Searxng

Using SearXNG with RAGFlow

Introduction to SearXNG in RAGFlow Agents

RAGFlow provides powerful agent capabilities, including built-in web search tools for real-time information retrieval during reasoning or chat sessions. The default web search often relies on cloud providers like Tavily, which require API keys and may incur costs.

SearXNG serves as a privacy-focused, open-source metasearch engine that aggregates results from multiple sources (e.g., Google, Bing, DuckDuckGo) without tracking users. Deploying SearXNG locally offers a fully self-hosted alternative, ideal for enterprises prioritizing data privacy, avoiding external API dependencies, or operating in restricted networks.

Although RAGFlow does not yet include native SearXNG support in its UI (as of early 2026 releases), users can integrate it easily via custom agent tools or by pointing the web search component to a local SearXNG instance. This approach mirrors integrations in similar open-source tools like Open WebUI, AnythingLLM, and Perplexica.

Benefits of Using SearXNG with RAGFlow

• Full Privacy and Control — All searches stay within your infrastructure. No data leaks to third-party APIs, making it perfect for sensitive or regulated environments.
• No Costs or Rate Limits — Unlike Tavily or SerpAPI, SearXNG runs free with unlimited queries (limited only by your hardware).
• Customizable Search Engines — Configure SearXNG to use specific engines, add restrictions, or focus on reliable sources.
• Hybrid Real-Time Retrieval — Agents combine internal knowledge bases (private documents) with fresh web results, reducing hallucinations on current events or external facts.
• Intranet Compatibility — Deploy SearXNG in air-gapped or intranet-only setups. For purely internal “browsing,” configure it to search indexed intranet pages (via custom engines or plugins) or pair it with RAGFlow’s built-in web page ingestion for static internal sites.

Benefits for Intranet Page Browsing

RAGFlow excels at ingesting and querying private/internal documents, but dynamic intranet pages (e.g., wikis, dashboards) may require real-time access.

• With Internet Access: SearXNG provides standard web search while keeping queries internal to your server.
• Without Internet (Pure Intranet):
  • Use RAGFlow’s native web crawling/ingestion to periodically index intranet URLs into knowledge bases.
  • Configure SearXNG with custom “site:” restrictions or internal search engines (e.g., via plugins for mediawiki or confluence) to query only intranet domains.
  • This creates a “local web search” experience where agents retrieve up-to-date content from internal sites without exposing data externally.

Deployment in Docker Containers

Both RAGFlow and SearXNG deploy easily with Docker, often on the same network for seamless communication.

1. Deploy SearXNG

Use the official searxng/searxng image or the dedicated docker repo.

Example docker-compose.yml for SearXNG:

version: ‘3’ services: searxng: image: searxng/searxng:latest container_name: searxng ports: - “8080:8080” volumes: - ./searxng:/etc/searxng environment: - BASE_URL=http://localhost:8080 restart: unless-stopped

Key configuration in searxng/settings.yml (mounted volume):

search: formats: - html - json # Required for API/tool calls server: limiter: false # Disable rate limiting for agent use secret_key: your_strong_secret

Run: docker compose up -d

2. Integrate with RAGFlow

• Deploy RAGFlow normally (via its official docker-compose.yml).
• Place both in the same Docker network or use host networking.
• In RAGFlow’s Agent builder:
  • Create a custom tool that calls your local SearXNG API (http://searxng:8080/search?q={query}&format=json).
  • Or, if using the built-in Websearch Agent, configure it (via env vars or custom code) to route to your SearXNG endpoint instead of Tavily.
• Many users report success with similar custom integrations in agent frameworks.

Combined Example Network

Extend RAGFlow’s compose file to include SearXNG:

services: # … existing RAGFlow services … searxng: image:
searxng/searxng:latest ports: - “8080:8080” volumes: -
./searxng:/etc/searxng networks: - ragflow_network

This setup keeps everything containerized, secure, and scalable.

example on homelab

services:
  caddy:
    container_name: caddy
    image: docker.io/library/caddy:2-alpine
    network_mode: host
    restart: unless-stopped
    volumes:
      - ./Caddyfile:/etc/caddy/Caddyfile:ro
      - caddy-data:/data:rw
      - caddy-config:/config:rw
    environment:
      - SEARXNG_HOSTNAME=${SEARXNG_HOSTNAME:-http://192.168.0.213}
      - SEARXNG_TLS=${LETSENCRYPT_EMAIL:-internal}
    logging:
      driver: "json-file"
      options:
        max-size: "1m"
        max-file: "1"

  searxng:
    container_name: searxng
    image: docker.io/searxng/searxng:latest
    restart: unless-stopped
    networks:
      - searxng
    ports:
      - "192.168.0.213:8080:8080"
    volumes:
      - ./searxng:/etc/searxng:rw
    environment:
      - SEARXNG_BASE_URL=https://${SEARXNG_HOSTNAME:-192.168.0.213}/
    depends_on:
      - caddy
    logging:
      driver: "json-file"
      options:
        max-size: "1m"
        max-file: "1"

networks:

searxng:

volumes:

caddy-data: caddy-config:

Conclusion

Integrating SearXNG with RAGFlow delivers a powerful, private alternative to cloud-based web search in agents—especially valuable for intranet deployments where data sovereignty matters. While awaiting potential native support, the custom tool approach works reliably and aligns with RAGFlow’s extensible design.

For the latest updates, monitor RAGFlow’s GitHub issues/releases or community discussions. This combination empowers fully self-hosted, real-time augmented agents without compromising privacy.