Web search
Han AI searches the open web by scraping the DuckDuckGo HTML endpoint directly. No API key, no per-query fee, no third-party SaaS in the loop.
What it does
Returns a ranked list of search results — title, URL, and a short snippet — for any query the AI decides to run.
| Field | Value |
|---|---|
| Schema name | web_search |
| Powered by | DuckDuckGo HTML scrape |
| API key required | No |
| Cost | Free |
When Han AI uses it
- A question requires current information your captured profile does not contain.
- A draft needs sourcing — a competitor’s pricing page, a regulator’s guidance, a news article.
- A research task spans multiple sources and needs candidates before deeper reading.
Search rarely runs alone. Han AI typically follows it up with page fetch or browser on the most promising result.
Examples
- “What did the central bank announce about reserve requirements last week?”
- “Find three suppliers for matte black anodised aluminium trims in Southeast Asia.”
- “Who currently advertises in that property segment?”
Limits
- Results are a snapshot of what DuckDuckGo returns at that moment.
- Heavily JavaScript-driven sites only expose a teaser snippet — Han AI escalates to the browser tool when needed.
- Rate limits are unpublished. If DuckDuckGo throttles, Han AI surfaces the failure rather than silently degrading.
Why this stack
DuckDuckGo HTML works without an account or API key, which means it costs nothing and never expires. Paid search APIs (Tavily, Perplexity, SerpAPI) only enter the conversation when a customer needs something they cannot get from open scraping.