PostgreSQL pgvector Vector Database

This is a repackaged open source software product wherein additional charges apply for cloudimg support services.

## Overview

PostgreSQL with the pgvector extension turns a proven SQL database into a first-class vector store. Alongside ordinary tables you store high-dimensional embeddings in a dedicated vector column, build approximate nearest-neighbour indexes, and run similarity queries that rank rows by distance to a query vector - all while keeping transactions, joins, SQL and the entire PostgreSQL driver and tooling ecosystem. This AMI delivers PostgreSQL 17 with pgvector fully installed and enabled as a system service, so a production-ready vector database is running within minutes of launch. The extension is pre-enabled in a ready-to-use database named vectordb, seeded with a small demo table and an HNSW index so you can run your first nearest-neighbour query immediately.

## Why This AMI Over DIY or Unsupported Images

Standing PostgreSQL and pgvector up by hand on EC2 means adding an APT repository, compiling or locating the matching extension build, configuring TLS, hardening authentication and writing systemd units - hours of work that introduces security risk if done incorrectly. Many community images ship with a shared or blank database password baked into the image. This AMI eliminates that effort and risk: there is no default password to change. On first boot each instance generates its own PostgreSQL password and its own TLS certificate, written to a file only the root user can read, and remote connections are accepted only over TLS. You get a hardened, production-grade deployment with unique-per-instance credentials, a dedicated data volume and automatic service recovery - all backed by 24/7 expert support from cloudimg.

## Application Stack

PostgreSQL 17 and the matching postgresql-17-pgvector package are installed from the official PostgreSQL PGDG APT repository (GPG-verified). The database listens on port 5432 and is managed by systemd, which starts it on boot and restarts it on failure. The cluster data directory lives on a dedicated, independently resizable EBS volume mounted at /var/lib/postgresql, kept separate from the operating system disk and snapshotted with the AMI.

## Secure By Default

There is no shared or default credential in this image. The postgres superuser role ships password-less (usable only via the local Unix socket for on-box administration), TLS is off and no server certificate is baked in. On its first boot the instance mints a per-instance self-signed TLS certificate, enables TLS, generates a unique password for the postgres role and writes it to a root-only file. Remote clients are permitted only over TLS (hostssl) authenticated with scram-sha-256; non-TLS remote connections are rejected. Restrict the source range on port 5432 in your security group to your application tier.

## Ready To Use

Retrieve your per-instance password with a single command, then connect from any PostgreSQL client or driver over TLS to the vectordb database. Create a table with a vector column, add your embeddings, build an HNSW or IVFFlat index, and run ORDER BY embedding distance queries for nearest-neighbour search. The image ships no embedding model and is CPU-only - bring your own embeddings or generate them from your preferred model.

## AWS Integrations

• Amazon S3 - Back up your PostgreSQL data volume or logical dumps to S3 for durable, cost-effective storage and disaster recovery.
• Amazon CloudWatch - Monitor instance health and database logs with the CloudWatch agent and alarms.
• Amazon Bedrock - Generate embeddings with Bedrock foundation models and store them directly in pgvector for retrieval-augmented generation.
• LangChain and LlamaIndex - Use PostgreSQL with pgvector as the vector store in popular orchestration frameworks to build end-to-end RAG pipelines.

## Use Cases

• Retrieval-augmented generation - Store embeddings of your own documents and retrieve the most relevant passages at query time to ground an LLM, keeping your data inside your own VPC.
• Semantic and hybrid search - Combine keyword filters in SQL with vector similarity to rank results by meaning, not just exact matches.
• Recommendations and deduplication - Rank items by embedding similarity for recommendation, near-duplicate detection and clustering, next to your relational data.
• Data residency - A self-hosted, in-your-own-VPC vector database for teams with compliance or data-residency requirements.

## cloudimg Support

24/7 technical support by email and live chat. Our engineers help with PostgreSQL and pgvector deployment, index selection (HNSW vs IVFFlat), embedding workflows, backups, TLS and connection configuration, scaling and AWS integration.

All product and company names are trademarks or registered trademarks of their respective holders. Use of them does not imply any affiliation with or endorsement by them.