[Bug]: No 'batch_root' folder in tutorial notebook of vector search index

[BanJeoker]

File Name

generative-ai/blob/main/embeddings/intro-textemb-vectorsearch.ipynb

What happened?

In this sentence, Then, create a new Cloud Storage bucket and copy the file to it. BUCKET_URI = f"gs://{PROJECT_ID}-embvs-tutorial-{UID}" ! gsutil mb -l $LOCATION -p {PROJECT_ID} {BUCKET_URI} ! gsutil cp questions.json {BUCKET_URI}

You need to create a 'batch_root' folder in the bucket you just created, otherwise the index created will not use that json file, it will be an empty index.

Relevant log output

No response

Code of Conduct

• [X] I agree to follow this project's Code of Conduct

[kazunori279]

Hi @BanJeoker , did you actually ran the notebook and saw an empty index created? I've been creating index without the batch_root folder for many times, and haven't seen any issues..

[BanJeoker]

@kazunori279 yes I ran the notebook, the index can be created without the batch_root folder, however, in that case, the index will have no dense count, a query will return empty list back. Also, isn't creating the batch_root folder emphasized in the official documentation here

https://cloud.google.com/vertex-ai/docs/vector-search/setup/format-st

Thanks

[kazunori279]

Hi @BanJeoker , I just went through the notebook from the start, and got a proper result:

So the use of the batch_root folder may not be the cause why you are getting an empty result. But agree, it's better to add the use of batch_root folder to this sample to be consistent with the document.

[BanJeoker]

thank you Kaz