Capabilities

• Dense near-vector search using precomputed embeddings.
• Weaviate v3 collections API integration.
• Batch object ingestion with deterministic UUIDs.
• Cosine, dot, or L2-squared distance configuration.
• Equality, comparison, and, and or filter translation.
• Multi-embedding logical result collapsing.
• search plus createVectorSearchTool() tool integration from @anvia/core/vector-store.

The adapter creates only baseline Anvia properties. It does not manage collection aliases, replication, tenancy, property indexing, backups, or data deletion. close() closes the adapter-created client; injected clients are left open.

Batch ingestion uses replacement semantics: upsert() deletes existing objects by __anvia_document_id before inserting, so re-upserting a document replaces its physical embeddings. See API reference.