IdanR
12/30/2025, 9:24 AMparams = {
"model": "gpt-4.1-2025-04-14",
"tools": [{"type": "web_search"}], # OpenAI's native web search
"text": {"format": {"type": "json_schema", ...}}
}
response = client.responses.create(**params)Vaibhav (BAML 🐑)
12/30/2025, 1:39 PMIdanR
12/30/2025, 4:11 PMclient<llm> WebSearchClient {
provider openai-responses
options {
model "gpt-4.1"
tools [
{
type "web_search_preview"
}
]
}
}
Function (Notice the last sentence instruction):
function ResearchRecruitingMarket(
job_title: string,
location: string,
company_name: string?,
company_domain: string?
) -> string {
client WebSearchClient
prompt #"
You are a recruiting market researcher analyzing cold email reply rates for hiring campaigns.
# Research Task
...
# Instructions
...
# Output Format
Provide a concise summary (3-5 sentences) with estimated percentages.
Focus on actionable data that can inform funnel calculations.
If you hadn't use web search then state it in your title, so the user will know your base
"#
}
The result:
### 2025 Cold Email Reply Rates for Senior Backend Engineer at Google – San Francisco (Based on Expert Knowledge, No Web Search)\n\n Summary: ...IdanR
12/30/2025, 4:12 PMVaibhav (BAML 🐑)
12/30/2025, 4:44 PMVaibhav (BAML 🐑)
12/30/2025, 4:44 PM