Skip to main content

Vector Store API Reference

VectorStore (ABC)​

All vector stores implement this interface.

from synapsekit.retrieval.base import VectorStore

class VectorStore(ABC):
    async def add(
        self,
        texts: list[str],
        metadata: list[dict] | None = None,
    ) -> None: ...

    async def search(
        self,
        query: str,
        top_k: int = 5,
    ) -> list[dict]: ...

    def save(self, path: str) -> None: ...   # optional — raises NotImplementedError if unsupported
    def load(self, path: str) -> None: ...   # optional — raises NotImplementedError if unsupported

search() returns a list of dicts: {"text": str, "score": float, "metadata": dict}.

InMemoryVectorStore​

from synapsekit import InMemoryVectorStore, SynapsekitEmbeddings

InMemoryVectorStore(embedding_backend: SynapsekitEmbeddings)
MethodDescription
add(texts, metadata=None)Embed and store texts
search(query, top_k=5)Cosine similarity search (L2-normalised vectors)
save(path)Persist to .npz file
load(path)Load from .npz file
__len__()Number of stored texts

ChromaVectorStore​

from synapsekit.retrieval.chroma import ChromaVectorStore

ChromaVectorStore(
    embedding_backend: SynapsekitEmbeddings,
    collection_name: str = "synapsekit",
    persist_directory: str | None = None,
)

Pass persist_directory for on-disk persistence.

FAISSVectorStore​

from synapsekit.retrieval.faiss import FAISSVectorStore

FAISSVectorStore(embedding_backend: SynapsekitEmbeddings)
MethodDescription
save(path)Writes path.faiss, path_texts.npy, path_meta.json
load(path)Loads the three files above

QdrantVectorStore​

from synapsekit.retrieval.qdrant import QdrantVectorStore

QdrantVectorStore(
    embedding_backend: SynapsekitEmbeddings,
    collection_name: str = "synapsekit",
    url: str = "http://localhost:6333",
    api_key: str | None = None,
)

Collection is auto-created on first add().

PineconeVectorStore​

from synapsekit.retrieval.pinecone import PineconeVectorStore

PineconeVectorStore(
    embedding_backend: SynapsekitEmbeddings,
    index_name: str,
    api_key: str,
    environment: str = "us-east-1",
)

Index must exist in your Pinecone project before calling add().

SynapsekitEmbeddings​

from synapsekit import SynapsekitEmbeddings

SynapsekitEmbeddings(model: str = "all-MiniLM-L6-v2")

Uses sentence-transformers under the hood (lazy import).

MethodReturnsDescription
embed(texts)np.ndarray (N, D)Embed a list of texts
embed_one(text)np.ndarray (D,)Embed a single text