Return large number of matching documents

[richardehughes]

Here is my scenario, which I have implemented using FAISS. It works but is painfully slow (~5 minutes to completion). So I want to know if I can do this faster using Milvus.

• 100M documents in the dataset
• I want to return all documents with a cosine similarity of 0.7 or greater to a vector that I supply
• Sometimes the returned dataset has 10k documents. Sometimes it has 100k documents.

I would like all of the documents, as well as the cosine similarity to the search vector I am using. Is this possible in milvus? Reading through the documentation it was not clear exactly how to do this.

[xiaofan-luan]

checkout https://milvus.io/docs/range-search.md#Range-Search and https://milvus.io/docs/with-iterators.md

that's probably what you want~