Setting the file. One moment.
Subchapter 91.16
references/vector-api-migration.mdMarkdown1 KBView on GitHub
Use vectors.topK.byEmbedding with a single configuration map for
embedding-based vector similarity queries:
CALL neptune.algo.vectors.topK.byEmbedding({
embedding: $embedding,
topK: $top_k
})
YIELD node, score
RETURN node, scoreThe deprecated procedure did not expose vertexFilter. When migrating a
label-scoped search, move any post-YIELD label predicate into the supported
procedure so the algorithm selects topK results from the intended label:
CALL neptune.algo.vectors.topK.byEmbedding({
embedding: $embedding,
topK: $top_k,
vertexFilter: '{"equals":{"property":"~label","value":"Chunk"}}'
})
YIELD node, score
RETURN node, scoreDo not apply the label filter with WHERE after YIELD. That filters an
already-selected global topK result and can return fewer matching vertices
than requested.
| Old (deprecated) | Replacement |
|---|---|
vectors.topKByEmbedding($embedding, {topK: $top_k}) | vectors.topK.byEmbedding({embedding: $embedding, topK: $top_k}) |
vectors.upsert(node, embedding) is unchanged.