RAG Pipeline: Embedding Search + LLM Inference

Scenario

Deploy an embedding model in OpenSearch for vector similarity retrieval, then deploy a large language model via PAI for online inference, combining both into a retrieval-augmented generation (RAG) pipeline.

How the products combine

1. pai · pai-deploy-inference — Platform for AI (PAI) — Deploy a model for online inference

See pai/pai-deploy-inference.

2. opensearch · opensearch-deploy-model — OpenSearch — Deploy embedding model for inference

See opensearch/opensearch-deploy-model.

Typical questions

• build RAG system
• deploy RAG pipeline
• embedding search plus LLM
• vector search and model inference
• retrieval augmented generation deploy
• 部署RAG系统
• 向量检索加大模型推理
• 构建检索增强生成

Q: How do I build and deploy a retrieval-augmented generation pipeline using embedding search and LLM inference? A: You can deploy a retrieval-augmented generation pipeline by combining an embedding model in OpenSearch for vector similarity retrieval with a large language model on PAI for online inference. This cross-product integration links vector search directly to model inference to create a complete RAG workflow.