Faiss

Langchain vector store for FAISS. Supports save/load to disk.

Install

pip install langchain-faiss-vectorstore

or

uv add langchain-faiss-vectorstore

API Reference

__init__

def __init__(
    self,
    embedding_function: Embeddings | None = None,
    index: faiss.Index | None = None,
    **kwargs,
)

from_texts

def from_texts(
    cls,
    texts: list[str],
    embedding: Embeddings | None = None,
    metadatas: list[dict] | None = None,
    ids: list[str] | None = None,
    **kwargs,
) -> Faiss

Returns: Faiss — a new vector store with the texts indexed.

add_texts

def add_texts(
    self,
    texts: list[str],
    metadatas: list[dict] | None = None,
    ids: list[str] | None = None,
    **kwargs,
) -> list[str]

Returns: list[str] — the IDs of the added texts.

delete

def delete(
    self,
    ids: list[str] | None = None,
    **kwargs,
) -> bool | None

Returns: bool — True if deletion succeeded. Raises ValueError if ids is None.

similarity_search

def similarity_search(
    self,
    query: str,
    k: int = 4,
    **kwargs,
) -> list[Document]

Returns: list[Document] — documents most similar to the query, ordered by L2 distance (ascending).

similarity_search_with_score

def similarity_search_with_score(
    self,
    query: str,
    k: int = 4,
    **kwargs,
) -> list[tuple[Document, float]]

Returns: list[tuple[Document, float]] — tuples of (Document, L2_distance). Lower distance = more similar.

save_local

def save_local(
    self,
    folder_path: str | Path,
) -> None

Writes two files into folder_path:

• index.faiss — the serialized FAISS index.
• docs.json — document contents and metadata.

load_local

def load_local(
    cls,
    folder_path: str | Path,
    embedding_function: Embeddings,
    allow_dangerous_deserialization: bool = False,
) -> Faiss

Returns: Faiss — a restored vector store instance.