To use TensorFlow.js to get the embeddings of text and use KNN to classify them, you can follow the following steps:

1. First, load the pre-trained Universal Sentence Encoder model into TensorFlow.js by using the following code:

```

const model = await use.load();

```

2. Next, preprocess your text data by tokenizing it and converting it into a tensor. You can use the following code to preprocess the text:

```

const sentences = ['Hello there', 'How are you?'];

const embeddings = await model.embed(sentences).array();

const tensors = embeddings.map(embedding => tf.tensor1d(embedding));

```

3. Once you have the preprocessed text data in the form of tensors, you can use the KNN algorithm to classify the data. You can use the following code to train the KNN classifier:

```

const knn = knnClassifier.create();

tensors.forEach((tensor, i) => knn.addExample(tensor, label[i]));

```

4. Finally, to classify new text data, you can use the following code:

```

const newSentence = 'Howdy';

const newEmbedding = await model.embed(newSentence).array();

const newTensor = tf.tensor1d(newEmbedding);

const result = await knn.predictClass(newTensor);

console.log(result.label); // Output: 'How are you?'

```

In the above code, the `predictClass()` method of the KNN classifier is used to predict the label of the new text data based on its embedding.