opea/erag-torchserve_embedding repository overview

⁠OPEA ERAG TorchServe Embedding Model Server

Part of the Intel® AI for Enterprise RAG (ERAG) ecosystem.

⁠🔍 Overview

The OPEA ERAG TorchServe Embedding Model Server hosts embedding models using TorchServe, providing a scalable and efficient endpoint for generating vector embeddings from text or documents. It serves as the backend for the ERAG Embedding Microservice.

TorchServe⁠ is a lightweight, scalable, and easy-to-use model serving library for PyTorch models. It provides a RESTful API for serving trained models, allowing users to deploy and serve their models in production environments. Moreover, Torchserve supports Intel® Extension for PyTorch*⁠ for a performance boost on Intel-based Hardware.

⁠🔗 Related Components

This service integrates with other OPEA ERAG components:

• OPEA ERAG Embedding Microservice sends requests to this model server to obtain embeddings
• Retriever & Reranker Microservices use embeddings generated here for improved search relevance

⁠License

OPEA ERAG is licensed under the Apache License, Version 2.0.

Copyright © 2024–2026 Intel Corporation. All rights reserved.