Web Search
Give a chat-completions request access to current web information.
Use gonkagate:web_search when a request may need current information from the web. GonkaGate runs its existing Web Search capability and gives the model web-grounded context for the final answer.
Quick start
export GONKAGATE_API_KEY="gp-your-api-key"
curl https://api.gonkagate.com/v1/chat/completions \
-H "Authorization: Bearer $GONKAGATE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen/qwen3-235b-a22b-instruct-2507-fp8",
"messages": [
{
"role": "user",
"content": "Find the latest public information about Gonka Network and summarize it with sources."
}
],
"tools": [
{ "type": "gonkagate:web_search" }
]
}'How it works
- Add
{ "type": "gonkagate:web_search" }totools. - GonkaGate uses the same web-search path as the existing Web Search feature.
- The model receives web results and uses them to produce the final answer.
Configuration
The server tool does not accept request-level parameters.
{
"tools": [
{ "type": "gonkagate:web_search" }
]
}Do not send parameters on the server tool. If you need request-level Web Search plugin fields, use Web Search Plugin.
Compatibility
- Works with streaming and non-streaming chat completions.
- Can be used next to user-defined function tools.
- Can be used next to
gonkagate:datetime. - Do not add both
gonkagate:web_searchand thewebplugin unless you intentionally want the same request to use the existing plugin path. New server-tool requests should usually use onlytools.
Common errors
| Error code | What it means | Fix |
|---|---|---|
invalid_server_tool_parameters | The request sent unsupported parameters for gonkagate:web_search. | Remove parameters or switch to the Web Search plugin if you need plugin-specific fields. |
unsupported_server_tool | The request used an unknown gonkagate:* tool type. | Use gonkagate:web_search, gonkagate:web_fetch, or gonkagate:datetime. |
See also
- Server Tools Overview
- Web Search Plugin for plugin fields and saved provider settings.
- Web Fetch when the model should read a specific URL instead of searching for sources.
Was this page helpful?