agentsociety.vectorstore.vectorstore

Module Contents

Classes

VectorStore

A class for handling similarity searches and document management using Qdrant and FastEmbed.

Data

__all__

API

agentsociety.vectorstore.vectorstore.__all__

[‘VectorStore’]

class agentsociety.vectorstore.vectorstore.VectorStore(embedding: fastembed.SparseTextEmbedding)

A class for handling similarity searches and document management using Qdrant and FastEmbed.

  • Description:

    • This class provides functionalities to manage embeddings and perform similarity searches over a set of documents.

    • It uses FastEmbed for generating embeddings and Qdrant for vector storage and retrieval.

    • The class initializes with an optional embedding model and Qdrant client settings.

Initialization

Initialize the VectorStore instance.

  • Parameters:

    • embedding (SparseTextEmbedding): The embedding model to use.

property embeddings
async add_documents(documents: list[str], extra_tags: Optional[dict] = None) → list[str]

Add documents to the vector store with metadata.

  • Description:

    • Asynchronously adds one or more documents to the vector store, associating them with an agent ID and optional extra tags.

    • Each document is converted into a vector using FastEmbed before being added to Qdrant.

  • Args:

    • documents (list[str]): A list of document strings to add.

    • extra_tags (Optional[dict], optional): Additional metadata tags to associate with the documents. Defaults to None.

  • Returns:

    • list[str]: List of document IDs (UUIDs) that were added to the vector store.

async delete_documents(to_delete_ids: list[str])

Delete documents from the vector store by IDs.

  • Description:

    • Asynchronously deletes documents from the vector store based on provided document IDs.

  • Args:

    • to_delete_ids (list[str]): List of document IDs (UUIDs) to delete from the vector store.

Perform a similarity search for documents related to the given query.

  • Description:

    • Conducts an asynchronous search for the top-k documents most similar to the query text.

    • Uses FastEmbed to generate query embedding and Qdrant for vector search.

  • Args:

    • query (str): The text to look up documents similar to.

    • k (int, optional): The number of top similar contents to return. Defaults to 4.

    • fetch_k (int, optional): The number of documents to fetch before applying any filters. Defaults to 20.

    • filter (Optional[dict], optional): The filter dict for metadata.

  • Returns:

    • list[tuple[str, float, dict]]: List of tuples containing content, score, and metadata.