Welcome to Catsu

High-performance embeddings client for multiple providers, powered by Rust.

Catsu is a unified client for accessing multiple embedding providers through a single, consistent API. It supports 11 major embedding providers with over 60 models, providing:

• Unified Interface: Single API for all providers
• High Performance: Rust core with Python bindings
• Automatic Retry Logic: Built-in exponential backoff
• Cost Tracking: Automatic usage and cost calculation
• Async Support: Both sync and async methods
• NumPy Integration: Convert embeddings to numpy arrays

Quick Example

from catsu import Client

# Initialize the client (reads API keys from environment)
client = Client()

# Generate embeddings
response = client.embed(
    "openai:text-embedding-3-small",
    ["Hello, embeddings!"]
)

# Access your embeddings
embedding = response.embeddings[0]  # List of floats
print(f"Embedding: {embedding[:5]}...")  # First 5 dimensions
print(f"Dimensions: {response.dimensions}")

# Check usage
print(f"Tokens used: {response.usage.tokens}")

Get Started

Why Catsu?

The world of embedding API clients is broken. Provider-specific libraries are inconsistent or broken, universal clients don't focus on embeddings, and there's no single source of truth for model metadata. Catsu fixes this with a clean, consistent interface across all major providers while handling the complexity of retries, error handling, and cost tracking for you.

Ready to get started? Head to the Installation page!