Skip to content

Conversation

simon-dew
Copy link
Contributor

@simon-dew simon-dew commented Sep 12, 2025

Jira issue: DOC-12294

Adds all the documentation for Hyperscale Vector Indexes and Composite Vector Indexes.
This includes the following pull requests — follow the links for previous reviews and discussion:

Docs preview:

Credentials: Preview docs for internal review

Important

  • Avoid merging small edits directly into this PR. Where possible, try to use separate PRs to merge any further changes into the DOC-12294-GSI-vector-index branch.
  • Rebase and merge this PR! Do not squash and merge!

Merge at the same time as the following PR:

@simon-dew simon-dew marked this pull request as ready for review September 12, 2025 16:00
ggray-cb and others added 6 commits September 24, 2025 14:12
Full draft of the vector index concepts.
---------

Co-authored-by: Simon Dew <[email protected]>
* Add vector index syntax to EBNF
* Update script to use Railroad Diagram Generator 2.1
* Update syntax diagrams
* Update primary index functionality
* Update nav
* Update indexing overview
* Update index types
* Add ALTER VECTOR INDEX
* Add CREATE VECTOR INDEX
* Add DROP VECTOR INDEX
* Update ALTER INDEX
* Update BUILD INDEX
* Update CREATE INDEX
* Update CREATE PRIMARY INDEX
* Update DROP INDEX
* UPDATE DROP PRIMARY INDEX
* Update toc depth
* Adjusting related links
* Add composite vector index example
* Add examples to ALTER VECTOR INDEX
* Add clarification about base64 decoding
* Update CREATE INDEX examples

* DOC-12940: GSI Vector Index not supported on Windows (#406)
* Update Functions overview page
* Update nav
* Add Vector Functions page
* Include ISVECTOR in the Type functions
* Add Vector Functions
* Tweak type functions
* Add formulas
* TEMP: Preview config for feature branch
* Fix links, tags, anchors, and includes
* Tweak examples for consistency and formatting
* Move composite examples to separate files — use the View button as intended
… for L2 and Euclidean (#422)

* Added note to Euclidean Distance to explain that Couchbase uses Euclidean Squared Distance instead behind the scenes.
* Changed "FTS Vector Index" to "Search Vector Index" which is PM's current nomenclature.

* Fixing a typo spotted by Nischal
* Use sequence diagrams for clearer readability
* Minor wording tweaks
* Apply Vale suggestions for DDL and DML
* Update Search Vector workflow diagram
* Rename Vector Indexes top-level topic
* Rename Choose the Right Vector Index topic
* An Search Vector → a Search Vector
* Make vector index topic names consistent
* Simplify xrefs in nav
@simon-dew simon-dew requested a review from a team October 7, 2025 21:10
@ksaikrishnateja
Copy link

Is it possible to have dedicated page for Composite Vector index under Create Index Tab like that of Array Index and Partitioned Index

@kaminijagtiani
Copy link

@simon-dew
We would need to update the section on optimizer as well and add some context around how the appropriate index would be selected.

@ksaikrishnateja
Copy link

persist_full_vector needs to be defined in the with nodes clause for Create Vector Index its only in the example. We will also need information on how to enable reranking at scan time.

@simon-dew
Copy link
Contributor Author

persist_full_vector needs to be defined in the with nodes clause for Create Vector Index its only in the example. We will also need information on how to enable reranking at scan time.

Fixed

@simon-dew
Copy link
Contributor Author

We would need to update the section on optimizer as well and add some context around how the appropriate index would be selected.

This is now mentioned on both the Hyperscale Vector Indexes page and the Composite Vector Indexes page, and covered in detail in the documentation on the APPROX_VECTOR_DISTANCE() function.

@simon-dew
Copy link
Contributor Author

Is it possible to have dedicated page for Composite Vector index under Create Index Tab like that of Array Index and Partitioned Index

I did consider this, but creating a Composite Vector index is so similar to creating a secondary index that I decided to document them together in the CREATE INDEX page.

* Explain how the appropriate index is selected
* Explain all ANN_DISTANCE arguments in Hyperscale Vector index guide
* Index Service not Query Service
* ORDER BY not GROUP BY
* Similarity is optional
* Description is optional
* Composite Vector index with leading scalar key
* HVI and CVI are supported on Windows
* Include persist_full_vector
* Remove trailing spaces
* Scan for typos
* Capitalization of Hyperscale Vector and Composite Vector
* Examples on this page
* Remove feature branch preview config
@simon-dew
Copy link
Contributor Author

All comments addressed. I'm now merging this into the release/8.0 branch. Any further issues can be addressed there before GA.

@simon-dew simon-dew merged commit 3a6c4d4 into release/8.0 Oct 13, 2025
6 checks passed
@simon-dew simon-dew deleted the DOC-12294-GSI-vector-index branch October 13, 2025 11:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants