---
title: "Discussion on \"Building a Production RAG Pipeline on AWS Lambda and pgvector\" | Hashnode"
description: "Discussion on \"Building a Production RAG Pipeline on AWS Lambda and pgvector\". How to store embeddings in PostgreSQL with pgvector, run retrieval on Lambda, and wire it all together with Bedrock or OpenAI at near-zero upfront cost"
url: "https://hashnode.com/posts/building-a-production-rag-pipeline-on-aws-lambda-pgvector/69d0d31df5bdeea1546a34b1"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Building a Production RAG Pipeline on AWS Lambda and pgvector](https://blog.harun.dev/building-a-production-rag-pipeline-on-aws-lambda-pgvector?utm_source=hashnode&utm_medium=feed)

How to store embeddings in PostgreSQL with pgvector, run retrieval on Lambda, and wire it all together with Bedrock or OpenAI at near-zero upfront cost

Apr 4 · 22 min read · blog.harun.dev  
Read the article

-   [#aws](https://hashnode.com/tag/aws)
-   [#serverless](https://hashnode.com/tag/serverless)
-   [#ai](https://hashnode.com/tag/ai)
-   [#postgresql](https://hashnode.com/tag/postgresql)
-   [#nodejs](https://hashnode.com/tag/nodejs)

## Written by

[Harun R Rayhan](https://hashnode.com/@HarunRRayhan)

@HarunRRayhan

  
[View profile](https://hashnode.com/@HarunRRayhan)

Senior Software Engineer | AWS Community Builder | Cloud Consultant | Serverless Enthusiast | 10x AWS & Laravel Certified | CSM®, CSD®

He is a Senior Software Engineer. An AWS Community Builder and 10x AWS Certified. He loves to work on Laravel and is a Certified Laravel Developer. He is an agile/Scrum practitioner with CSM® &amp; CSD®. He mostly writes about AWS, Serverless, Container, Laravel, SRE, and DevOps in General. You can follow him on Social media.

## Responses 1

Markdown supported · ⌘↵ to post

[CapeStart](https://hashnode.com/@capestart)

[Apr 4](https://hashnode.com/posts/building-a-production-rag-pipeline-on-aws-lambda-pgvector/69d0d31df5bdeea1546a34b1/comment/69d1153f6792e486f6700c02)

This is a great reminder that RAG is more about retrieval quality than fancy models.
