Embedding Suite

A simple, unified interface for generating text embeddings from various providers. (This is heavily modeled on Andrew Ng's AI Suite, which does the same thing for Large Language Models.)

This is a work-in-progress, though is fully functional as described below. On the roadmap are new providers, an easier way to optionally also install providers' packages, perhaps use of HTTP endpoints in lieu of providers' packages, perhaps multimodal embeddings...

Installation

(You'll need to install the provider's SDK separately.)

Pip

pip install embedding-suite

Poetry

poetry add embedding-suite

UV (recommended)

uv add embedding-suite

Example Usage

inputs = ["First sentence", "Second sentence"]

You can serve API keys from your .env or explicitly add them as in the below examples.

OpenAI

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"openai": {
    "api_key": "XXX"}})

embeddings = client.generate_embeddings(
    model="openai:text-embedding-3-large", inputs=inputs)

Cohere

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"cohere": {
    "api_key": "XXX"}})

embeddings = client.generate_embeddings(
    model="cohere:embed-english-v3.0", inputs=inputs)

VoyageAI

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"voyageai": {
    "api_key": "XXX"}})

embeddings = client.generate_embeddings(
    model="voyage:voyage-3", inputs=inputs)

Sentence Transformers

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"sentencetransformers": {}})

embeddings = client.generate_embeddings(
    model="sentencetransformers:all-mpnet-base-v2", inputs=inputs)

Huggingface

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"huggingface": {
    "api_key": "XXX"}})

embeddings = client.generate_embeddings(
    model="huggingface:sentence-transformers/sentence-t5-large", inputs=inputs)

Mistral

from embedding_suite import EmbeddingSuiteClient

client = EmbeddingSuiteClient(provider_configs={"mistral": {
    "api_key": "XXX"}})

embeddings = client.generate_embeddings(
    model="mistral:mistral-embed", inputs=inputs)