MelowRAG

MelowRAG is a modular Python framework for semantic search, vector indexing, and retrieval-augmented generation (RAG). It provides a unified interface for embedding, indexing, searching, and managing data using dense, sparse, and hybrid vector models.

Features

• Embeddings Management: Transform data into embeddings using various backends.
• Flexible Indexing: Build, update, and search indexes with support for dense, sparse, and hybrid models.
• Database Integration: Store and retrieve content using pluggable database backends.
• Graph Algorithms: Advanced graph-based search and topic modeling.
• Pipelines: Modular pipelines for text, audio, and image processing.
• Remote Storage: Archive and load indexes from local or cloud storage.
• Extensible: Easily add new models, scoring functions, or storage backends.

Quick Start

from melowrag import Embeddings

# Initialize embeddings
embedding = Embeddings()

# Index some texts
texts = ["The cat sat on the mat.", "Dogs are wonderful companions."]
embedding.index(texts)

# Search for similar content
results = embedding.search("animal companions", 1)
for result in results:
    print(f"Index: {result.index}, Score: {result.score}")

Installation

pip install -e .

License

This project is licensed under the terms of the MIT license.