Skip to main content

VectorStore

What it is​

  • A minimal in-memory vector store backed by Qdrant (using QdrantClient(":memory:")).
  • Supports inserting text + metadata with pre-computed embeddings and running cosine similarity search.

Public API​

class VectorStore​

In-memory Qdrant collection wrapper.

__init__(dimension: int = 1536)​

  • Creates an in-memory Qdrant client and a collection named "documents".
  • Configures vectors with:
    • size = dimension
    • distance = COSINE

add_texts(texts, metadatas=None, embeddings=None) -> List[str]​

  • Inserts points into the "documents" collection.
  • Payload stored per point:
    • "text" plus any metadata keys.
  • Returns:
    • List of string IDs assigned sequentially starting at 0.
  • Requires:
    • embeddings must be provided (otherwise raises ValueError).

similarity_search(query_embedding, k: int = 4) -> List[dict]​

  • Runs a similarity query against the stored vectors.
  • Returns a list of dicts:
    • text: stored text
    • metadata: payload keys excluding "text"
    • score: Qdrant similarity score

Configuration/Dependencies​

  • Requires qdrant-client:
    • pip install qdrant-client
  • Uses an in-memory Qdrant instance (QdrantClient(":memory:")), so data is not persisted.
  • Collection name is fixed to "documents".

Usage​

from naas_abi_core.services.agent.beta.VectorStore import VectorStore

# Create store (dimension must match your embeddings)
store = VectorStore(dimension=3)

texts = ["hello world", "goodbye world"]
metadatas = [{"source": "a"}, {"source": "b"}]
embeddings = [
    [1.0, 0.0, 0.0],
    [0.9, 0.1, 0.0],
]

ids = store.add_texts(texts=texts, metadatas=metadatas, embeddings=embeddings)

results = store.similarity_search(query_embedding=[1.0, 0.0, 0.0], k=2)
print(ids)
print(results)

Caveats​

  • add_texts does not compute embeddings; you must provide embeddings.
  • No explicit validation that embedding lengths match dimension.
  • Data is ephemeral due to in-memory Qdrant usage.
  • ID generation is local and sequential; IDs reset when a new VectorStore is created.