Skip to content

feat: add hsg-embedding vector field for hybrid search - #102

Draft
duncdrum wants to merge 1 commit into
masterfrom
feat/hsg-embedding-vector-field
Draft

duncdrum wants to merge 1 commit into
masterfrom
feat/hsg-embedding-vector-field

Conversation

@duncdrum

Copy link
Copy Markdown
Contributor

Summary

  • Add hsg-embedding (all-MiniLM-L6-v2, 384-d, cosine, local ONNX) beside each hsg-fulltext field, and set vector-store="db".
  • The vector uses that field's expression when it has one, and the text extractor when it does not.
  • vector-field does not read if. Where the keyword field has an if test, the same test is the vector expression, so other nodes store no vector.

Test plan

  • Install on an eXist build with the vector module and reindex the collection with mode all
  • Confirm hybrid search can return hits from this package without changing keyword results

Made with Cursor

Hybrid search ranks a hit only when that node stores this vector beside
hsg-fulltext. The vector uses the keyword field's expression when it has
one. Where that field is limited by an if test, the same test gates the
vector, because vector-field does not read if.
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.

1 participant