agrag.vectordb

External vector indexes with graph-owned visibility and placement.

Modules:

• base – The asynchronous contract for an exclusive external vector index.
• errors – Errors exposed by external vector indexes.
• kinds – Map public graph labels to external collection kinds.
• milvus – Milvus dense and native BM25 indexes for external vector placement.
• placement – Validate that a graph continues to use its original vector placement.
• qdrant – Qdrant indexes with dense vectors and locally computed BM25 sparse vectors.
• settings – Validation shared by external vector store settings.
• weaviate – Weaviate collections with self-provided vectors and native BM25 fusion.

Classes:

• CollectionMismatchError – An existing collection has incompatible dimensions or distance.
• MilvusSettings – Milvus or Zilliz Cloud connection and collection namespace.
• MilvusVectorStore – Store UUID-keyed vectors with server-generated BM25 sparse vectors.
• QdrantSettings – Qdrant connection and per-graph collection names.
• QdrantVectorStore – An exclusive vector index on Qdrant 1.17 or newer.
• VectorPlacementError – The requested vector placement disagrees with the graph's recorded placement.
• VectorStore – An external index whose hits must be joined to visible graph nodes.
• VectorStoreError – An external vector index operation failed.
• VectorStoreMissingExtraError – The selected vector store's optional client is not installed.
• WeaviateSettings – Weaviate connection and collection isolation settings.
• WeaviateVectorStore – An exclusive external index backed by Weaviate.

Functions:

• build_vector_store – Construct a named vector index or return an existing instance unchanged.
• validate_placement – Compare placement, creating it on first open only when write is true.

Attributes:

• VectorStoreName –