[Question] Example for using DiskANN for VectorDB on Python

[sleepingcat4]

I wanted to know if I could be provided some examples to understand how we can create a vectorDB using embeddings and correspodning url. And index it using Python. We have memory constraints like we have 8-16GBish RAM and SSD.

We are currently using Intel Xeon Processors and some super computing clusters but when we are finished we want to index it for cosine similarity on fairly okaysih spec laptop. Any guidance/help?

[daxpryce]

@sleepingcat4 this is entirely too late for my tests tastes - my apologies for not seeing this.

I have many examples of just this - I'm going to put them together now and add them as part of our documentation in workflows and will inform you there - I'm sure you've already moved on from this point, but I also wanted to keep you in the loop.

[daxpryce]

https://github.com/microsoft/DiskANN/pull/592/files should contain a mad dash through using the API. Most of the documentation is much more detailed and appears @ https://microsoft.github.io/DiskANN/docs/python/latest/diskannpy.html, but the python.md file in workflows in the linked PR should pull it all together in a bit of a cookbook example.

[ucas010]

have the same question
so great pr, it's beautiful if could be used in real

[daxpryce]

have the same question so great pr, it's beautiful if could be used in real

now that I have updated the docs in #600, can you try it again with the new examples?