Search model / interactive preview
Describe what you want to find. The model turns your query into a search tool call with filters, tags, language, and intent, entirely in your browser.
Model input
Model output
Generate a search tool call from the example query, or try your own.
Search tool calls
Decode trace
Under the hood
A packaged language model and its inference engine execute inside your browser. Your content never enters a server-side inference queue.
The 5.6 MB model artifact enters runtime memory and initializes inside a dedicated worker for this interactive preview.
The IR runtime tokenizes the structured prompt and uses WebAssembly SIMD acceleration when supported, with a JavaScript fallback.
Inspect the JSON and expand the timing and token diagnostics. Tool calls are displayed here; no search is executed.