hybrid semantic search: embeddings + KNN + RRF fusion (#1) - #33

Open

arreyder wants to merge 9 commits into

mainfrom

feat/semantic-search

Commits on Jun 10, 2026

hybrid semantic search: embeddings + KNN + RRF fusion (#1)

arreyder
and
claude
committed
d1333b2
embed: truncate input to model context limit (EMBED_MAX_CHARS, default 6000)

arreyder
and
claude
committed
ce9c731
knn: force defType=lucene so {!knn} isn't parsed as edismax text

arreyder
and
claude
committed
8dffd4c
fuse: NumFound reflects returned docs (semantic-only hits aren't in lexical count)

arreyder
and
claude
committed
e5e5173
embed: nomic task prefixes (search_document/search_query) for asymmetric retrieval

arreyder
and
claude
committed
5294b3a
semantic search: upgrade to mxbai-embed-large (1024d) via new embedding1024 field

arreyder
and
claude
committed
e4c62ec
revert to nomic-embed-text: mxbai's 512-tok context truncated our corpus too hard

arreyder
and
claude
committed
5020757

Commits on Aug 13, 2026

fix: stop EnsureCollection from destroying a corrupt core

arreyder
and
claude
committed
fa38b5a

Commits on Aug 17, 2026

fix(parser): resolve physical line numbers, not //line-adjusted ones

arreyder
and
claude
committed
bb782b2