Connecting OpenWebUI to a Search API for Grounded Responses

OpenWebUI defaults to SearXNG for web search. Connecting it to a commercial search API improves reliability and adds structured result types — shopping, news, Reddit, YouTube — that SearXNG cannot provide. The tradeoff is a small per-query cost vs. free but higher-maintenance self-hosted search.

The SearXNG Problem

SearXNG is OpenWebUI's recommended self-hosted search backend. It aggregates results from Google, Bing, DuckDuckGo, and others. Problems in production:

• Google blocks SearXNG aggressively when queries come from a single server IP
• Results degrade silently when upstream sources block the instance
• Maintaining a reliable SearXNG instance requires proxy configuration and updates
• No structured result types beyond basic web search

For a personal OpenWebUI instance with low query volume, SearXNG works acceptably. For a team or production deployment, the maintenance burden exceeds the cost of a commercial API.

OpenWebUI Search API Configuration

OpenWebUI supports custom search endpoints via the admin settings. The endpoint must accept a query and return a specific JSON format.

Create a lightweight adapter that translates between OpenWebUI's expected format and your search API:

from flask import Flask, request, jsonify
import requests

app = Flask(__name__)

@app.route('/search', methods=['GET'])
def search():
    query = request.args.get('q', '')
    if not query:
        return jsonify({"results": []})

    # Call Scavio search API
    resp = requests.post(
        "https://api.scavio.dev/api/v1/search",
        headers={"x-api-key": SCAVIO_KEY},
        json={"query": query, "num": 5}
    )
    data = resp.json()

    # Transform to OpenWebUI expected format
    results = []
    for r in data.get("organic_results", []):
        results.append({
            "title": r.get("title"),
            "url": r.get("link"),
            "content": r.get("snippet")
        })

    return jsonify({"results": results})

if __name__ == '__main__':
    app.run(port=8080)

Deploy this adapter on your OpenWebUI server. In OpenWebUI admin settings, set:

• Web Search Engine: custom
• Search API URL: http://localhost:8080/search?q={query}

Adding Brave Search as Alternative

Brave Search's API response format is close to OpenWebUI's expected format with minimal transformation:

@app.route('/brave-search', methods=['GET'])
def brave_search():
    query = request.args.get('q', '')
    resp = requests.get(
        "https://api.search.brave.com/res/v1/web/search",
        headers={"X-Subscription-Token": BRAVE_KEY},
        params={"q": query, "count": 5}
    )
    data = resp.json()
    results = [{
        "title": r["title"],
        "url": r["url"],
        "content": r.get("description", "")
    } for r in data.get("web", {}).get("results", [])]
    return jsonify({"results": results})

Brave's $5/month free credit gives ~1,000 searches. For a personal OpenWebUI instance averaging 30 searches/day, that is roughly 900 searches/month — staying within the free tier.

Cost for a Team Deployment

For a team of 10 using OpenWebUI with 20 web searches per person per day:

• 200 searches/day
• 6,000 searches/month
• At Scavio $0.005/credit: $30/month
• At Brave PAYG $5/1k: $30/month

Both are comparable at this volume. Scavio adds multi-platform coverage (Amazon, YouTube, Reddit results) which Brave does not have. For a team that needs platform-specific search results from within OpenWebUI, multi-platform is worth the equivalent cost.

Production Considerations

For a production OpenWebUI deployment:

1. Add API key rotation if using provider rate limits (unlikely to need for team-scale)
2. Cache frequent searches (common queries in an internal tool repeat often)
3. Add query logging to understand which searches are triggering web lookup — this tells you what to add to your LLM system prompt to avoid unnecessary searches
4. Set a per-user daily search limit if cost control is important

The adapter pattern above also lets you switch providers without changing OpenWebUI configuration — just update the adapter's target API.