Hello qdrant people. I used the engine to create a semantic search application using a corpus of 4 Million Terms and their embeddings and adding in the payload the source these terms are coming from. These terms come from 5 sources, where one source is dominent (3_900_000 terms come from this specific source). The hsnw parameters used to create that collection are ef=128, m=32. I index at the end of uploading all the data in that collection. When i do an unfiltered search with a vector that i know exists in the collection and comes from a non dominent source, the results are consistent and correct. When i do a search with filter specifying only that non dominent source , i dont retrieve back the correct result. Do you know how can i resolve this, because the major selling point is the filterable part of the search.
