Web Search

Not yet available. The Web Search server tool is not yet available.

The Web Search server tool gives any model access to live search results without you writing a search-and-inject loop. The gateway runs the search, formats the results, injects them into the conversation context, and lets the model answer — all within your single API call.

Enabling web search#

Add opendunes.web_search to the tools array on any chat-completions request:

{
  "model": "anthropic/claude-opus-4.8",
  "messages": [
    { "role": "user", "content": "What are the latest AI model releases this week?" }
  ],
  "tools": [
    { "type": "opendunes.web_search" }
  ]
}

The model decides when to search. If the question is fully answerable from its training data, it may skip the tool entirely.

Configuration#

You can tune web search behavior with an options object inside the tool entry:

{
  "type": "opendunes.web_search",
  "options": {
    "max_results": 5,
    "safe_search": true,
    "region": "dz"
  }
}

What the model receives#

Each search result is injected as a structured snippet:

{
  "title": "...",
  "url": "https://...",
  "snippet": "..."
}

The model uses these snippets to ground its answer and may cite sources inline.

Billing#

Web Search charges a flat fee per search invocation (regardless of result count), plus the token cost of injecting results into the context. Both line items appear in your per-request usage breakdown.

Combining with other tools#

Web Search can run alongside Web Fetch in the same request. A common pattern is to search for relevant URLs, then fetch one in a second tool call for deeper content — the gateway handles both hops automatically.