qA_Ap.app.ai.vectorstore.VectorStore

class documentation

class VectorStore:

Constructors: VectorStore.from_bytes(raw), VectorStore(documents)

A class for storing and querying document embeddings using FAISS and SentenceTransformer.

Class Method	`from_bytes`	Deserializes a VectorStore from bytes.
Method	`__init__`	Initializes the VectorStore with optional documents.
Method	`query`	Query the vector store for the most similar document chunks.
Instance Variable	`documents`	Undocumented
Instance Variable	`ids`	Undocumented
Instance Variable	`metadatas`	Undocumented
Instance Variable	`model`	Undocumented
Instance Variable	`store`	Undocumented
Property	`as_bytes`	Serializes the vector store to bytes.
Property	`metas_json`	Returns a JSON string of ids, metas, and documents.
Method	`_setup`	Splits documents into chunks of maximum 1000 characters.

@classmethod

def from_bytes(cls, raw: bytes) -> VectorStore: ¶

Deserializes a VectorStore from bytes. Args: raw (bytes): Serialized vector store. Returns: VectorStore: The deserialized instance.

def __init__(self, documents: list[dict[str, str]] = None): ¶

Initializes the VectorStore with optional documents. Args: documents (list[dict[str, str]], optional): List of documents with 'title' and 'content'.

def query(self, query: str) -> list[dict]: ¶

Query the vector store for the most similar document chunks. Args: query (str): The query string. Returns: list[dict]: List of results with document, metadata, and distance. Raises: ValueError: If the vector store is not initialized. RuntimeError: If the query fails.

documents = ¶

Undocumented

ids = ¶

Undocumented

metadatas = ¶

Undocumented

model = ¶

Undocumented

store = ¶

Undocumented

@property

as_bytes: bytes = ¶

Serializes the vector store to bytes. Returns: bytes: Serialized vector store.

@property

metas_json: str = ¶

Returns a JSON string of ids, metas, and documents. Returns: str: JSON representation.

def _setup(self, documents: list[dict[str, str]]) -> tuple[faiss.IndexFlatL2, list[str], list[str], list[dict[str, str]]]: ¶

Splits documents into chunks of maximum 1000 characters. For each chunk: create a unique ID string made of the document title and chunk index, and a metadata dict with the document title. Creates a FAISS index for the embeddings. Args: documents (list[dict[str, str]]): List of documents. Returns: tuple: (FAISS index, list of chunks, list of ids, list of metas)