0.7.0 Plan
ankane opened this issue · comments
Andrew Kane commented
Hi all, just wanted to share the initial plan for 0.7.0, with a goal for the end of April. If you want to help test or have any thoughts, check out the open issues, or create a new issue if there isn't one.
HNSW indexing support for more types
- float16 vectors -
halfvec
type - early results - binary vectors -
bit
type (already in Postgres) - sparse vectors -
sparsevec
type
New functions (all of these are currently possible with SQL but less convenient)
binary_quantize
- convert a vector into a binary vectorl2_normalize
- normalize a vector to length 1subvector
- slice a vector
New operators
- Hamming distance -
<~>
- Jaccard distance -
<%>
- L1 distance -
<+>
- concatenate vectors -
||
Other changes
- HNSW
and IVFFlatindexing support for L1 distance - dynamic dispatch for Linux
- change comparison operators to not throw an error for vectors with different dimensions
- change handling of range error to be consistent with
real
Andrew Kane commented
Quick update: 0.7.0 is on-track to be released early next week (pending final testing).
Andrew Kane commented
0.7.0 is now released 🎉