Feather overview

Feather is the easiest way to start building vector search and applications powered by embeddings. With under 10 lines of code, you can connect to OpenAI, Anthropic, local models, and more. Feather provides a pre-built vector database architecture and model integrations to help you get started quickly and seamlessly incorporate vector search into your applications.

We recommend you use Feather if you want to quickly build vector search applications. Use Feather Core, our low-level vector database framework, when you have more advanced needs that require a combination of deterministic and vector workflows, heavy customization, and carefully controlled latency.

Feather databases are built on top of HNSW indexing in order to provide fast search, persistence, quantization, and more. You do not need to know HNSW for basic Feather usage.

Install

pip install -U feather-py
# Requires Python 3.10+

Create a vector database

from feather import DB
import numpy as np

# Create a new database
db = DB.open("vectors.feather", dim=128)

# Add vectors
vector = np.random.rand(128).astype(np.float32)
db.add(1, vector)

# Search
query = np.random.rand(128).astype(np.float32)
results = db.search(query, k=5)

# Save
db.save()

Core benefits