Odysseus SearXNG setup for Deep Research

Why SearXNG matters

When an Odysseus research agent searches the web, it needs a search endpoint that returns usable results consistently. A self-hosted SearXNG instance can keep this under your control, but it does not magically bypass every search engine limit.

Basic local test

Before connecting Odysseus, open your SearXNG instance directly and run a normal browser query. If the web UI returns empty or blocked results, Odysseus will inherit that problem.

• Confirm the SearXNG web UI loads.
• Search for a simple query such as Odysseus AI GitHub.
• Check whether results come from multiple engines, not only Google.
• Review logs for rate limit, CAPTCHA, or blocked engine messages.

Connect it to Odysseus

In Odysseus, use the SearXNG base URL expected by the search or Deep Research setting. If Odysseus runs in Docker and SearXNG runs on the host, remember the same Docker networking rule:

http://host.docker.internal:<searxng-port>

If Google gets blocked

Google can block automated or high-volume search patterns. Do not treat that as an Odysseus bug. Use multiple engines, lower request frequency, and prefer citation quality over raw result count.

• Disable engines that constantly return CAPTCHA pages.
• Use conservative rate limits.
• Prefer engines with stable HTML/API behavior for your region.
• Keep research prompts narrow so agents do fewer searches.

Research quality checklist

Good Deep Research is not just search access. Require citations, dates, and source comparison in the agent prompt, especially for fast-moving open-source projects.