Update js demo README.md with link to Gemma 4 prompt formatting guide

Author: tyrmullen

examples/llm_inference/js/README.md

@@ -13,5 +13,5 @@ This web sample demonstrates how to use the LLM Inference API to run common text
 
 Follow the following instructions to run the sample on your device:
 1. Copy the [index.html](https://github.com/googlesamples/mediapipe/blob/main/examples/llm_inference/js/index.html) file to your computer.
-2. Download a pre-converted Gemma model, like [Gemma 3 4B](https://huggingface.co/litert-community/Gemma3-4B-IT/resolve/main/gemma3-4b-it-q4_0-web.task?download=true). See the [Web Models](https://huggingface.co/collections/litert-community/web-models) LiteRT community collection for more supported models. For image and audio inputs, see our [multimodal documentation](https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference/web_js#multimodal).
+2. Download a pre-converted Gemma model, like [Gemma 3 4B](https://huggingface.co/litert-community/Gemma3-4B-IT/resolve/main/gemma3-4b-it-q4_0-web.task?download=true). See the [Web Models](https://huggingface.co/collections/litert-community/web-models) LiteRT community collection for more supported models. For image and audio inputs, see our [multimodal documentation](https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference/web_js#multimodal). For Gemma 4 models, see [new prompt formatting](https://ai.google.dev/gemma/docs/core/prompt-formatting-gemma4).
 3. Open the index.html file in your browser and upload your Gemma model using the "Choose File" button. The model will start loading. After a few seconds, a 'Get Response' button will be enabled and you can send queries to the LLM.