Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
7.6.0
-
Security Level: Public
-
None
Description
Vector search requests can be used directly by embedding the search request into N1QL's SEARCH() function.
The ask here is to go a little beyond that for on boarding users easily with the a CREATE VECTOR INDEX and a KNN() search function so then can skip coming up to speed on the search index definition syntaxes.
- Introduce CREATE INDEX capability to build vector indexes in the search space -
CREATE [VECTOR] INDEX `idx` ON `key-space`({"field": expression, "dimension": constant, "similarity": string}, {"field": expression, "dimension": constant, "similarity": string}) USING FTS;
- Add a KNN function as an added convenience that would spare users the SEARCH() function syntax -
SELECT * FROM `key-space` WHERE KNN(`key-space`, {object}, {object [optional]})
Attachments
Issue Links
- relates to
-
MB-48298 N1QL's `CREATE INDEX` command to setup FTS indexes
- Open
Gerrit Reviews
For Gerrit Dashboard: MB-58441 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
199290,9 | MB-58441. FTS Index creation | master | query | Status: ABANDONED | 0 | 0 |
199481,4 | MB-58441. FTS index | master | query | Status: MERGED | +2 | +1 |