searchSearxng function accepts query strings and optional parameters:

import { searchSearxng } from '@/lib/searxng';

// Simple query
const { results, suggestions } = await searchSearxng('latest AI research');

// Advanced options
const { results } = await searchSearxng('open-source LLMs', {
  categories: ['general'],
  engines: ['google', 'bing'],
  language: 'en',
  pageno: 2,
});

Available options include categories for result types, engines for specific search providers, language for localization, and pageno for pagination.

Architecture Overview

According to the project's docs/architecture/README.md, Vane employs "A meta search backend... to fetch relevant web results." This refers specifically to the SearXNG integration, which acts as a privacy-respecting intermediary between Vane and various search providers. The architecture allows operators to self-host SearXNG or connect to any reachable instance, ensuring data sovereignty and customization flexibility.

Summary

• Vane uses SearXNG as its exclusive web search backend, implemented through the searchSearxng helper in src/lib/searxng.ts.
• Configuration occurs via the SEARXNG_API_URL environment variable or the search.searxngURL JSON config key, retrieved at runtime by getSearxngURL().
• API requests target <searxngURL>/search?format=json and support filtering by categories, engines, language, and pagination.
• Architecture documentation explicitly identifies SearXNG as the meta-search backend powering web queries.

Frequently Asked Questions

Does Vane implement its own web crawler?

No. According to the source code in src/lib/searxng.ts, Vane delegates all web search operations to a configured SearXNG instance. SearXNG acts as the meta-search engine that aggregates results from multiple providers, while Vane handles the presentation and AI-enhanced processing of those results.

How do I change the SearXNG instance Vane connects to?

Modify the SEARXNG_API_URL environment variable or update the search.searxngURL value in your data/config.json file. The ConfigManager loads this value at startup, making it available through getSearxngURL() in src/lib/config/serverRegistry.ts. Restart the server after making changes.

What parameters can I pass to the searchSearxng function?

The function accepts a query string and an options object supporting categories (array of strings), engines (array of provider names), language (ISO code), and pageno (number). These map directly to SearXNG's API parameters, allowing fine-grained control over result sources and pagination.

Can I use Vane without hosting my own SearXNG instance?

Yes. While self-hosting SearXNG is recommended for privacy, you can configure Vane to use any publicly accessible SearXNG instance by setting the searxngURL configuration value to the instance's base URL. The application will route all web searches through that endpoint.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}