XHToken/Spark-X2.6 ? reverse-engineered prompt

Build me a simple local demo for Spark X2.6 that lets me run the model on my own machine and chat with it in a clean web page or lightweight app.

I want it to use the Spark X2.6 4B model by default, with a way to switch to the 1.7B version if needed. Please include the basic setup for running it locally, and make sure it works with the common local model options mentioned in the docs, like Ollama or a direct Python setup. If there’s a good way to expose the model as an API and a chat interface, do that too.

It would be great if the app shows the answer clearly, supports long conversations, and lets me choose between no thinking, low thinking, and high thinking so I can trade off speed and quality. Keep it practical and easy to run, and look up current docs online if you need to.