PyVectorDB

Born to be simple. Simple Python wrapper provides efficient support for CRUD operations and querying with vector databases.".

---

🚀 Getting Started

Installation

To install all vector database support depedencies (require a lot of disk space, not recommended)

pip install pyvectordb[all]

if you only need a specific vector database engine, you can use (recommended)

pip install pyvectordb[pgvector]

pip install pyvectordb[qdrant]

pip install pyvectordb[chromadb]

Usage examples

1. PGVector

PGvector is an extension for PostgreSQL that allows the storage, indexing, and querying of vector embeddings. It is designed to support vector similarity search, which is useful in machine learning applications like natural language processing, image recognition, and recommendation systems. By storing vector embeddings as a data type, PGvector enables efficient similarity searches using distance metrics such as cosine similarity, Euclidean distance, inner product, etc.

from dotenv import load_dotenv
load_dotenv()

import os
from pyvectordb import Vector
from pyvectordb.pgvector.pgvector import PgvectorDB
from pyvectordb.distance_function import DistanceFunction

v1 = Vector(
    embedding=[2., 2., 1.],
    metadata={"text": "hellow from pyvectordb"}
)
v2 = Vector(
    embedding=[2., 2., 2.],
    metadata={"text": "hi"}
)
v3 = Vector(
    embedding=[2., 2., 3.],
    metadata={"text": "good morning!"}
)

vector_db = PgvectorDB(
    user=os.getenv("PG_USER"),
    password=os.getenv("PG_PASSWORD"),
    host=os.getenv("PG_HOST"),
    port=os.getenv("PG_PORT"),
    db_name=os.getenv("PG_NAME"),
    collection=os.getenv("PG_COLLECTION"),
    distance_function=DistanceFunction.L2,
)

# insert new vector
vector_db.insert_vector(v1)
vector_db.insert_vectors([v2, v3])

# read v1
v_from_db = vector_db.read_vector(v1.get_id())

# update v1 embedding
new_embedding = [2., 2., 4.]
v_from_db.embedding = new_embedding
vector_db.update_vector(v_from_db)

# read updated embedding and check
v_from_db_updated = vector_db.read_vector(v1.get_id())
assert list(v_from_db_updated.embedding) == list(new_embedding), "updated embedding not equal"

# re-update v1 embedding to the v1, check
vector_db.update_vectors([v1, v2, v3])
re_updated_embedding = vector_db.read_vector(v1.get_id()).embedding
assert list(re_updated_embedding) == list(v1.embedding), "re-updated embedding not equal"

for x in vector_db.get_neighbor_vectors(v1, 3):
    print(f"{x}")

vector_db.delete_vector(v1.get_id())
vector_db.delete_vectors([v2, v3])

2. Qdrant

Qdrant “is a vector similarity search engine that provides a production-ready service with a convenient API to store, search, and manage points (i.e. vectors) with an additional payload.” You can think of the payloads as additional pieces of information that can help you hone in on your search and also receive useful information that you can give to your users.

Using Qdrant in pyvectordb is simple, you only need to change the client to QdrantDB

from pyvectordb import QdrantDB

vector_db = QdrantDB(
    host=os.getenv("Q_HOST"),
    api_key=os.getenv("Q_API_KEY"),
    port=os.getenv("Q_PORT"),
    collection=os.getenv("Q_COLLECTION"),
    vector_size=int(os.getenv("Q_SIZE")),
    distance_function=DistanceFunction.COSINE,
)

3. Chroma DB

Chroma is the AI-native open-source vector database. Chroma makes it easy to build LLM apps by making knowledge, facts, and skills pluggable for LLMs.

from pyvectordb import ChromaDB

vector_db = ChromaDB(
    host=os.getenv("CH_HOST"),
    port=os.getenv("CH_PORT"),
    auth_provider=os.getenv("CH_AUTH_PROVIDER"),
    auth_credentials=os.getenv("CH_AUTH_CREDENTIALS"),
    collection_name=os.getenv("CH_COLLECTION_NAME"),
    distance_function=DistanceFunction.L2,
)

Available functions

These are available functions in this simple tool

def insert_vector(self, vector: Vector) -> None: ...
def insert_vectors(self, vectors: List[Vector]) -> None: ...
def read_vector(self, id: str) -> Vector | None: ...
def update_vector(self, vector: Vector) -> None: ...
def update_vectors(self, vectors: List[Vector]) -> None: ...
def delete_vector(self, id: str) -> None: ...
def delete_vectors(self, ids: Union[List[str], List[Vector]]) -> None: ...
def get_neighbor_vectors(self, vector: Vector, n: int) -> List[VectorDistance]: ...

---

💬 Support & Contact

If you have any questions, feedback, or need support, feel free to reach out:

📧 Email: My Email
🌐 GitHub Issues: Submit an Issue
💼 LinkedIn: LinkedIn Profile

---

🙏 Support the Project

If you find this project helpful, consider supporting it by:

• ⭐ Starring this repository
• 🍴 Forking the project and contributing
• 🗨 Sharing your feedback or feature requests

Every contribution helps make the project better!