documentation
documentation copied to clipboard
v1.11: New embedder option: `binaryQuantized`
This release introduces a new embedder setting, binaryQuantized
. This setting converts embeddings of floating point numbers into boolean values. This reduces relevancy of semantic search results, but greatly reduces database size.
Activating binaryQuantized
is irreversible. Once a user sets it to true
, all relevant information is irrevocably and the only way to get it back is to re-vectorize the whole index in a new embedder.
Task
- [x] Update
reference/api/settings#body-18
References
- PRD: https://www.notion.so/meilisearch/Handle-a-larger-vector-dataset-50f2d10618bf480bba0d994e6571be0c
- SME: @irevoire
- Engine issue: https://github.com/meilisearch/meilisearch/issues/4873