Use an accordion as the <details> didn't work

app.py

@@ -179,9 +179,12 @@ with gr.Blocks(title="Quantized Retrieval") as demo:
 <h1 style='margin-top: 0;'>Quantized Retrieval - Binary Search with Scalar (int8) Rescoring</h1>
 
 This demo showcases retrieval using [quantized embeddings](https://huggingface.co/blog/embedding-quantization) on a CPU. The corpus consists of [41 million texts](https://huggingface.co/datasets/sentence-transformers/quantized-retrieval-data) from Wikipedia articles.
-
-<details><summary>Click to learn about the retrieval process</summary>
-
+</div>
+"""
+            )
+            with gr.Accordion("Click to learn about the retrieval process", open=False):
+                gr.Markdown(
+                    """
 Details:
 1. The query is embedded using the [`mixedbread-ai/mxbai-embed-large-v1`](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1) SentenceTransformer model.
 2. The query is quantized to binary using the `quantize_embeddings` function from the SentenceTransformers library.
@@ -191,8 +194,8 @@ Details:
 6. The top 20 documents are sorted by score.
 7. The titles and texts of the top 20 documents are loaded on the fly from disk and displayed.
 
-This process is designed to be memory efficient and fast, with the binary index being small enough to fit in memory and the int8 index being loaded as a view to save memory. 
-In total, this process requires keeping 1) the model in memory, 2) the binary index in memory, and 3) the int8 index on disk. With a dimensionality of 1024, 
+This process is designed to be memory efficient and fast, with the binary index being small enough to fit in memory and the int8 index being loaded as a view to save memory.
+In total, this process requires keeping 1) the model in memory, 2) the binary index in memory, and 3) the int8 index on disk. With a dimensionality of 1024,
 we need `1024 / 8 * num_docs` bytes for the binary index and `1024 * num_docs` bytes for the int8 index.
 
 This is notably cheaper than doing the same process with float32 embeddings, which would require `4 * 1024 * num_docs` bytes of memory/disk space for the float32 index, i.e. 32x as much memory and 4x as much disk space.
@@ -202,10 +205,6 @@ Feel free to check out the [code for this demo](https://huggingface.co/spaces/se
 
 Notes:
 - The approximate search index (a binary Inverted File Index (IVF)) is in beta and has not been trained with a lot of data.
-
-</details>
-
-</div>
 """
             )
             query = gr.Textbox(