OpenHuman Guide
← Back to Guides

OpenHuman Web Search Setup — Real-Time Internet Search for Your AI

By default, OpenHuman uses a managed backend proxy that handles search requests through a curated set of search sources. You can also configure a custom or self-hosted search API for full control.

How it works

When the AI needs current information, it calls the web search tool. The search request is sent to the configured search provider, and the results are returned and compressed via TokenJuice before entering the model context.

Default setup

Out of the box, OpenHuman uses a managed backend proxy. No configuration needed — search works immediately after installation. This ensures privacy is maintained while still getting real-time results.

Self-hosted SearXNG

For maximum privacy, deploy your own SearXNG instance:

[web_search]
provider = "searxng"
base_url = "http://localhost:8888/search"
engines = ["google", "bing", "duckduckgo"]
max_results = 5

Tips

  • Prompt the AI with "search for" or "look up" to trigger web search explicitly
  • The AI can automatically search when it determines current data is needed
  • Search results are compressed via TokenJuice to minimize API costs
  • With a local-only setup (Ollama + no network), web search requires a network connection to the search backend