3 ms·
Looks great. How does it compare to hnswlib? How many vectors can it handle for subsecond queries? Is it approximate or precise nearest neighbor?
by leobg 5y ago
Looks great. How does it compare to hnswlib? How many vectors can it handle for subsecond queries? Is it approximate or precise nearest neighbor?
- neumll 5y agoThank you. The indexes are built on top of either Faiss, Hnswlib or Annoy depending on the settings - https://neuml.github.io/txtai/embeddings/configuration/#backend https://neuml.github.io/txtai/embeddings/configuration/#back... I've primarily focused on single query times, which have averaged around 5ms - 25ms depending on the size of the index. Queries can be batched so query times wouldn't increase linearly. You could batch quite a few and still get subsecond response times. All three libraries are approximate nearest neighbors but I know at least for Faiss, it can be configured to effectively be a precise query.
- leobg 5y agoThank you. I’ll check it out!