Prompt API playground

Prompts an on-device small language model through the standard Prompt API shape (window.LanguageModel). Uses the browser's native Prompt API when it is present (for example Chrome or Edge with the on-device model flag enabled), and otherwise falls back to the experimental prompt-api-polyfill running a small model locally through Transformers.js, which works in any modern browser. The same code path is used either way.

Detecting provider...
The model can call the tools below. It emits a TOOL_CALL, the app runs the tool and feeds the result back, and the model reprocesses it into the answer (this is emulated, not native tool calling; see the README). Try "What is today's date?" for the date tool, or ask a two-tool question like "What did I miss at the offsite, and how long ago was it?" (compound questions are harder and work best on the 3B or native model). The model picker only affects the polyfill path; WebGPU is needed for usable speed (CPU/WASM is very slow).
Tools

Output



      

        Provider: -
        Model: -
        Tools: -
        Total time: - ms