The post asks whether the llama-index library uses KNN or HNSW for its automerge engine. The community members discuss this, with some indicating that the default vector store in llama-index is a simple numpy array with pairwise cosine similarity, and that HNSW would only be available if the vector store used supports it. One community member suggests that if Chroma, which uses HNSW by default, is integrated with the llama-index engine, then HNSW could be used. However, there is no explicitly marked answer in the comments.
