Vector Search

HNSW-based approximate nearest neighbor search.

# Create a vector index
db.create_vector_index("items", "embedding", dimension=384, metric="cosine")

# Insert documents with embeddings
db.insert("items", {
    "title": "Introduction to Rust",
    "embedding": [0.1, 0.5, -0.3, ...]  # 384-dim vector
})

# Search for similar items
results = db.vector_search("items", "embedding", query_vector, limit=10)

Distance Metrics

Metric	Formula	Range
`cosine`	1 - cos(a, b)	[0, 2]
`euclidean`	L2 distance	[0, inf)
`dot`	-dot(a, b)	(-inf, inf)

Results include document ID, raw distance, and a normalized similarity score in [0, 1].